sync with bigtop master'

author: ganeshraju <ganeshraju@gmail.com> 2016-06-21 16:10:43 -0500
committer: ganeshraju <ganeshraju@gmail.com> 2016-06-21 16:10:43 -0500
commit: 9d62f51ec3684c75d0fa4b8d2da9301e0a18f661 (patch)
tree: 2a255c6f11a6358f76e701b1357ee8a758d18998 /bigtop-deploy
parent: cff413decf877c03c670de67e002bf1d6d15a5fc (diff)
44 files changed, 2947 insertions, 204 deletions
diff --git a/bigtop-deploy/juju/hadoop-processing/.gitignore b/bigtop-deploy/juju/hadoop-processing/.gitignore
new file mode 100644
index 00000000..a295864e
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/.gitignore
@@ -0,0 +1,2 @@
+*.pyc
+__pycache__
diff --git a/bigtop-deploy/juju/hadoop-processing/README.md b/bigtop-deploy/juju/hadoop-processing/README.md
new file mode 100644
index 00000000..69425771
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/README.md
@@ -0,0 +1,194 @@
+<!--
+  Licensed to the Apache Software Foundation (ASF) under one or more
+  contributor license agreements.  See the NOTICE file distributed with
+  this work for additional information regarding copyright ownership.
+  The ASF licenses this file to You under the Apache License, Version 2.0
+  (the "License"); you may not use this file except in compliance with
+  the License.  You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+  Unless required by applicable law or agreed to in writing, software
+  distributed under the License is distributed on an "AS IS" BASIS,
+  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+  See the License for the specific language governing permissions and
+  limitations under the License.
+-->
+## Overview
+
+The Apache Hadoop software library is a framework that allows for the
+distributed processing of large data sets across clusters of computers
+using a simple programming model.
+
+It is designed to scale up from single servers to thousands of machines,
+each offering local computation and storage. Rather than rely on hardware
+to deliver high-avaiability, the library itself is designed to detect
+and handle failures at the application layer, so delivering a
+highly-availabile service on top of a cluster of computers, each of
+which may be prone to failures.
+
+This bundle provides a complete deployment of the core components of the
+[Apache Bigtop](http://bigtop.apache.org/)
+platform to perform distributed data analytics at scale.  These components
+include:
+
+  * NameNode (HDFS)
+  * ResourceManager (YARN)
+  * Slaves (DataNode and NodeManager)
+  * Client (Bigtop hadoop client)
+    * Plugin (subordinate cluster facilitator)
+
+Deploying this bundle gives you a fully configured and connected Apache Bigtop
+cluster on any supported cloud, which can be easily scaled to meet workload
+demands.
+
+
+## Deploying this bundle
+
+In this deployment, the aforementioned components are deployed on separate
+units. To deploy this bundle, simply use:
+
+    juju deploy hadoop-processing
+
+This will deploy this bundle and all the charms from the [charm store][].
+
+> Note: With Juju versions < 2.0, you will need to use [juju-deployer][] to
+deploy the bundle.
+
+You can also build all of the charms from their source layers in the
+[Bigtop repository][].  See the [charm package README][] for instructions
+to build and deploy the charms.
+
+The default bundle deploys three slave nodes and one node of each of
+the other services. To scale the cluster, use:
+
+    juju add-unit slave -n 2
+
+This will add two additional slave nodes, for a total of five.
+
+[charm store]: https://jujucharms.com/
+[Bigtop repository]: https://github.com/apache/bigtop
+[charm package README]: ../../../bigtop-packages/src/charm/README.md
+[juju-deployer]: https://pypi.python.org/pypi/juju-deployer/
+
+
+## Status and Smoke Test
+
+The services provide extended status reporting to indicate when they are ready:
+
+    juju status --format=tabular
+
+This is particularly useful when combined with `watch` to track the on-going
+progress of the deployment:
+
+    watch -n 0.5 juju status --format=tabular
+
+The charms for each master component (namenode, resourcemanager)
+also each provide a `smoke-test` action that can be used to verify that each
+component is functioning as expected.  You can run them all and then watch the
+action status list:
+
+    juju action do namenode/0 smoke-test
+    juju action do resourcemanager/0 smoke-test
+    watch -n 0.5 juju action status
+
+Eventually, all of the actions should settle to `status: completed`.  If
+any go instead to `status: failed` then it means that component is not working
+as expected.  You can get more information about that component's smoke test:
+
+    juju action fetch <action-id>
+
+
+## Monitoring
+
+This bundle includes Ganglia for system-level monitoring of the namenode,
+resourcemanager, and slave units. Metrics are sent to a central
+ganglia unit for easy viewing in a browser. To view the ganglia web interface,
+first expose the service:
+
+    juju expose ganglia
+
+Now find the ganglia public IP address:
+
+    juju status ganglia
+
+The ganglia web interface will be available at:
+
+    http://GANGLIA_PUBLIC_IP/ganglia
+
+
+## Benchmarking
+
+This charm provides several benchmarks to gauge the performance of your
+environment.
+
+The easiest way to run the benchmarks on this service is to relate it to the
+[Benchmark GUI][].  You will likely also want to relate it to the
+[Benchmark Collector][] to have machine-level information collected during the
+benchmark, for a more complete picture of how the machine performed.
+
+[Benchmark GUI]: https://jujucharms.com/benchmark-gui/
+[Benchmark Collector]: https://jujucharms.com/benchmark-collector/
+
+However, each benchmark is also an action that can be called manually:
+
+        $ juju action do resourcemanager/0 nnbench
+        Action queued with id: 55887b40-116c-4020-8b35-1e28a54cc622
+        $ juju action fetch --wait 0 55887b40-116c-4020-8b35-1e28a54cc622
+
+        results:
+          meta:
+            composite:
+              direction: asc
+              units: secs
+              value: "128"
+            start: 2016-02-04T14:55:39Z
+            stop: 2016-02-04T14:57:47Z
+          results:
+            raw: '{"BAD_ID": "0", "FILE: Number of read operations": "0", "Reduce input groups":
+              "8", "Reduce input records": "95", "Map output bytes": "1823", "Map input records":
+              "12", "Combine input records": "0", "HDFS: Number of bytes read": "18635", "FILE:
+              Number of bytes written": "32999982", "HDFS: Number of write operations": "330",
+              "Combine output records": "0", "Total committed heap usage (bytes)": "3144749056",
+              "Bytes Written": "164", "WRONG_LENGTH": "0", "Failed Shuffles": "0", "FILE:
+              Number of bytes read": "27879457", "WRONG_MAP": "0", "Spilled Records": "190",
+              "Merged Map outputs": "72", "HDFS: Number of large read operations": "0", "Reduce
+              shuffle bytes": "2445", "FILE: Number of large read operations": "0", "Map output
+              materialized bytes": "2445", "IO_ERROR": "0", "CONNECTION": "0", "HDFS: Number
+              of read operations": "567", "Map output records": "95", "Reduce output records":
+              "8", "WRONG_REDUCE": "0", "HDFS: Number of bytes written": "27412", "GC time
+              elapsed (ms)": "603", "Input split bytes": "1610", "Shuffled Maps ": "72", "FILE:
+              Number of write operations": "0", "Bytes Read": "1490"}'
+        status: completed
+        timing:
+          completed: 2016-02-04 14:57:48 +0000 UTC
+          enqueued: 2016-02-04 14:55:14 +0000 UTC
+          started: 2016-02-04 14:55:27 +0000 UTC
+
+
+## Deploying in Network-Restricted Environments
+
+Charms can be deployed in environments with limited network access. To deploy
+in this environment, you will need a local mirror to serve required packages.
+
+
+### Mirroring Packages
+
+You can setup a local mirror for apt packages using squid-deb-proxy.
+For instructions on configuring juju to use this, see the
+[Juju Proxy Documentation](https://juju.ubuntu.com/docs/howto-proxies.html).
+
+
+## Contact Information
+
+- <bigdata@lists.ubuntu.com>
+
+
+## Resources
+
+- [Apache Bigtop](http://bigtop.apache.org/) home page
+- [Apache Bigtop issue tracking](http://bigtop.apache.org/issue-tracking.html)
+- [Apache Bigtop mailing lists](http://bigtop.apache.org/mail-lists.html)
+- [Juju Bigtop charms](https://jujucharms.com/q/apache/bigtop)
+- [Juju mailing list](https://lists.ubuntu.com/mailman/listinfo/juju)
+- [Juju community](https://jujucharms.com/community)
diff --git a/bigtop-deploy/juju/hadoop-processing/bundle-dev.yaml b/bigtop-deploy/juju/hadoop-processing/bundle-dev.yaml
new file mode 100644
index 00000000..abc18513
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/bundle-dev.yaml
@@ -0,0 +1,68 @@
+services:
+  openjdk:
+    charm: cs:trusty/openjdk
+    annotations:
+      gui-x: "500"
+      gui-y: "400"
+    options:
+      java-type: "jdk"
+      java-major: "8"
+  namenode:
+    charm: cs:~bigdata-dev/trusty/hadoop-namenode
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "800"
+    constraints: mem=7G
+  resourcemanager:
+    charm: cs:~bigdata-dev/trusty/hadoop-resourcemanager
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "0"
+    constraints: mem=7G
+  slave:
+    charm: cs:~bigdata-dev/trusty/hadoop-slave
+    num_units: 3
+    annotations:
+      gui-x: "0"
+      gui-y: "400"
+    constraints: mem=7G
+  plugin:
+    charm: cs:~bigdata-dev/trusty/hadoop-plugin
+    annotations:
+      gui-x: "1000"
+      gui-y: "400"
+  client:
+    charm: cs:trusty/hadoop-client
+    num_units: 1
+    annotations:
+      gui-x: "1250"
+      gui-y: "400"
+  ganglia-node:
+    charm: cs:trusty/ganglia-node
+    annotations:
+      gui-x: "250"
+      gui-y: "400"
+  ganglia:
+    charm: cs:trusty/ganglia
+    num_units: 1
+    annotations:
+      gui-x: "750"
+      gui-y: "400"
+series: trusty
+relations:
+  - [openjdk, namenode]
+  - [openjdk, resourcemanager]
+  - [openjdk, slave]
+  - [openjdk, client]
+  - [resourcemanager, namenode]
+  - [namenode, slave]
+  - [resourcemanager, slave]
+  - [plugin, namenode]
+  - [plugin, resourcemanager]
+  - [client, plugin]
+  - ["ganglia:node", ganglia-node]
+  - [ganglia-node, namenode]
+  - [ganglia-node, resourcemanager]
+  - [ganglia-node, slave]
diff --git a/bigtop-deploy/juju/hadoop-processing/bundle-local.yaml b/bigtop-deploy/juju/hadoop-processing/bundle-local.yaml
new file mode 100644
index 00000000..3947f829
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/bundle-local.yaml
@@ -0,0 +1,68 @@
+services:
+  openjdk:
+    charm: cs:trusty/openjdk
+    annotations:
+      gui-x: "500"
+      gui-y: "400"
+    options:
+      java-type: "jdk"
+      java-major: "8"
+  namenode:
+    charm: local:trusty/hadoop-namenode
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "800"
+    constraints: mem=7G
+  resourcemanager:
+    charm: local:trusty/hadoop-resourcemanager
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "0"
+    constraints: mem=7G
+  slave:
+    charm: local:trusty/hadoop-slave
+    num_units: 3
+    annotations:
+      gui-x: "0"
+      gui-y: "400"
+    constraints: mem=7G
+  plugin:
+    charm: local:trusty/hadoop-plugin
+    annotations:
+      gui-x: "1000"
+      gui-y: "400"
+  client:
+    charm: cs:trusty/hadoop-client
+    num_units: 1
+    annotations:
+      gui-x: "1250"
+      gui-y: "400"
+  ganglia-node:
+    charm: cs:trusty/ganglia-node
+    annotations:
+      gui-x: "250"
+      gui-y: "400"
+  ganglia:
+    charm: cs:trusty/ganglia
+    num_units: 1
+    annotations:
+      gui-x: "750"
+      gui-y: "400"
+series: trusty
+relations:
+  - [openjdk, namenode]
+  - [openjdk, resourcemanager]
+  - [openjdk, slave]
+  - [openjdk, client]
+  - [resourcemanager, namenode]
+  - [namenode, slave]
+  - [resourcemanager, slave]
+  - [plugin, namenode]
+  - [plugin, resourcemanager]
+  - [client, plugin]
+  - ["ganglia:node", ganglia-node]
+  - [ganglia-node, namenode]
+  - [ganglia-node, resourcemanager]
+  - [ganglia-node, slave]
diff --git a/bigtop-deploy/juju/hadoop-processing/bundle.yaml b/bigtop-deploy/juju/hadoop-processing/bundle.yaml
new file mode 100644
index 00000000..dcc5bd99
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/bundle.yaml
@@ -0,0 +1,68 @@
+services:
+  openjdk:
+    charm: cs:trusty/openjdk-1
+    annotations:
+      gui-x: "500"
+      gui-y: "400"
+    options:
+      java-type: "jdk"
+      java-major: "8"
+  namenode:
+    charm: cs:trusty/hadoop-namenode-3
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "800"
+    constraints: mem=7G
+  resourcemanager:
+    charm: cs:trusty/hadoop-resourcemanager-3
+    num_units: 1
+    annotations:
+      gui-x: "500"
+      gui-y: "0"
+    constraints: mem=7G
+  slave:
+    charm: cs:trusty/hadoop-slave-4
+    num_units: 3
+    annotations:
+      gui-x: "0"
+      gui-y: "400"
+    constraints: mem=7G
+  plugin:
+    charm: cs:trusty/hadoop-plugin-3
+    annotations:
+      gui-x: "1000"
+      gui-y: "400"
+  client:
+    charm: cs:trusty/hadoop-client-4
+    num_units: 1
+    annotations:
+      gui-x: "1250"
+      gui-y: "400"
+  ganglia-node:
+    charm: cs:trusty/ganglia-node-2
+    annotations:
+      gui-x: "250"
+      gui-y: "400"
+  ganglia:
+    charm: cs:trusty/ganglia-2
+    num_units: 1
+    annotations:
+      gui-x: "750"
+      gui-y: "400"
+series: trusty
+relations:
+  - [openjdk, namenode]
+  - [openjdk, resourcemanager]
+  - [openjdk, slave]
+  - [openjdk, client]
+  - [resourcemanager, namenode]
+  - [namenode, slave]
+  - [resourcemanager, slave]
+  - [plugin, namenode]
+  - [plugin, resourcemanager]
+  - [client, plugin]
+  - ["ganglia:node", ganglia-node]
+  - [ganglia-node, namenode]
+  - [ganglia-node, resourcemanager]
+  - [ganglia-node, slave]
diff --git a/bigtop-deploy/juju/hadoop-processing/copyright b/bigtop-deploy/juju/hadoop-processing/copyright
new file mode 100644
index 00000000..e900b97c
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/copyright
@@ -0,0 +1,16 @@
+Format: http://dep.debian.net/deps/dep5/
+
+Files: *
+Copyright: Copyright 2015, Canonical Ltd., All Rights Reserved.
+License: Apache License 2.0
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+ .
+     http://www.apache.org/licenses/LICENSE-2.0
+ .
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
diff --git a/bigtop-deploy/juju/hadoop-processing/tests/01-bundle.py b/bigtop-deploy/juju/hadoop-processing/tests/01-bundle.py
new file mode 100755
index 00000000..176ff748
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/tests/01-bundle.py
@@ -0,0 +1,96 @@
+#!/usr/bin/env python3
+
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import unittest
+
+import yaml
+import amulet
+
+
+class TestBundle(unittest.TestCase):
+    bundle_file = os.path.join(os.path.dirname(__file__), '..', 'bundle.yaml')
+
+    @classmethod
+    def setUpClass(cls):
+        # classmethod inheritance doesn't work quite right with
+        # setUpClass / tearDownClass, so subclasses have to manually call this
+        cls.d = amulet.Deployment(series='trusty')
+        with open(cls.bundle_file) as f:
+            bun = f.read()
+        bundle = yaml.safe_load(bun)
+        cls.d.load(bundle)
+        cls.d.setup(timeout=3600)
+        cls.d.sentry.wait_for_messages({'client': 'Ready'}, timeout=3600)
+        cls.hdfs = cls.d.sentry['namenode'][0]
+        cls.yarn = cls.d.sentry['resourcemanager'][0]
+        cls.slave = cls.d.sentry['slave'][0]
+        cls.client = cls.d.sentry['client'][0]
+
+    def test_components(self):
+        """
+        Confirm that all of the required components are up and running.
+        """
+        hdfs, retcode = self.hdfs.run("pgrep -a java")
+        yarn, retcode = self.yarn.run("pgrep -a java")
+        slave, retcode = self.slave.run("pgrep -a java")
+        client, retcode = self.client.run("pgrep -a java")
+
+        assert 'NameNode' in hdfs, "NameNode not started"
+        assert 'NameNode' not in yarn, "NameNode should not be running on resourcemanager"
+        assert 'NameNode' not in slave, "NameNode should not be running on slave"
+
+        assert 'ResourceManager' in yarn, "ResourceManager not started"
+        assert 'ResourceManager' not in hdfs, "ResourceManager should not be running on namenode"
+        assert 'ResourceManager' not in slave, "ResourceManager should not be running on slave"
+
+        assert 'JobHistoryServer' in yarn, "JobHistoryServer not started"
+        assert 'JobHistoryServer' not in hdfs, "JobHistoryServer should not be running on namenode"
+        assert 'JobHistoryServer' not in slave, "JobHistoryServer should not be running on slave"
+
+        assert 'NodeManager' in slave, "NodeManager not started"
+        assert 'NodeManager' not in yarn, "NodeManager should not be running on resourcemanager"
+        assert 'NodeManager' not in hdfs, "NodeManager should not be running on namenode"
+
+        assert 'DataNode' in slave, "DataServer not started"
+        assert 'DataNode' not in yarn, "DataNode should not be running on resourcemanager"
+        assert 'DataNode' not in hdfs, "DataNode should not be running on namenode"
+
+    def test_hdfs(self):
+        """Smoke test validates mkdir, ls, chmod, and rm on the hdfs cluster."""
+        unit_name = self.hdfs.info['unit_name']
+        uuid = self.d.action_do(unit_name, 'smoke-test')
+        result = self.d.action_fetch(uuid)
+        # hdfs smoke-test sets outcome=success on success
+        if (result['outcome'] != "success"):
+            error = "HDFS smoke-test failed"
+            amulet.raise_status(amulet.FAIL, msg=error)
+
+    def test_yarn(self):
+        """Smoke test validates teragen/terasort."""
+        unit_name = self.yarn.info['unit_name']
+        uuid = self.d.action_do(unit_name, 'smoke-test')
+        result = self.d.action_fetch(uuid)
+        # yarn smoke-test only returns results on failure; if result is not
+        # empty, the test has failed and has a 'log' key
+        if result:
+            error = "YARN smoke-test failed: %s" % result['log']
+            amulet.raise_status(amulet.FAIL, msg=error)
+
+
+if __name__ == '__main__':
+    unittest.main()
diff --git a/bigtop-deploy/juju/hadoop-processing/tests/tests.yaml b/bigtop-deploy/juju/hadoop-processing/tests/tests.yaml
new file mode 100644
index 00000000..8a4cf6f1
--- /dev/null
+++ b/bigtop-deploy/juju/hadoop-processing/tests/tests.yaml
@@ -0,0 +1,4 @@
+reset: false
+packages:
+  - amulet
+  - python3-yaml
diff --git a/bigtop-deploy/puppet/README.md b/bigtop-deploy/puppet/README.md
index 6364b9e8..692c3952 100644
--- a/bigtop-deploy/puppet/README.md
+++ b/bigtop-deploy/puppet/README.md
@@ -137,7 +137,7 @@ hadoop_cluster_node::cluster_components:
   - yarn
   - zookeeper
 bigtop::jdk_package_name: "openjdk-7-jre-headless"
-bigtop::bigtop_repo_uri: "http://bigtop.s3.amazonaws.com/releases/1.0.0/ubuntu/trusty/x86_64"
+bigtop::bigtop_repo_uri: "http://bigtop-repos.s3.amazonaws.com/releases/1.1.0/ubuntu/trusty/x86_64"
 ```
 
 And finally execute
diff --git a/bigtop-deploy/puppet/hieradata/bigtop/cluster.yaml b/bigtop-deploy/puppet/hieradata/bigtop/cluster.yaml
index de98502c..5c2c5f12 100644
--- a/bigtop-deploy/puppet/hieradata/bigtop/cluster.yaml
+++ b/bigtop-deploy/puppet/hieradata/bigtop/cluster.yaml
@@ -16,8 +16,8 @@
 # "$components" list. If $components isn't set then everything in the stack will
 # be installed as usual. Otherwise only a specified list will be set
 # Possible elements:
-# hadoop,yarn,hbase,tachyon,flume,solrcloud,spark,oozie,hcat,sqoop,sqoop2,httpfs,
-# hue,mahout,giraph,crunch,pig,hive,zookeeper, ycsb
+# hadoop,yarn,hbase,alluxio,flink,flume,solrcloud,spark,oozie,hcat,sqoop,sqoop2,httpfs,
+# hue,mahout,giraph,crunch,pig,hive,zookeeper,ycsb,qfs
 # Example (to deploy only HDFS and YARN server and gateway parts)
 # This can be a comma-separated list or an array.
 #hadoop_cluster_node::cluster_components:
@@ -54,6 +54,13 @@
 
 #bigtop::bigtop_repo_uri: "http://mirror.example.com/path/to/mirror/"
 
+# Use a pre-installed java environment. The default value of 'false' will cause
+# the configured 'bigtop::jdk_package_name' package to be installed. Setting
+# this to 'true' will ignore the configured 'bigtop::jdk_package_name' but
+# requires a compatible java environment be avaialble prior to Bigtop
+# installation.
+#bigtop::jdk_preinstalled: false
+
 # Test-only variable controls if user hdfs' sshkeys should be installed to allow
 # for passwordless login across the cluster. Required by some integration tests
 #hadoop::common_hdfs::testonly_hdfs_sshkeys: "no"
@@ -98,6 +105,10 @@ hadoop::common_yarn::hadoop_rm_port: "8032"
 
 hadoop::common_mapred_app::jobtracker_host: "%{hiera('bigtop::hadoop_head_node')}"
 hadoop::common_mapred_app::mapreduce_jobhistory_host: "%{hiera('bigtop::hadoop_head_node')}"
+# actually default but needed for hadoop::common_yarn::yarn_log_server_url here
+bigtop::hadoop_history_server_port: "19888"
+bigtop::hadoop_history_server_url: "http://%{hiera('hadoop::common_mapred_app::mapreduce_jobhistory_host')}:%{hiera('bigtop::hadoop_history_server_port')}"
+hadoop::common_yarn::yarn_log_server_url: "%{hiera('bigtop::hadoop_history_server_url')}/jobhistory/logs"
 
 # actually default but needed for hue::server::webhdfs_url here
 hadoop::httpfs::hadoop_httpfs_port: "14000"
@@ -125,7 +136,14 @@ hcatalog::webhcat::server::kerberos_realm: "%{hiera('kerberos::site::realm')}"
 
 spark::common::master_host: "%{hiera('bigtop::hadoop_head_node')}"
 
-tachyon::common::master_host: "%{hiera('bigtop::hadoop_head_node')}"
+alluxio::common::master_host: "%{hiera('bigtop::hadoop_head_node')}"
+
+# qfs
+qfs::common::metaserver_host: "%{hiera('bigtop::hadoop_head_node')}"
+qfs::common::metaserver_port: "30000"
+qfs::common::chunkserver_port: "30000"
+qfs::common::metaserver_client_port: "20000"
+qfs::common::chunkserver_client_port: "22000"
 
 hadoop_zookeeper::server::myid: "0"
 hadoop_zookeeper::server::ensemble:
@@ -135,7 +153,6 @@ hadoop_zookeeper::server::kerberos_realm: "%{hiera('kerberos::site::realm')}"
 # those are only here because they were present as extlookup keys previously
 bigtop::hadoop_rm_http_port: "8088"
 bigtop::hadoop_rm_proxy_port: "8088"
-bigtop::hadoop_history_server_port: "19888"
 bigtop::sqoop2_server_port: "12000"
 bigtop::hbase_thrift_port: "9090"
 bigtop::hadoop_oozie_port: "11000"
@@ -143,8 +160,8 @@ bigtop::hadoop_oozie_port: "11000"
 hue::server::rm_host: "%{hiera('hadoop::common_yarn::hadoop_rm_host')}"
 hue::server::rm_port: "%{hiera('hadoop::common_yarn::hadoop_rm_port')}"
 hue::server::rm_url: "http://%{hiera('bigtop::hadoop_head_node')}:%{hiera('bigtop::hadoop_rm_http_port')}"
-hue::server::rm_proxy_url: "http://%{hiera('bigtop::hadoop_head_node')}:%{hiera('bigtop::hadoop_rm_proxy_port')}"
-hue::server::history_server_url: "http://%{hiera('bigtop::hadoop_head_node')}:%{hiera('bigtop::hadoop_history_server_port')}"
+hue::server::rm_proxy_url: "http://%{hiera('hadoop::common_yarn::hadoop_ps_host')}:%{hiera('hadoop::common_yarn::hadoop_ps_port')}"
+hue::server::history_server_url: "%{hiera('bigtop::hadoop_history_server_url')}"
 # those use fqdn instead of hadoop_head_node because it's only ever activated
 # on the gatewaynode
 hue::server::webhdfs_url: "http://%{fqdn}:%{hiera('hadoop::httpfs::hadoop_httpfs_port')}/webhdfs/v1"
@@ -174,3 +191,10 @@ zeppelin::server::spark_master_url: "yarn-client"
 zeppelin::server::hiveserver2_url: "jdbc:hive2://%{hiera('hadoop-hive::common::hiveserver2_host')}:%{hiera('hadoop-hive::common::hiveserver2_port')}"
 zeppelin::server::hiveserver2_user: "%{hiera('bigtop::hiveserver2_user')}"
 zeppelin::server::hiveserver2_password: "%{hiera('bigtop::hiveserver2_password')}"
+
+# Flink
+flink::common::jobmanager_host: "%{hiera('bigtop::hadoop_head_node')}"
+flink::common::jobmanager_port: "6123"
+
+flink::common::ui_port: "8081"
+flink::common::storage_dirs: "%{hiera('hadoop::hadoop_storage_dirs')}"
diff --git a/bigtop-deploy/puppet/hieradata/site.yaml b/bigtop-deploy/puppet/hieradata/site.yaml
index e44cb500..b9ed6f98 100644
--- a/bigtop-deploy/puppet/hieradata/site.yaml
+++ b/bigtop-deploy/puppet/hieradata/site.yaml
@@ -12,7 +12,10 @@ hadoop::hadoop_storage_dirs:
   - /data/4
 
 #hadoop_cluster_node::cluster_components:
+#  - alluxio
+#  - apex
 #  - crunch
+#  - flink
 #  - flume
 #  - giraph
 #  - ignite_hadoop
@@ -25,11 +28,11 @@ hadoop::hadoop_storage_dirs:
 #  - mapred-app
 #  - oozie
 #  - pig
+#  - qfs
 #  - solrcloud
 #  - spark
 #  - sqoop
 #  - sqoop2
-#  - tachyon
 #  - tez
 #  - yarn
 #  - zookeeper
diff --git a/bigtop-deploy/puppet/manifests/cluster.pp b/bigtop-deploy/puppet/manifests/cluster.pp
index a0be5678..9ff424cd 100644
--- a/bigtop-deploy/puppet/manifests/cluster.pp
+++ b/bigtop-deploy/puppet/manifests/cluster.pp
@@ -14,6 +14,9 @@
 # limitations under the License.
 
 $roles_map = {
+  apex => {
+    client => ["apex-client"],
+  },
   hdfs-non-ha => {
     master => ["namenode"],
     worker => ["datanode"],
@@ -49,9 +52,13 @@ $roles_map = {
     master => ["spark-master"],
     worker => ["spark-worker"],
   },
-  tachyon => {
-    master => ["tachyon-master"],
-    worker => ["tachyon-worker"],
+  alluxio => {
+    master => ["alluxio-master"],
+    worker => ["alluxio-worker"],
+  },
+  flink => {
+    master => ["flink-jobmanager"],
+    worker => ["flink-taskmanager"],
   },
   flume => {
     worker => ["flume-agent"],
@@ -105,6 +112,11 @@ $roles_map = {
   zeppelin => {
     master => ["zeppelin-server"],
   },
+  qfs => {
+    master => ["qfs-metaserver"],
+    worker => ["qfs-chunkserver"],
+    client => ["qfs-client"],
+  },
 }
 
 class hadoop_cluster_node (
@@ -150,7 +162,10 @@ class node_with_roles ($roles = hiera("bigtop::roles")) inherits hadoop_cluster_
   }
 
   $modules = [
+    "alluxio",
+    "apex",
     "crunch",
+    "flink",
     "giraph",
     "hadoop",
     "hadoop_hbase",
@@ -166,7 +181,7 @@ class node_with_roles ($roles = hiera("bigtop::roles")) inherits hadoop_cluster_
     "mahout",
     "solr",
     "spark",
-    "tachyon",
+    "qfs",
     "tez",
     "ycsb",
     "kerberos",
diff --git a/bigtop-deploy/puppet/manifests/site.pp b/bigtop-deploy/puppet/manifests/site.pp
index 6e5fd06b..4862b223 100644
--- a/bigtop-deploy/puppet/manifests/site.pp
+++ b/bigtop-deploy/puppet/manifests/site.pp
@@ -28,6 +28,7 @@ case $operatingsystem {
     }
 }
 
+$jdk_preinstalled = hiera("bigtop::jdk_preinstalled", false)
 $jdk_package_name = hiera("bigtop::jdk_package_name", "jdk")
 
 stage {"pre": before => Stage["main"]}
@@ -70,6 +71,7 @@ case $operatingsystem {
 package { $jdk_package_name:
   ensure => "installed",
   alias => "jdk",
+  noop => $jdk_preinstalled,
 }
 
 import "cluster.pp"
diff --git a/bigtop-deploy/puppet/modules/alluxio/manifests/init.pp b/bigtop-deploy/puppet/modules/alluxio/manifests/init.pp
new file mode 100644
index 00000000..66151cad
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/alluxio/manifests/init.pp
@@ -0,0 +1,79 @@
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+class alluxio {
+
+  class deploy ($roles) {
+    if ("alluxio-master" in $roles) {
+      include alluxio::master
+    }
+
+    if ("alluxio-worker" in $roles) {
+      include alluxio::worker
+    }
+  }
+
+  class common ($master_host){
+    package { "alluxio":
+      ensure => latest,
+    }
+
+    # add logging into /var/log/..
+    file {
+        "/etc/alluxio/conf/log4j.properties":
+        content => template("alluxio/log4j.properties"),
+        require => [Package["alluxio"]]
+    }
+
+    # add alluxio-env.sh to point to alluxio master
+    file { "/etc/alluxio/conf/alluxio-env.sh":
+        content => template("alluxio/alluxio-env.sh"),
+        require => [Package["alluxio"]]
+    }
+  }
+
+  class master {
+    include common
+
+   exec {
+        "alluxio formatting":
+           command => "/usr/lib/alluxio/bin/alluxio format",
+           require => [ Package["alluxio"], File["/etc/alluxio/conf/log4j.properties"], File["/etc/alluxio/conf/alluxio-env.sh"] ]
+    }
+
+    if ( $fqdn == $alluxio::common::master_host ) {
+      service { "alluxio-master":
+        ensure => running,
+        require => [ Package["alluxio"], Exec["alluxio formatting"] ],
+        hasrestart => true,
+        hasstatus => true,
+      }
+    }
+
+  }
+
+  class worker {
+    include common
+
+   if ( $fqdn == $alluxio::common::master_host ) {
+      notice("alluxio ---> master host")
+      # We want master to run first in all cases
+      Service["alluxio-master"] ~> Service["alluxio-worker"]
+   }
+
+    service { "alluxio-worker":
+      ensure => running,
+      require => [ Package["alluxio"], File["/etc/alluxio/conf/log4j.properties"], File["/etc/alluxio/conf/alluxio-env.sh"] ],
+      hasrestart => true,
+      hasstatus => true,
+    }
+  }
+}
diff --git a/bigtop-deploy/puppet/modules/alluxio/templates/alluxio-env.sh b/bigtop-deploy/puppet/modules/alluxio/templates/alluxio-env.sh
new file mode 100755
index 00000000..27c7fb2b
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/alluxio/templates/alluxio-env.sh
@@ -0,0 +1,78 @@
+#!/usr/bin/env bash
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This file contains environment variables required to run Alluxio. Copy it as alluxio-env.sh and
+# edit that to configure Alluxio for your site. At a minimum,
+# the following variables should be set:
+#
+# - JAVA_HOME, to point to your JAVA installation
+# - ALLUXIO_MASTER_ADDRESS, to bind the master to a different IP address or hostname
+# - ALLUXIO_UNDERFS_ADDRESS, to set the under filesystem address.
+# - ALLUXIO_WORKER_MEMORY_SIZE, to set how much memory to use (e.g. 1000mb, 2gb) per worker
+# - ALLUXIO_RAM_FOLDER, to set where worker stores in memory data
+# - ALLUXIO_UNDERFS_HDFS_IMPL, to set which HDFS implementation to use (e.g. com.mapr.fs.MapRFileSystem,
+#   org.apache.hadoop.hdfs.DistributedFileSystem)
+
+# The following gives an example:
+
+if [[ `uname -a` == Darwin* ]]; then
+  # Assuming Mac OS X
+  export JAVA_HOME=${JAVA_HOME:-$(/usr/libexec/java_home)}
+  export ALLUXIO_RAM_FOLDER=/Volumes/ramdisk
+  export ALLUXIO_JAVA_OPTS="-Djava.security.krb5.realm= -Djava.security.krb5.kdc="
+else
+  # Assuming Linux
+  if [ -z "$JAVA_HOME" ]; then
+    export JAVA_HOME=/usr/lib/jvm/java-7-oracle
+  fi
+  export ALLUXIO_RAM_FOLDER=/mnt/ramdisk
+fi
+
+export JAVA="$JAVA_HOME/bin/java"
+
+echo "Starting alluxio w/ java = $JAVA "
+
+export ALLUXIO_MASTER_ADDRESS=<%= @master_host %>
+export ALLUXIO_UNDERFS_ADDRESS=$ALLUXIO_HOME/underfs
+#export ALLUXIO_UNDERFS_ADDRESS=hdfs://localhost:9000
+export ALLUXIO_WORKER_MEMORY_SIZE=1GB
+export ALLUXIO_UNDERFS_HDFS_IMPL=org.apache.hadoop.hdfs.DistributedFileSystem
+
+echo "ALLUXIO master => $ALLUXIO_MASTER_ADDRESS "
+
+CONF_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+
+export ALLUXIO_JAVA_OPTS+="
+  -Dlog4j.configuration=file:$CONF_DIR/log4j.properties
+  -Dalluxio.debug=false
+  -Dalluxio.underfs.address=$ALLUXIO_UNDERFS_ADDRESS
+  -Dalluxio.underfs.hdfs.impl=$ALLUXIO_UNDERFS_HDFS_IMPL
+  -Dalluxio.data.folder=$ALLUXIO_UNDERFS_ADDRESS/tmp/alluxio/data
+  -Dalluxio.workers.folder=$ALLUXIO_UNDERFS_ADDRESS/tmp/alluxio/workers
+  -Dalluxio.worker.memory.size=$ALLUXIO_WORKER_MEMORY_SIZE
+  -Dalluxio.worker.data.folder=$ALLUXIO_RAM_FOLDER/alluxioworker/
+  -Dalluxio.master.worker.timeout.ms=60000
+  -Dalluxio.master.hostname=$ALLUXIO_MASTER_ADDRESS
+  -Dalluxio.master.journal.folder=$ALLUXIO_HOME/journal/
+  -Dorg.apache.jasper.compiler.disablejsr199=true
+  -Djava.net.preferIPv4Stack=true
+"
+
+# Master specific parameters. Default to ALLUXIO_JAVA_OPTS.
+export ALLUXIO_MASTER_JAVA_OPTS="$ALLUXIO_JAVA_OPTS"
+
+# Worker specific parameters that will be shared to all workers. Default to ALLUXIO_JAVA_OPTS.
+export ALLUXIO_WORKER_JAVA_OPTS="$ALLUXIO_JAVA_OPTS"
diff --git a/bigtop-deploy/puppet/modules/tachyon/templates/log4j.properties b/bigtop-deploy/puppet/modules/alluxio/templates/log4j.properties
index e3c5f04c..7ef17e68 100644
--- a/bigtop-deploy/puppet/modules/tachyon/templates/log4j.properties
+++ b/bigtop-deploy/puppet/modules/alluxio/templates/log4j.properties
@@ -13,9 +13,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # May get overridden by System Property
-tachyon.logger.type=Console
+alluxio.logger.type=Console
 
-log4j.rootLogger=INFO, ${tachyon.logger.type}
+log4j.rootLogger=INFO, ${alluxio.logger.type}
 
 log4j.appender.Console=org.apache.log4j.ConsoleAppender
 log4j.appender.Console.Target=System.out
@@ -23,8 +23,8 @@ log4j.appender.Console.layout=org.apache.log4j.PatternLayout
 log4j.appender.Console.layout.ConversionPattern=%d{ISO8601} %-5p %c{1} (%F:%M) - %m%n
 
 # Appender for Master
-log4j.appender.MASTER_LOGGER=tachyon.Log4jFileAppender
-log4j.appender.MASTER_LOGGER.File=/var/log/tachyon/master.log
+log4j.appender.MASTER_LOGGER=alluxio.Log4jFileAppender
+log4j.appender.MASTER_LOGGER.File=/var/log/alluxio/master.log
 
 log4j.appender.MASTER_LOGGER.MaxFileSize=10
 log4j.appender.MASTER_LOGGER.MaxBackupIndex=100
@@ -34,8 +34,8 @@ log4j.appender.MASTER_LOGGER.layout.ConversionPattern=%d{ISO8601} %-5p %c{2} (%F
 #log4j.appender.MASTER_LOGGER.layout.ConversionPattern=%d{ISO8601} %-5p %c{2} (%F:%M(%L)) - %m%n
 
 # Appender for Workers
-log4j.appender.WORKER_LOGGER=tachyon.Log4jFileAppender
-log4j.appender.WORKER_LOGGER.File=/var/log/tachyon/worker.log
+log4j.appender.WORKER_LOGGER=alluxio.Log4jFileAppender
+log4j.appender.WORKER_LOGGER.File=/var/log/alluxio/worker.log
 
 log4j.appender.WORKER_LOGGER.MaxFileSize=10
 log4j.appender.WORKER_LOGGER.MaxBackupIndex=100
@@ -45,8 +45,8 @@ log4j.appender.WORKER_LOGGER.layout.ConversionPattern=%d{ISO8601} %-5p %c{2} (%F
 #log4j.appender.WORKER_LOGGER.layout.ConversionPattern=%d{ISO8601} %-5p %c{2} (%F:%M(%L)) - %m%n
 
 # Appender for User
-log4j.appender.USER_LOGGER=tachyon.Log4jFileAppender
-log4j.appender.USER_LOGGER.File=/var/log/tachyon/logs/user.log
+log4j.appender.USER_LOGGER=alluxio.Log4jFileAppender
+log4j.appender.USER_LOGGER.File=/var/log/alluxio/logs/user.log
 log4j.appender.USER_LOGGER.MaxFileSize=10
 log4j.appender.USER_LOGGER.MaxBackupIndex=10
 log4j.appender.USER_LOGGER.DeletionPercentage=20
diff --git a/bigtop-deploy/puppet/modules/apex/manifests/init.pp b/bigtop-deploy/puppet/modules/apex/manifests/init.pp
new file mode 100644
index 00000000..c4b91c91
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/apex/manifests/init.pp
@@ -0,0 +1,30 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+class apex {
+
+  class deploy ($roles) {
+    if ("apex-client" in $roles) {
+      include apex::client
+    }
+  }
+
+  class client {
+    package { "apex":
+      ensure => latest,
+      require => Package["hadoop"],
+    }
+  }
+}
diff --git a/bigtop-deploy/puppet/modules/apex/tests/init.pp b/bigtop-deploy/puppet/modules/apex/tests/init.pp
new file mode 100644
index 00000000..0885cbfa
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/apex/tests/init.pp
@@ -0,0 +1,17 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+include apex
+apex::client { "test-apex": }
diff --git a/bigtop-deploy/puppet/modules/flink/manifests/init.pp b/bigtop-deploy/puppet/modules/flink/manifests/init.pp
new file mode 100644
index 00000000..ae79fe45
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/flink/manifests/init.pp
@@ -0,0 +1,62 @@
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+class flink {
+
+  class deploy ($roles) {
+    if ("flink-jobmanager" in $roles) {
+      include flink::jobmanager
+    }
+
+    if ("flink-taskmanager" in $roles) {
+      include flink::taskmanager
+    }
+  }
+
+  class common($jobmanager_host, $jobmanager_port, $ui_port, $storage_dirs) {
+    # make sure flink is installed
+    package { "flink":
+      ensure => latest
+    }
+
+    # set values in flink-conf.yaml
+    file { "/etc/flink/conf/flink-conf.yaml":
+        content => template("flink/flink-conf.yaml"),
+        require => Package["flink"]
+    }
+  }
+
+  class jobmanager {
+    include common
+
+    service { "flink-jobmanager":
+      ensure => running,
+      require => Package["flink"],
+      subscribe => File["/etc/flink/conf/flink-conf.yaml"],
+      hasrestart => true,
+      hasstatus => true
+     
+    }
+
+  }
+
+  class taskmanager {
+    include common
+
+    service { "flink-taskmanager":
+      ensure => running,
+      require => Package["flink"],
+      subscribe => File["/etc/flink/conf/flink-conf.yaml"],
+      hasrestart => true,
+      hasstatus => true,
+    }
+  }
+}
diff --git a/bigtop-deploy/puppet/modules/flink/templates/flink-conf.yaml b/bigtop-deploy/puppet/modules/flink/templates/flink-conf.yaml
new file mode 100644
index 00000000..14e8853b
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/flink/templates/flink-conf.yaml
@@ -0,0 +1,31 @@
+################################################################################
+#  Licensed to the Apache Software Foundation (ASF) under one
+#  or more contributor license agreements.  See the NOTICE file
+#  distributed with this work for additional information
+#  regarding copyright ownership.  The ASF licenses this file
+#  to you under the Apache License, Version 2.0 (the
+#  "License"); you may not use this file except in compliance
+#  with the License.  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+# limitations under the License.
+################################################################################
+
+
+# Configuration values managed by puppet:
+jobmanager.rpc.address: <%= @jobmanager_host %>
+jobmanager.rpc.port: <%= @jobmanager_port %>
+jobmanager.web.port: <%= @ui_port %>
+
+<% if defined?(storage_dirs) %>
+taskmanager.tmp.dirs: <%= @storage_dirs.join(":") %>
+<% end %>
+
+
+# For performance reasons its highly recommended to allocate as much memory to the
+# Flink TaskManager as possible by setting 'taskmanager.heap.mb'.
diff --git a/bigtop-deploy/puppet/modules/hadoop/manifests/init.pp b/bigtop-deploy/puppet/modules/hadoop/manifests/init.pp
index 64b3ee69..90c9505b 100644
--- a/bigtop-deploy/puppet/modules/hadoop/manifests/init.pp
+++ b/bigtop-deploy/puppet/modules/hadoop/manifests/init.pp
@@ -61,11 +61,9 @@ class hadoop ($hadoop_security_authentication = "simple",
       include hadoop::historyserver
       include hadoop::proxyserver
 
-      Class['Hadoop::Init_hdfs'] -> Class['Hadoop::Resourcemanager']
       if ("nodemanager" in $roles) {
         Class['Hadoop::Resourcemanager'] -> Class['Hadoop::Nodemanager']
       }
-      Class['Hadoop::Init_hdfs'] -> Class['Hadoop::Historyserver']
     }
 
     if ($hadoop::common_hdfs::ha == "disabled" and "secondarynamenode" in $roles) {
@@ -140,6 +138,7 @@ class hadoop ($hadoop_security_authentication = "simple",
       $hadoop_rm_webapp_port = "8088",
       $hadoop_rt_port = "8025",
       $hadoop_sc_port = "8030",
+      $yarn_log_server_url = undef,
       $yarn_nodemanager_resource_memory_mb = undef,
       $yarn_scheduler_maximum_allocation_mb = undef,
       $yarn_scheduler_minimum_allocation_mb = undef,
@@ -827,6 +826,7 @@ class hadoop ($hadoop_security_authentication = "simple",
 
 
   class nodemanager {
+    include common_mapred_app
     include common_yarn
 
     package { "hadoop-yarn-nodemanager":
@@ -868,6 +868,7 @@ class hadoop ($hadoop_security_authentication = "simple",
 
   class client {
       include common_mapred_app
+      include common_yarn
 
       $hadoop_client_packages = $operatingsystem ? {
         /(OracleLinux|CentOS|RedHat|Fedora)/  => [ "hadoop-doc", "hadoop-hdfs-fuse", "hadoop-client", "hadoop-libhdfs", "hadoop-debuginfo" ],
diff --git a/bigtop-deploy/puppet/modules/hadoop/templates/yarn-site.xml b/bigtop-deploy/puppet/modules/hadoop/templates/yarn-site.xml
index 6f6c4648..054ae653 100644
--- a/bigtop-deploy/puppet/modules/hadoop/templates/yarn-site.xml
+++ b/bigtop-deploy/puppet/modules/hadoop/templates/yarn-site.xml
@@ -145,7 +145,12 @@
     <name>yarn.log-aggregation-enable</name>
     <value>true</value>
   </property>
-
+<% if @yarn_log_server_url %>
+  <property>
+    <name>yarn.log.server.url</name>
+    <value><%= @yarn_log_server_url %></value>
+  </property>
+<% end %>
   <property>
     <name>yarn.dispatcher.exit-on-error</name>
     <value>true</value>
diff --git a/bigtop-deploy/puppet/modules/hadoop_hbase/templates/hbase-env.sh b/bigtop-deploy/puppet/modules/hadoop_hbase/templates/hbase-env.sh
index 2f965813..a2dc9aff 100644
--- a/bigtop-deploy/puppet/modules/hadoop_hbase/templates/hbase-env.sh
+++ b/bigtop-deploy/puppet/modules/hadoop_hbase/templates/hbase-env.sh
@@ -1,7 +1,5 @@
 #
 #/**
-# * Copyright 2007 The Apache Software Foundation
-# *
 # * Licensed to the Apache Software Foundation (ASF) under one
 # * or more contributor license agreements.  See the NOTICE file
 # * distributed with this work for additional information
@@ -21,15 +19,24 @@
 
 # Set environment variables here.
 
-# The java implementation to use.  Java 1.6 required.
+# This script sets variables multiple times over the course of starting an hbase process,
+# so try to keep things idempotent unless you want to take an even deeper look
+# into the startup scripts (bin/hbase, etc.)
+
+# The java implementation to use.  Java 1.7+ required.
 # export JAVA_HOME=/usr/java/jdk1.6.0/
 
 # Extra Java CLASSPATH elements.  Optional.
 export HBASE_CLASSPATH=/etc/hadoop/conf
 
-# The maximum amount of heap to use, in MB. Default is 1000.
+# The maximum amount of heap to use. Default is left to JVM default.
+# export HBASE_HEAPSIZE=1G
 export HBASE_HEAPSIZE=<%= @heap_size %>
 
+# Uncomment below if you intend to use off heap cache. For example, to allocate 8G of 
+# offheap, set the value to "8G".
+# export HBASE_OFFHEAPSIZE=1G
+
 # Extra Java runtime options.
 # Below are what we set by default.  May only work with SUN JVM.
 # For more on why as well as other possible settings,
@@ -42,28 +49,70 @@ export HBASE_MASTER_OPTS="$HBASE_MASTER_OPTS -Djava.security.auth.login.config=/
 export HBASE_REGIONSERVER_OPTS="$HBASE_REGIONSERVER_OPTS -Djava.security.auth.login.config=/etc/hbase/conf/jaas.conf"
 <% end -%>
 
-# Uncomment below to enable java garbage collection logging.
-# export HBASE_OPTS="$HBASE_OPTS -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:$HBASE_HOME/logs/gc-hbase.log" 
+# Uncomment one of the below three options to enable java garbage collection logging for the server-side processes.
+
+# This enables basic gc logging to the .out file.
+# export SERVER_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps"
+
+# This enables basic gc logging to its own file.
+# If FILE-PATH is not replaced, the log file(.gc) would still be generated in the HBASE_LOG_DIR .
+# export SERVER_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:<FILE-PATH>"
+
+# This enables basic GC logging to its own file with automatic log rolling. Only applies to jdk 1.6.0_34+ and 1.7.0_2+.
+# If FILE-PATH is not replaced, the log file(.gc) would still be generated in the HBASE_LOG_DIR .
+# export SERVER_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:<FILE-PATH> -XX:+UseGCLogFileRotation -XX:NumberOfGCLogFiles=1 -XX:GCLogFileSize=512M"
+
+# Uncomment one of the below three options to enable java garbage collection logging for the client processes.
+
+# This enables basic gc logging to the .out file.
+# export CLIENT_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps"
+
+# This enables basic gc logging to its own file.
+# If FILE-PATH is not replaced, the log file(.gc) would still be generated in the HBASE_LOG_DIR .
+# export CLIENT_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:<FILE-PATH>"
+
+# This enables basic GC logging to its own file with automatic log rolling. Only applies to jdk 1.6.0_34+ and 1.7.0_2+.
+# If FILE-PATH is not replaced, the log file(.gc) would still be generated in the HBASE_LOG_DIR .
+# export CLIENT_GC_OPTS="-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:<FILE-PATH> -XX:+UseGCLogFileRotation -XX:NumberOfGCLogFiles=1 -XX:GCLogFileSize=512M"
+
+# See the package documentation for org.apache.hadoop.hbase.io.hfile for other configurations
+# needed setting up off-heap block caching. 
 
 # Uncomment and adjust to enable JMX exporting
 # See jmxremote.password and jmxremote.access in $JRE_HOME/lib/management to configure remote password access.
 # More details at: http://java.sun.com/javase/6/docs/technotes/guides/management/agent.html
-#
+# NOTE: HBase provides an alternative JMX implementation to fix the random ports issue, please see JMX
+# section in HBase Reference Guide for instructions.
+
 # export HBASE_JMX_BASE="-Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false"
 # export HBASE_MASTER_OPTS="$HBASE_MASTER_OPTS $HBASE_JMX_BASE -Dcom.sun.management.jmxremote.port=10101"
 # export HBASE_REGIONSERVER_OPTS="$HBASE_REGIONSERVER_OPTS $HBASE_JMX_BASE -Dcom.sun.management.jmxremote.port=10102"
 # export HBASE_THRIFT_OPTS="$HBASE_THRIFT_OPTS $HBASE_JMX_BASE -Dcom.sun.management.jmxremote.port=10103"
 # export HBASE_ZOOKEEPER_OPTS="$HBASE_ZOOKEEPER_OPTS $HBASE_JMX_BASE -Dcom.sun.management.jmxremote.port=10104"
+# export HBASE_REST_OPTS="$HBASE_REST_OPTS $HBASE_JMX_BASE -Dcom.sun.management.jmxremote.port=10105"
 
 # File naming hosts on which HRegionServers will run.  $HBASE_HOME/conf/regionservers by default.
 # export HBASE_REGIONSERVERS=${HBASE_HOME}/conf/regionservers
 
+# Uncomment and adjust to keep all the Region Server pages mapped to be memory resident
+#HBASE_REGIONSERVER_MLOCK=true
+#HBASE_REGIONSERVER_UID="hbase"
+
+# File naming hosts on which backup HMaster will run.  $HBASE_HOME/conf/backup-masters by default.
+# export HBASE_BACKUP_MASTERS=${HBASE_HOME}/conf/backup-masters
+
 # Extra ssh options.  Empty by default.
 # export HBASE_SSH_OPTS="-o ConnectTimeout=1 -o SendEnv=HBASE_CONF_DIR"
 
 # Where log files are stored.  $HBASE_HOME/logs by default.
 # export HBASE_LOG_DIR=${HBASE_HOME}/logs
 
+# Enable remote JDWP debugging of major HBase processes. Meant for Core Developers 
+# export HBASE_MASTER_OPTS="$HBASE_MASTER_OPTS -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=8070"
+# export HBASE_REGIONSERVER_OPTS="$HBASE_REGIONSERVER_OPTS -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=8071"
+# export HBASE_THRIFT_OPTS="$HBASE_THRIFT_OPTS -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=8072"
+# export HBASE_ZOOKEEPER_OPTS="$HBASE_ZOOKEEPER_OPTS -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=8073"
+
 # A string representing this instance of hbase. $USER by default.
 # export HBASE_IDENT_STRING=$USER
 
@@ -80,3 +129,12 @@ export HBASE_REGIONSERVER_OPTS="$HBASE_REGIONSERVER_OPTS -Djava.security.auth.lo
 
 # Tell HBase whether it should manage it's own instance of Zookeeper or not.
 # export HBASE_MANAGES_ZK=true
+
+# The default log rolling policy is RFA, where the log file is rolled as per the size defined for the 
+# RFA appender. Please refer to the log4j.properties file to see more details on this appender.
+# In case one needs to do log rolling on a date change, one should set the environment property
+# HBASE_ROOT_LOGGER to "<DESIRED_LOG LEVEL>,DRFA".
+# For example:
+# HBASE_ROOT_LOGGER=INFO,DRFA
+# The reason for changing default to RFA is to avoid the boundary case of filling out disk space as 
+# DRFA doesn't put any cap on the log size. Please refer to HBase-5655 for more context.
diff --git a/bigtop-deploy/puppet/modules/hadoop_pig/manifests/init.pp b/bigtop-deploy/puppet/modules/hadoop_pig/manifests/init.pp
index 0a09dec3..9582ffd0 100644
--- a/bigtop-deploy/puppet/modules/hadoop_pig/manifests/init.pp
+++ b/bigtop-deploy/puppet/modules/hadoop_pig/manifests/init.pp
@@ -22,11 +22,13 @@ class hadoop_pig {
   }
 
   class client {
+    include hadoop::common
+
     package { "pig":
       ensure => latest,
       require => Package["hadoop"],
-    } 
- 
+    }
+
     file { "/etc/pig/conf/pig.properties":
       content => template('hadoop_pig/pig.properties'),
       require => Package["pig"],
diff --git a/bigtop-deploy/puppet/modules/hue/manifests/init.pp b/bigtop-deploy/puppet/modules/hue/manifests/init.pp
index fa6af886..2b646877 100644
--- a/bigtop-deploy/puppet/modules/hue/manifests/init.pp
+++ b/bigtop-deploy/puppet/modules/hue/manifests/init.pp
@@ -29,6 +29,13 @@ class hue {
   class server($sqoop2_url = "http://localhost:12000/sqoop", $solr_url = "http://localhost:8983/solr/", $hbase_thrift_url = "",
                $webhdfs_url, $rm_host, $rm_port, $oozie_url, $rm_proxy_url, $history_server_url,
                $hive_host = "", $hive_port = "10000",
+		$zookeeper_host_port = "localhost:2181",
+               $force_username_lowercase = "false",
+               $group_filter_value = "objectclass=groupOfEntries",
+               $nt_domain = undef,
+               $use_ldap_username_pattern = false,
+               $ldap_username_pattern = undef,
+               $remote_deployement_dir = "/user/hue/oozie/deployments",
                $rm_logical_name = undef, $rm_api_port = "8088", $app_blacklist = "impala, security",
                $hue_host = "0.0.0.0", $hue_port = "8888", $hue_timezone = "America/Los_Angeles",
                $default_fs = "hdfs://localhost:8020",
@@ -38,7 +45,7 @@ class hue {
                $base_dn = undef , $bind_dn = undef, $bind_password = undef,
                $user_name_attr = undef, $user_filter = undef,
                $group_member_attr = undef, $group_filter = undef,
-               $hue_apps = "all" ) {
+               $hue_apps = "all", $default_hdfs_superuser = "hdfs" ) {
 
     $hue_packages = $hue_apps ? {
       "all"     => [ "hue", "hue-server" ], # The hue metapackage requires all apps
diff --git a/bigtop-deploy/puppet/modules/hue/templates/hue.ini b/bigtop-deploy/puppet/modules/hue/templates/hue.ini
index 8c81b69e..51c03160 100644
--- a/bigtop-deploy/puppet/modules/hue/templates/hue.ini
+++ b/bigtop-deploy/puppet/modules/hue/templates/hue.ini
@@ -70,7 +70,7 @@
   ## default_user=hue
 
   # This should be the hadoop cluster admin
-  ## default_hdfs_superuser=hdfs
+  default_hdfs_superuser=<%= @default_hdfs_superuser %>
 
   # If set to false, runcpserver will not actually start the web server.
   # Used if Apache is being used as a WSGI container.
@@ -198,6 +198,11 @@
     # The search base for finding users and groups
     base_dn="<%= @base_dn %>"
 
+<%     if @nt_domain -%>
+    # The NT domain to connect to (only for use with Active Directory)
+    nt_domain=<%= @nt_domain %>
+<%     end -%>
+
     # URL of the LDAP server
     ldap_url=<%= @ldap_url %> 
 
@@ -229,8 +234,12 @@
 <%     else -%>
     # Pattern for searching for usernames -- Use <username> for the parameter
     # For use when using LdapBackend for Hue authentication
-    #    ldap_username_pattern="uid=<username>,ou=People,dc=mycompany,dc=com"
 
+<%     if @use_ldap_username_pattern -%>
+    # for example, ldap_username_pattern=uid=<username>,ou=People,dc=mycompany,dc=com
+    ldap_username_pattern="<%= @ldap_username_pattern %>"
+<%     end -%> 
+ 
     search_bind_authentication=false
 <%     end -%>
     # Execute this script to produce the bind user password. This will be used
@@ -248,7 +257,7 @@
     ignore_username_case=true
 
     # Force usernames to lowercase when creating new users from LDAP.
-    ## force_username_lowercase=false
+    force_username_lowercase=<%= @force_username_lowercase %>
 
     # Choose which kind of subgrouping to use: nested or suboordinate (deprecated).
     ## subgroups=suboordinate
@@ -282,7 +291,7 @@
 
       # Base filter for searching for groups
 <%     if @group_filter -%>
-      group_filter="objectclass=groupOfEntries"
+      group_filter="<%= @group_filter_value %>"
 <%     end -%>
 
       # The group name attribute in the LDAP schema
@@ -359,7 +368,7 @@
     # Kerberos principal name for Hue
     hue_principal=hue/<%= @fqdn %>@<%= @kerberos_realm %>
     # Path to kinit
-    kinit_path=<%= (@operatingsystem == 'ubuntu' || @operatingsystem == 'Debian') ? '/usr/bin' : '/usr/kerberos/bin' %>/kinit
+    kinit_path=<%= (@operatingsystem == 'ubuntu' || @operatingsystem == 'Debian' || @operatingsystem == 'CentOS' ) ? '/usr/bin' : '/usr/kerberos/bin' %>/kinit
 
 <% end -%>
   # Configuration options for using OAuthBackend (core) login
@@ -660,7 +669,7 @@
   security_enabled=<%= if (@kerberos_realm != "") ; "true" else "false" end %>
 
   # Location on HDFS where the workflows/coordinator are deployed when submitted.
-  remote_deployement_dir=/user/hue/oozie/deployments
+  remote_deployement_dir=<%= @remote_deployement_dir %>
 
 
 ###########################################################################
@@ -908,7 +917,7 @@
     [[[default]]]
       # Zookeeper ensemble. Comma separated list of Host/Port.
       # e.g. localhost:2181,localhost:2182,localhost:2183
-      host_ports=localhost:2181
+      host_ports=<%= @zookeeper_host_port %>
 
       # The URL of the REST contrib service (required for znode browsing)
       rest_url=http://localhost:9998
diff --git a/bigtop-deploy/puppet/modules/mahout/manifests/init.pp b/bigtop-deploy/puppet/modules/mahout/manifests/init.pp
index 0c55a9be..cfb08dfb 100644
--- a/bigtop-deploy/puppet/modules/mahout/manifests/init.pp
+++ b/bigtop-deploy/puppet/modules/mahout/manifests/init.pp
@@ -22,6 +22,8 @@ class mahout {
   }
 
   class client {
+    include hadoop::common
+
     package { "mahout":
       ensure => latest,
       require => Package["hadoop"],
diff --git a/bigtop-deploy/puppet/modules/qfs/README.md b/bigtop-deploy/puppet/modules/qfs/README.md
new file mode 100644
index 00000000..22254203
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/README.md
@@ -0,0 +1,80 @@
+QFS puppet module
+=================
+This module contains puppet recipes for various qfs components, e.g. metaserver,
+chunkserver, and the webui. It is responsible for the following items:
+    - Installing all required packages, configuration, init scripts, etc
+    - Wiring everything up so that each service can talk to the other
+
+hadoop-qfs
+==========
+Furthermore, this module also installs a compatibility wrapper script called
+hadoop-qfs to make the hadoop and qfs integration and interaction easier.
+
+In order to tell the main hadoop command to use qfs as the underlying
+filesystem, extra options must be specified. For example, to issue a `hadoop fs`
+command, the full command line would look like this:
+
+    $ JAVA_LIBRARY_PATH=/usr/lib/qfs hadoop fs
+        -Dfs.qfs.impl=com.quantcast.qfs.hadoop.QuantcastFileSystem \
+        -Dfs.default.name=qfs://localhost:20000 \
+        -Dfs.qfs.metaServerHost=localhost \
+        -Dfs.qfs.metaServerPort=20000 \
+        -ls /
+
+This (a) is cumbersome and (b) exposes low level details, e.g. metaserver port
+numbers, to the user who likely doesn't care. In order to avoid a poor user
+experience, we provide a wrapper script around the main hadoop command called
+`hadoop-qfs` which handles all the boilerplate for you. Using `hadoop-qfs`, the
+command above is reduced to the following:
+
+    $ hadoop-qfs fs -ls /
+
+The `hadoop-qfs` command also supports submitting jobs to hadoop which will use
+qfs as the underlying filesystem instead of the default HDFS. The process is the
+exact same as the standard `hadoop` command, except with using `hadoop-qfs`.
+
+    $ hadoop-qfs jar hadoop-mapreduce-examples.jar pi 100 100
+
+Any output data the job writes will be stored in qfs instead of HDFS. Note that
+when submitting jobs through the `hadoop-qfs` command, both the path to the jar
+file as well as the main class are required options.
+
+See the usage for `hadoop-qfs` (`hadoop-qfs --help`) for more information.
+
+Python
+======
+To use the qfs bindings in python, you must set the LD_LIBRARY_PATH to where the
+qfs shared objects are stored so that they can be found.
+
+    [root@bigtop1 /]# LD_LIBRARY_PATH=/usr/lib/qfs python
+    Python 2.6.6 (r266:84292, Jul 23 2015, 15:22:56)
+    [GCC 4.4.7 20120313 (Red Hat 4.4.7-11)] on linux2
+    Type "help", "copyright", "credits" or "license" for more information.
+    >>> import qfs
+    >>>
+
+Why Not Update Hadoop's `core-site.xml`?
+========================================
+The following is a little discussion on the rationale behind why this script has
+been implemented the way it has and some of the other options considered.
+
+One option is to include these configuration options in the core-site.xml file
+in a hadoop configuration directory. If we modify the core-site.xml file that
+the hadoop component installs, qfs silently takes precedence over hdfs. This
+results in a bad user experience since users should have a clear way to choose
+which filesystem they want to interact with.
+
+Another option is to provide our own core-site.xml specific to qfs in an
+alternate location and use the --config option to the hadoop command. This works
+but all of the other configuration in place for hadoop will either be missing or
+have to be symlinked in order to be carried over since we are overriding *all*
+configuration. This is annoying and brittle -- new configuration will have to be
+carried over whenever it is added.  Having two places where configuration needs
+to live is usually a bad idea.
+
+The final option considered, and the one used, is to provide a wrapper around
+the hadoop command, don't touch any hadoop configuration, and set the necessary
+override parameters on the command line here. This seemed the cleanest solution
+and also made sure that a qfs installation is as non-confrontational as
+possible. Unfortunately, the hadoop command is very picky about where the
+parameter overrides go depending on the subcommand used.
diff --git a/bigtop-deploy/puppet/modules/qfs/manifests/init.pp b/bigtop-deploy/puppet/modules/qfs/manifests/init.pp
new file mode 100644
index 00000000..c1602719
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/manifests/init.pp
@@ -0,0 +1,161 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+class qfs {
+  class deploy($roles) {
+    if ("qfs-metaserver" in $roles) {
+      include qfs::metaserver
+    }
+
+    if ("qfs-chunkserver" in $roles) {
+      include qfs::chunkserver
+    }
+
+    if ("qfs-client" in $roles) {
+      include qfs::client
+    }
+  }
+
+  class common($metaserver_host, $metaserver_port, $chunkserver_port,
+      $metaserver_client_port, $chunkserver_client_port) {
+    $cluster_key = sha1($metaserver_host)
+    $storage_dirs = suffix($hadoop::hadoop_storage_dirs, "/qfs")
+
+    hadoop::create_storage_dir { $qfs::common::storage_dirs: } ->
+    file { $qfs::common::storage_dirs:
+      ensure => directory,
+      owner => root,
+      group => root,
+      mode => 0755,
+    }
+  }
+
+  class metaserver {
+    include common
+
+    package { "qfs-metaserver":
+      ensure => latest,
+    }
+
+    $metaserver_conf = "/etc/qfs/MetaServer.prp"
+    file { $metaserver_conf:
+      content => template("qfs/MetaServer.prp"),
+      require => Package["qfs-metaserver"],
+    }
+
+    file { [
+      "${qfs::common::storage_dirs[0]}/metaserver",
+      "${qfs::common::storage_dirs[0]}/metaserver/transaction_logs",
+      "${qfs::common::storage_dirs[0]}/metaserver/checkpoint",
+    ]:
+      ensure => directory,
+      owner => qfs,
+      group => qfs,
+      mode => 0755,
+      before => Service['qfs-metaserver'],
+      require => [
+        File[$qfs::common::storage_dirs[0]],
+        Package['qfs-metaserver']
+      ],
+    }
+
+    exec { "mkfs":
+      command => "/usr/bin/metaserver -c $metaserver_conf",
+      creates => "${qfs::common::storage_dirs[0]}/metaserver/checkpoint/latest",
+      user => qfs,
+      group => qfs,
+      require => [
+        Package["qfs-metaserver"],
+        File[$metaserver_conf],
+        File["${qfs::common::storage_dirs[0]}/metaserver/checkpoint"],
+      ],
+    }
+
+    if ($fqdn == $qfs::common::metaserver_host) {
+      service { "qfs-metaserver":
+        ensure => running,
+        require => [
+          Package["qfs-metaserver"],
+          File[$metaserver_conf],
+          Exec["mkfs"],
+        ],
+        hasrestart => true,
+        hasstatus => true,
+      }
+    }
+  }
+
+  class chunkserver {
+    include common
+
+    package { "qfs-chunkserver":
+      ensure => latest,
+    }
+
+    $chunkserver_conf = "/etc/qfs/ChunkServer.prp"
+    file { $chunkserver_conf:
+      content => template("qfs/ChunkServer.prp"),
+      require => Package["qfs-chunkserver"],
+    }
+
+    $cs_dirs = suffix($hadoop::hadoop_storage_dirs, "/qfs/chunkserver")
+    $cs_chunks_dirs = suffix($hadoop::hadoop_storage_dirs, "/qfs/chunkserver/chunks")
+    $storage_dirs = concat($cs_dirs, $cs_chunks_dirs)
+
+    file { $storage_dirs:
+      ensure => directory,
+      owner => qfs,
+      group => qfs,
+      mode => 0755,
+      before => Service['qfs-chunkserver'],
+      require => [
+        File[$qfs::common::storage_dirs],
+        Package['qfs-chunkserver']
+      ],
+    }
+
+    service { "qfs-chunkserver":
+      ensure => running,
+      require => [
+        Package["qfs-chunkserver"],
+        File[$chunkserver_conf]
+      ],
+      hasrestart => true,
+      hasstatus => true,
+    }
+  }
+
+  class client {
+    include common
+
+    package { [
+      "qfs-client",
+      "qfs-hadoop",
+      "qfs-java",
+    ]:
+      ensure => latest,
+    }
+
+    file { "/etc/qfs/QfsClient.prp":
+      content => template("qfs/QfsClient.prp"),
+      require => Package["qfs-client"],
+    }
+
+    file { "/usr/bin/hadoop-qfs":
+      content => template("qfs/hadoop-qfs"),
+      mode => 0755,
+    }
+  }
+}
diff --git a/bigtop-deploy/puppet/modules/qfs/templates/ChunkServer.prp b/bigtop-deploy/puppet/modules/qfs/templates/ChunkServer.prp
new file mode 100644
index 00000000..84875aea
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/templates/ChunkServer.prp
@@ -0,0 +1,280 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# The chunk server configuration.
+
+# The following parameters set must be specified at startup time. Other
+# parameters can be changed at runtime, and it usually more convenient to
+# specify these in meta server configuration. The meta server broadcasts the
+# corresponding chunk server parameters to all connected chunk chunk servers.
+
+# The meta server location.
+chunkServer.metaServer.hostname = <%= scope['qfs::common::metaserver_host'] %>
+chunkServer.metaServer.port     = <%= scope['qfs::common::metaserver_port'] %>
+
+# Client connection listener ip address to bind to.
+# Use :: to bind to ipv6 address any.
+# Default is empty, treaated as 0.0.0.0 ipv4 address any, unless the following
+# parameter chunkServer.clientIpV6Only set to 1
+# chunkServer.clientIp =
+
+# Accept ipv4 chunk servers connections.
+# Default is 0, enable acception ipv4 connection. Only has effect if
+# chunkServer.clientIp left empty, or set to ::
+# chunkServer.clientIpV6Only = 0
+
+# Port to open for client connections
+chunkServer.clientPort = <%= scope['qfs::common::chunkserver_client_port'] %>
+
+# Chunk server's rack id. Has effect only if meta server rack prefixes
+# (metaServer.rackPrefixes) are nots set, or chunk server's ip do not match
+# any of the prefix.
+# Default is no rack assigned.
+# chunkServer.rackId = -1
+
+# Space separated list of directories to store chunks (blocks).
+# Usually one directory per physical disk. More than one directory can
+# be used in the cases where the host file system has problems / limitations
+# large directories.
+# The directories that are "not available" (don't exists, io errors,
+# "evacuate.done" file exists, etc.) at a given moment directories are
+# periodically scanned.
+# If the directory becomes available while chunk server is running, chunk server
+# deletes all chunk files in this directory (if any), and starts using this
+# directory.
+# All available directories are periodically scanned, if the directory becomes
+# "unavailable" all chunks in this directory are declared lost, and the
+# gets added to "not available" directories which are periodically scanned as
+# described the above.
+chunkServer.chunkDir = <%= scope['qfs::common::storage_dirs'][0] %>/chunkserver/chunks
+
+# Number of io threads (max. number of disk io requests in flight) per host file
+# system.
+# The typical setup is to have one host file system per physical disk.
+# Even if raid controller is available jbod configuration is recommended,
+# leaving the failure handling and striping to the distributed file system.
+# The default is 2.
+# With large requests (~1MB) two io requests in flight should be sufficient.
+# chunkServer.diskQueue.threadCount = 2
+
+# Number of "client" / network io threads used to service "client" requests,
+# including requests from other chunk servers, handle synchronous replication,
+# chunk re-replication, and chunk RS recovery. Client threads allow to use more
+# than one cpu to perform network io, encryption / decryption, request parsing,
+# checksum and RS recovery calculation.
+# Multiple "client" threads might help to increase network throughput. With
+# modern Intel CPUs, maximum single cpu throughput can be expected roughly
+# 300MB/sec (under 3Gbps network) with 1MB average write request size, and
+# approximately 150MB/sec with network encryption enabled.
+# This parameter has effect only on startup. Please see meta server
+# configuration for related parameters description. The parameters described in
+# the meta server configuration can be changed at any time, and the changes
+# will have effect without chunk server restart.
+# Default is 0. All network io, replication and recovery is handled by the
+# "main" thread.
+# chunkServer.clientThreadCount = 0
+
+# Set client thread affinity to CPU, starting from the specified CPU index. The
+# first cpu index is 0.
+# If the number of CPUs is less than start index plus the number of threads, the
+# the threads affinity at start index plus CPU count will be set to the last
+# CPU.
+# Setting affinity might help to reduce processor ram cache misses.
+# The parameter has effect only on startup, and has effect only on Linux OS.
+# Default is -1, no cpu affinity set.
+# chunkServer.clientThreadFirstCpuIndex = -1
+
+# Set the cluster / fs key, to protect against data loss and "data corruption"
+# due to connecting to a meta server hosting different file system.
+chunkServer.clusterKey = <%= scope['qfs::common::cluster_key'] %>
+
+# Redirect stderror and out into /dev/null to handle the case where one or both
+# are written into a file and the host file system / disk where the file resides
+# exhibiting noticeable io stalls, or completely unavailable..
+# Normally all the log message output performed by the message writer (thread)
+# that deals with log io stall by dropping log messages. This redirection is for
+# extra safety to handle the case if some library function attempting to write
+# into stdout / stderror.
+chunkServer.stdout = /dev/null
+chunkServer.stderr = /dev/null
+
+# The following parameter defines max size of io buffer memory used by the chunk
+# server.
+# The value set here, 128K means 128K * 4K buffer = 512M of buffers
+# The default values are 64K (128MB) for 32 bit build, and 192K (768MB) for 64
+# bit build.
+# The optimal amount of memory depends on the number of disks in the system, and
+# the io (read, write) concurrency -- the number of concurrent "clients". The
+# memory should be increased if large number of concurrent write appenders is
+# expected. Ideally the disk io request should be around 1MB, thus for each
+# chunk opened for append at least 1MB of io buffers is recommended.
+chunkServer.ioBufferPool.partitionBufferCount = 131072
+
+# Buffer manager portion of all io buffers.
+# This value defines max amount of io buffers that can be used for servicing
+# "client" requests, chunk re-replication, and recovery.
+# The remaining (1 - chunkServer.bufferManager.maxRatio) used for write append
+# buffering and other "internal" purposes.
+# Default is 0.4 or 40%
+# chunkServer.bufferManager.maxRatio = 0.4
+
+# Set the following to 1 if no backward compatibility with the previous kfs
+# releases required. 0 is the default.
+# When set to 0 the 0 header checksum (all 8 bytes must be 0) is treated as
+# no checksum and therefore no chunk file header checksum verification
+# performed.
+# The downside of the compatibility mode is that chunk server might not detect
+# the cases where the host os zero fills the data during the host file system
+# recovery / journal / transaction log replay,
+# thus the data loss / corruption problem might not be detected.
+# chunkServer.requireChunkHeaderChecksum = 0
+
+# If set to a value greater than 0 then locked memory limit will be set to the
+# specified value, and mlock(MCL_CURRENT|MCL_FUTURE) invoked.
+# On linux running under non root user setting locked memory "hard" limit
+# greater or equal to the specified value required. ulimit -l can be used for
+# example.
+# Default is 0 -- no memory locking.
+# chunkServer.maxLockedMemory = 0
+
+# Mlock io buffers memory at startup, if set to non 0.
+# Default is 0 -- no io buffer memory locking.
+# chunkServer.ioBufferPool.lockMemory = 0
+
+# ---------------------------------- Message log. ------------------------------
+
+# Set reasonable log level, and other message log parameter to handle the case
+# when meta server not available, or doesn't accept this chunk server for any
+# reason.
+# The chunk servers message log configuration parameters including log level
+# level can be changed in the meta server configuration file.
+chunkServer.msgLogWriter.logLevel = INFO
+
+# Colon (:) separated file name prefixes to store log segments.
+# Default is empty list. The default is to use file name from the command line
+# or if none specified write into file descriptor 2 -- stderror.
+# chunkServer.msgLogWriter.logFilePrefixes =
+
+# Maximum log segment size.
+# Default is -1 -- unlimited.
+# chunkServer.msgLogWriter.maxLogFileSize = -1
+
+# Maximum number of log segments.
+# Default is -1 -- unlimited.
+# chunkServer.msgLogWriter.maxLogFiles = -1
+
+# Max. time to wait for the log buffer to become available.
+# When wait is enabled the request processing thread will wait for the log
+# buffer disk io to complete. If the disk subsystem cannot handle the
+# amount of logging it will slow down the request processing.
+# For chunk servers keeping the default is strongly recommended to minimize
+# dependency on the host's disk subsystem reliability and performance.
+# Default is -1. Do not wait, drop log record instead.
+# chunkServer.msgLogWriter.waitMicroSec = -1
+
+# -------------------- Chunk and meta server authentication. -------------------
+# By default chunk and meta server authentication is off.
+#
+# If any of the following meta authentication methods is configured then chunk
+# server requires QFS client connection to be authenticated. In other words, the
+# QFS client, and other chunk server acting as a client, must obtain from the
+# meta server chunk server access token and corresponding key and use this token
+# and the key to authenticate with the chunk server.
+
+# ================= X509 authentication ========================================
+#
+# Chunk server's X509 certificate file in PEM format.
+# chunkserver.meta.auth.X509.X509PemFile =
+
+# Password if X509 PEM file is encrypted.
+# chunkserver.meta.auth.X509.X509Password =
+
+# Chunk server's private key file.
+# chunkserver.meta.auth.X509.PKeyPemFile =
+
+# Password if private key PEM file is encrypted.
+# chunkserver.meta.auth.X509.PKeyPassword =
+
+# Certificate authorities file. Used for both meta server certificate
+# validation and to create certificate chain with chunk server's X509
+# certificate.
+# chunkserver.meta.auth.X509.CAFile =
+
+# Certificate authorities directory can be used in addition to CAFile.
+# For more detailed information please see SSL_CTX_load_verify_locations manual
+# page. CAFile/CADir corresponds to CAfile/CApath in the man page.
+# chunkserver.meta.auth.X509.CADir =
+
+# If set (the default) verify peer certificate, and declare error if peer, i.e.
+# meta server, does not preset "trusted" valid X509 certificate.
+# Default is on.
+# chunkserver.meta.auth.X509.verifyPeer = 1
+
+# OpenSSL cipher configuration.
+# chunkserver.meta.auth.X509.cipher = !ADH:!AECDH:!MD5:HIGH:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to SSL_OP_NO_COMPRESSION
+# chunkserver.meta.auth.X509.options =
+
+# ================= Kerberos authentication ====================================
+#
+# Kerberos principal: service/host@realm
+
+# Meta server's Kerberos principal [service/host@realm] service name part.
+# chunkserver.meta.auth.krb5.service =
+
+# Meta server's Kerberos principal [service/host@realm] host name part.
+# chunkserver.meta.auth.krb5.host =
+
+# Kerberos keytab file with the key(s) that corresponds to the chunk server's
+# principal.
+# chunkserver.meta.auth.krb5.keytab =
+
+# Chunk server's kerberos principal. krb5_parse_name() used to convert the name
+# into the Kerberos 5 internal principal representation.
+# chunkserver.meta.auth.krb5.clientName =
+
+# Force Kerberos client cache initialization during intialization.
+# Default is off.
+# chunkserver.meta.auth.krb5.initClientCache = 0
+
+# OpenSSL cipher configuration for TLS-PSK authentication method. This method
+# is used with delegation and with Kerberos authentication.
+# chunkserver.meta.auth.psk.cipherpsk = !ADH:!AECDH:!MD5:!3DES:PSK:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to the logical OR of
+# SSL_OP_NO_COMPRESSION and SSL_OP_NO_TICKET
+# metaServer.clientAuthentication.psk.options =
+
+# ================= PSK authentication =========================================
+#
+# PSK chunk server authentication is intended only for testing and possibly for
+# small [test] clusters with very few chunk servers, where the same
+# authentication credentials [PSK "key"] are used for for all chunk servers.
+
+# Chunk server's PSK key, Must be identical to the key the meta server
+# configured with. See metaServer.CSAuthentication.psk.key parameter description
+# in the annotated meta server configuration file.
+# chunkserver.meta.auth.psk.key =
+
+# Chunk server's PSK key id. See metaServer.CSAuthentication.psk.keyId parameter
+# description in the annotated meta server configuration file.
+# chunkserver.meta.auth.psk.keyId =
+
+#-------------------------------------------------------------------------------
diff --git a/bigtop-deploy/puppet/modules/qfs/templates/MetaServer.prp b/bigtop-deploy/puppet/modules/qfs/templates/MetaServer.prp
new file mode 100644
index 00000000..0bcba747
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/templates/MetaServer.prp
@@ -0,0 +1,1179 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# The meta server configuration.
+
+# Client listener port.
+metaServer.clientPort = <%= scope['qfs::common::metaserver_client_port'] %>
+
+# Client connection listener ip address to bind to.
+# Use :: to bind to ipv6 address any.
+# Default is empty, treaated as 0.0.0.0 ipv4 address any, unless the following
+# parameter metaServer.clientIpV6Only set to 1
+# metaServer.clientIp =
+
+# Accept ipv4 client connections.
+# Default is 0, enable acception ipv4 connection. Only has effect if
+# metaServer.clientIp left empty, or set to ::
+# metaServer.clientIpV6Only = 0
+
+# Chunk server connection listener ip address to bind to.
+# Use :: to bind to ipv6 address any.
+# Default is empty, treaated as 0.0.0.0 ipv4 address any, unless the following
+# parameter metaServer.clientIpV6Only set to 1
+# metaServer.clientIp =
+
+# Accept ipv4 chunk servers connections.
+# Default is 0, enable acception ipv4 connection. Only has effect if
+# metaServer.clientIp left empty, or set to ::
+# metaServer.clientIpV6Only = 0
+
+# Chunk server listener port.
+metaServer.chunkServerPort = <%= scope['qfs::common::chunkserver_port'] %>
+
+# Meta serve transactions log directory.
+metaServer.logDir = <%= scope['qfs::common::storage_dirs'][0] %>/metaserver/transaction_logs
+
+# Meta server checkpoint directory.
+metaServer.cpDir = <%= scope['qfs::common::storage_dirs'][0] %>/metaserver/checkpoint
+
+# Allow to automatically create an empty file system if checkpoint file does
+# not exist.
+# The default is 0, as under the normal circumstances where the file system
+# content is of value, completely losing checkpoint, transaction log, and
+# automatically creating an empty fs will have the same effect as conventional
+# "mkfs". All chunks (blocks) will get deleted, and restoring the checkpoint
+# and logs later won't be sufficient to recover the data.
+# Use "-c" command line option to create new empty file system. For exampe:
+# metaserver -c MetaServer.prp
+# metaServer.createEmptyFs = 0
+
+# Root directory permissions -- used only when the new file system created.
+# metaServer.rootDirUser  = 0
+# metaServer.rootDirGroup = 0
+# metaServer.rootDirMode  = 0755
+
+# Defaults for checkpoint and transaction log without permissions conversion on
+# startup.
+# metaServer.defaultLoadUser     = 0
+# metaServer.defaultLoadGroup    = 0
+# metaServer.defaultLoadFileMode = 0644
+# metaServer.defaultLoadDirMode  = 0755
+
+# The size of the "client" thread pool.
+# When set to greater than 0, dedicated threads to do client network io, request
+# parsing, and response assembly are created. The thread pool size should
+# usually be (at least one) less than the number of cpus. "Client" threads help
+# with processing large amount of ["short"] requests where more cpu used for
+# context switch, network io, request parsing, and response assembly, than
+# the cpu for the request processing itself. For example i-node attribute
+# lookup, or write append chunk allocations that can be satisfied from the write
+# append allocation cache.
+# Default is 0 -- no dedicated "client" threads.
+# metaServer.clientThreadCount = 0
+
+# Meta server threads affinity.
+# Presently only supported on linux.
+# The first cpu index to set thread affinity to.
+# The main thread will be assigned to the cpu at the specified index, then the
+# next "client" thread will be assigned to the cpu index plus one and so on.
+# For example with 2 client threads and start cpu index 0 the threads affinity
+# would be 0 1 2 respectively.
+# Useful on machines with more than one multi-core processor with shared dram
+# cache. Assigning the threads to the same processor might help minimize dram
+# cache misses.
+# Default is off (start index less than 0) no thread affinity set.
+# metaServer.clientThreadStartCpuAffinity = -1
+
+# Meta server process max. locked memory.
+# If set to a value greater than 0 then locked memory limit will be set to the
+# specified value, and mlock(MCL_CURRENT|MCL_FUTURE) invoked.
+# On linux running under non root user setting locked memory "hard" limit
+# greater or equal to the specified value required. ulimit -l can be used for
+# example.
+# Default is 0 -- no memory locking.
+# metaServer.maxLockedMemory = 0
+
+# Size of [network] io buffer pool.
+# The default buffer size is 4K, therefore the amount of memory is
+# 4K * metaServer.bufferPool.partionBuffers.
+# All io buffers are allocated at startup.
+# If memory locking enabled io buffers are locked in memory at startup.
+# Default is 256K or 1GB on 64 bit system, and 32K or 128MB on 32 bit system.
+# metaServer.bufferPool.partionBuffers = 262144
+
+# ==============================================================================
+# The parameters below this line can be changed at runtime by editing the
+# configuration file and sending meta server process HUP signal.
+# Note that to restore parameter to default at run time the default value must
+# be explicitly specified in the configuration file. In other words commenting
+# out the parameter will not have any effect until restart.
+
+# WORM mode.
+# Write once, read many mode.
+# In this mode only modification of files ".tmp" (without quotes) suffix is
+# allowed.
+# Typically the application would create and write the file with ".tmp" suffix,
+# and then rename it so the destination file name will not have ".tmp " suffix.
+# To delete a file without ".tmp" suffix the mode can be temporary turned off
+# by the administrator. "qfstoggleworm" utility, or temporary configuration
+# modification can be used to do that.
+# Default is 0.
+# metaServer.wormMode = 0
+
+# Mininum number of connected / functional chunk servers before the file system
+# can be used.
+# Default is 1.
+# metaServer.minChunkservers = 1
+
+# Wait 30 sec for chunk servers to connect back after restarting, before file
+# system considered fully functional.
+metaServer.recoveryInterval = 30
+
+# Ignore master/slave chunk server assignment for write append.
+# Master/slave assignment can help with append replication 2, so avoid
+# theoretically possible IO buffers resource deadlock when chunk server A is a
+# in one "AB" synchronous append replication chain, and chunk server B is a 
+# master in another chunk "BA" synchronous replication.
+# In practice such deadlocks should be rare enough to matter, and, if occur,
+# are resolve with replication timeout mechanism.
+# The downside of using master/slave assignment is that presently it only works
+# with replication 3, and only half of the chunk server population will be
+# accepting client's append requests.
+# Default is "on" -- ignore.
+# metaServer.appendPlacementIgnoreMasterSlave = 1
+
+# For write append use the low order bit of the IP address for the chunk servers
+# master/slave assignment. This scheme is works well if least significant bit of
+# ip address uniformly distributes masters and slaves withing the rack,
+# especially with "in rack" placement for append.
+# Default is 0. Assign master / slave to keep number of masters and slaves
+# equal. The obvious downside of this is that the assignment depends on the
+# chunk servers connection order.
+# metaServer.assignMasterByIp = 0
+
+# Chunk server executables md5 sums white list.
+# The chunk server sends its executable md5sum when it connects to the meta
+# server. If the following space separated list is not empty and does not
+# does not contain the chunk server executable md5 sum then the chunk server
+# is instructed to exit or restart itself.
+# This might be useful for upgrade or versions control.
+# While the chunk server is connected to the meta server no md5 sum verification
+# performed.
+# Default is empty list.
+# metaServer.chunkServerMd5sums =
+
+# Unique file system id -- some name that uniquely identifies distributed file
+# system instance.
+# This is used to protect data loss / and or corruption in the case where chunk
+# server(s) connect to the "wrong" meta server.
+# The meta server will not accept connections from the chunk servers with a
+# different "cluster key".
+# Default is empty string.
+metaServer.clusterKey = <%= scope['qfs::common::cluster_key'] %>
+
+# Assign rack id by ip prefix -- ip address treated as string.
+# The prefix can be positioned with trailing ??
+# For example: 10.6.34.2?
+# The rack id assigned on chunk server connect, and will not change until the
+# chunk server re-connect. Therefore the configuration file changes will not
+# have any effect until the chunk servers re-connect.
+# Default is empty -- use rack id assigned in the chunk server config.
+# metaServer.rackPrefixes =
+# Example:
+# 10.6.1.* -- rack 1, 10.6.2.* -- rack 2, 10.6.4.1? -- rack 4 etc.
+# metaServer.rackPrefixes = 10.6.1. 1  10.6.2. 2  10.6.4.1? 4  10.6.4.1 5
+
+# "Static" placement weights of the racks. The more weight and more chunk
+# servers are in the rack the more likely the rack will be chosen for chunk
+# allocation.
+# Default is empty -- all weight are default to 1.
+# metaServer.rackWeights =
+# Example: Racks 1 and 2 have weight 1, rack 3 -- 0.9, rack 4 weight 1.2,
+# rack 5 weight 1.5. All other rack weights are 1.
+# metaServer.rackWeights = 1 1  2 1  3 0.9  4 1.2  5 1.5
+
+# Various timeout settings.
+
+# Extend write lease expiration time by 30 sec. in the case of the write master
+# disconnect, to give it a chance to re-connect.
+# Default is 30 sec. Production value is 60 sec.
+# metaServer.leaseOwnerDownExpireDelay = 30
+
+# Re-replication or recovery delay in seconds on chunk server down, to give
+# chunk server a chance to re-connect.
+# Default is 120 sec.
+# metaServer.serverDownReplicationDelay = 120
+
+# Chunk server heartbeat interval.
+# Default is 30 sec.
+# metaServer.chunkServer.heartbeatInterval = 30
+
+# Chunk server operations timeouts.
+# Heartbeat timeout results in declaring chunk server non operational, and
+# closing connection.
+# All other operations timeout are interpreted as the operation failure.
+# The values are in seconds.
+# The defaults:
+# metaServer.chunkServer.heartbeatTimeout    = 60
+# metaServer.chunkServer.chunkReallocTimeout = 75
+# metaServer.chunkServer.chunkAllocTimeout   = 40
+# metaServer.chunkServer.chunkReallocTimeout = 75
+# metaServer.chunkServer.makeStableTimeout   = 330
+# metaServer.chunkServer.replicationTimeout  = 330
+
+# The current production values.
+# metaServer.chunkServer.heartbeatInterval   = 18
+# metaServer.chunkServer.heartbeatTimeout    = 30
+# metaServer.chunkServer.chunkReallocTimeout = 18
+# metaServer.chunkServer.chunkAllocTimeout   = 18
+# metaServer.chunkServer.makeStableTimeout   = 60
+
+# Other chunk server operations timeout.
+# metaServer.chunkServer.requestTimeout      = 600
+
+# Chunk server space utilization placement threshold.
+# Chunk servers with space utilization over this threshold are not considered
+# as candidates for the chunk placement.
+# Default is 0.95 or 95%.
+# metaServer.maxSpaceUtilizationThreshold = 0.95
+
+# Unix style permissions
+# Space separated list of ip addresses of hosts where root user is allowed.
+# Empty list means that root user is allowed on any host.
+# Default is empty.
+# metaServer.rootHosts =
+
+# File modification time update resolution. Increasing the value will reduce the
+# transaction log writes with large files.
+# Default is 1 sec.
+# metaServer.MTimeUpdateResolution = 1
+
+# --------------- File create limits. ------------------------------------------
+#
+# Disallow specific file types. The list is space separate file type ids.
+# Default is empty list. All valid file types are allowed.
+# metaServer.createFileTypeExclude =
+
+# Limit number of data stripes for all file types.If create attempt exceeds
+# the limit the meta server returns "permission denied".
+# Default is the max supported by the compile time constants.
+# metaServer.maxDataStripeCount = 511
+
+# Limit number of recovery stripes for all file types. If create attempt exceeds
+# the limit the meta server returns "permission denied".
+# Default is 32.
+# Max supported by the compile time constants in common/kfstypes.h is 127.
+# metaServer.maxRecoveryStripeCount = 32
+
+# Limit number of data stripes for files with recovery.
+# Default is 64.
+# Max supported by the compile time constants in common/kfstypes.h is 511.
+# metaServer.maxRSDataStripeCount = 64
+
+# Max number of replicas for "regular / replicated" file with no recovery.
+# If create, or change replication requests exceeds this limit then the meta
+# server replaces the value with the value specified.
+# metaServer.maxReplicasPerFile   = 64
+
+# Max number of replicas for RS (file with recovery).
+# If create, or change replication requests exceeds this limit then the meta
+# server replaces the value with the value specified.
+# metaServer.maxReplicasPerRSFile = 64
+
+# Force effective user to root. This effectively turns off all permissions
+# control.
+# Default is off.
+# metaServer.forceEUserToRoot = 0
+
+# Client backward compatibility.
+# Defaults are no user and no group -- no backward compatibility.
+# metaServer.defaultUser     = 0xFFFFFFFF
+# metaServer.defaultGroup    = 0xFFFFFFFF
+# metaServer.defaultFileMode = 0644
+# metaServer.defaultDirMode  = 0755
+
+# The chunk server disconnects history size. Useful for monitoring.
+# Default is 4096 slots / disconnect events.
+# metaServer.maxDownServersHistorySize = 4096
+
+# Space and placement re-balancing.
+# Space re-balancing is controlled by the next two parameters (thresholds) below.
+# Re-balancing constantly scans all chunks in the system and checks chunk
+# placement within the replication or RS groups, and moves chunks from chunk
+# servers that are above metaServer.maxRebalanceSpaceUtilThreshold to the chunk
+# servers that are below metaServer.minRebalanceSpaceUtilThreshold.
+# Default is 1 -- on.
+# metaServer.rebalancingEnabled = 1
+
+# Space re-balancing thresholds.
+# Move chunk from the servers that exceed the
+# metaServer.maxRebalanceSpaceUtilThreshold
+# Default is 0.82
+# metaServer.maxRebalanceSpaceUtilThreshold = 0.82
+
+# Move chunks to server below metaServer.minRebalanceSpaceUtilThreshold.
+# Default is 0.72.
+# metaServer.minRebalanceSpaceUtilThreshold = 0.72
+
+# Time interval in seconds between replication queues scans.
+# The more often the scan is scheduled the more cpu can potentially use.
+# Default is 5 sec.
+# metaServer.replicationCheckInterval = 5
+
+# Re-balance scan depth.
+# Max number of chunks to scan in one partial scan. The more chunks are scanned
+# the more cpu re-balance will use, and the "faster" it will scan the chunks.
+# metaServer.maxRebalanceScan = 1024
+
+# Single re-balance partial scan time limit.
+# Default is 0.03 sec.
+# metaServer.maxRebalanceRunTime = 0.03
+
+# Minimum time between two consecutive re-balance partial scans.
+# Default is 0.512 sec.
+# metaServer.rebalanceRunInterval = 0.512
+
+# Max. number of a single client connection requests in flight.
+# The higher value might reduce cpu and alleviate "head of the line blocking"
+# when single client connection shared between multiple concurrent file readers
+# and writers, potentially at the cost of reducing "fairness" between the client
+# connections. Increasing the value could also reduce number of context
+# switches, and os scheduling overhead with the "client" threads enabled.
+# Default is 16 if the "client" threads are enabled, and 1 otherwise.
+# metaServer.clientSM.maxPendingOps = 16
+
+# ------------------ Chunk placement parameters --------------------------------
+
+# The metaServer.sortCandidatesByLoadAvg and
+# metaServer.sortCandidatesBySpaceUtilization are mutially exclusive.
+# metaServer.sortCandidatesBySpaceUtilization takes precedence over
+# metaServer.sortCandidatesByLoadAvg if both set to 1
+
+# When allocating (placing) a chunk prefer chunk servers with lower "load"
+# metric over the chunk servers with the higher "load" metric.
+# For the write intensive file systems turning this mode on is
+# recommended.
+# Default is 0. Do not take chunk server "load" metric into the account.
+# metaServer.sortCandidatesByLoadAvg = 0
+
+# When allocating (placing) a chunk prefer chunk servers with lower disk space
+# utilizaiton.
+# Default is 0. Do not take space utilization into the account.
+# metaServer.sortCandidatesBySpaceUtilization = 0
+
+# When allocating (placing) a chunk do not consider chunk server with the "load"
+# exceeding average load multiplied by metaServer.maxGoodCandidateLoadRatio.
+# Default is 4.
+# metaServer.maxGoodCandidateLoadRatio = 4
+
+# When allocating (placing) a chunk do not consider chunk server with the "load"
+# exceeding average "master" chunk server load multiplied by
+# metaServer.maxGoodMasterLoadRatio if the chunk server is used as master (head
+# or synchronous replication chain).
+# Default is 4.
+# metaServer.maxGoodMasterLoadRatio = 4
+
+# When allocating (placing) a chunk do not consider chunk server with the "load"
+# exceeding average "slave" load multiplied by metaServer.maxGoodSlaveLoadRatio
+# if the chunk server is used as slave.
+# Default is 4.
+# metaServer.maxGoodSlaveLoadRatio = 4
+
+# When allocating (placing) a chunk do not consider chunk server with the
+# average number of chunks opened for write per drive (disk) exceeding average
+# number of chunks opened for write across all disks / chunks servers multiplied
+# by metaServer.maxWritesPerDriveRatio.
+# Default is 1.5.
+# metaServer.maxWritesPerDriveRatio = 1.5
+
+# When allocating (placing) a chunk do not consider chunk server running on the
+# same host as writer if the average number of chunks opened for write per drive
+# (disk) exceeding average number of chunks opened for write across all disks /
+# chunks servers multiplied by metaServer.maxLocalPlacementWeight.
+# Default is 1.0.
+# metaServer.maxLocalPlacementWeight = 1.0
+
+# "In rack" placement for append and non append chunk allocations.
+# Place chunk replicas on the same rack to save cross rack bandwidth at the cost
+# of reduced reliability. Useful for temporary / scratch file systems.
+# Default is 0.
+# metaServer.inRackPlacementForAppend = 0
+
+# "In rack" placement for non append files.
+# Default is 0 - place replicas and chunks from the same RS blocks on different
+# racks.
+# metaServer.inRackPlacement = 0
+
+# Limit number of re-replications (this does not include RS chunk recovery),
+# that the given chunk server can be used as replication "source".
+# Default is 10.
+# metaServer.maxConcurrentReadReplicationsPerNode = 10
+
+# Limit max concurrent chunk re-replications and RS recoveries per chunk server.
+# Default is 5.
+# metaServer.maxConcurrentWriteReplicationsPerNode = 5
+
+#-------------------------------------------------------------------------------
+
+# Order chunk replicas locations by the chunk "load average" metric in "get
+# alloc" responses. The read client logic attempts to use replicas in this
+# order.
+# Default is 0. The replicas locations are shuffled randomly.
+# metaServer.getAllocOrderServersByLoad = 0
+
+# Delay recovery for the chunks that are past the logical end of file in files
+# with Reed-Solomon redundant encoding.
+# The delay is required to avoid starting recovery while the file is being
+# written into, and the chunk sizes aren't known / final. The writer can stop
+# writing into a file, and the corresponding chunks write leases might timed
+# out, and will be automatically revoked. The existing writer logic sets logical
+# EOF when it closes the file, before that the logical file size remains 0
+# during write. (Unless it is re-write which is currently for all practical
+# purposes not supported with RS files). The timeout below should be set to
+# at least the max. practical file "write" time.
+# Setting the timeout to a very large value will prevent processing the chunks
+# sitting in the replication delayed queue from the "abandoned" files, i.e.
+# files that the writer wrote something and then exited without closing the
+# file.
+# The parameter and the corresponding "delay" logic will likely be removed in
+# future releases, and replaced with the write lease renew logic.
+# Default is 6 hours or 21600 seconds.
+# metaServer.pastEofRecoveryDelay = 21600
+
+# Periodic checkpointing.
+# If set to -1 checkpoint is disabled. In such case "logcompactor" can be used
+# periodically create new checkpoint from the transaction logs.
+# Default is 3600 sec.
+# metaServer.checkpoint.interval = 3600
+
+# Checkpoint lock file name. Can be used to serialize checkpoint write and load
+# with external programs, for example logcompactor.
+# Default is empty -- no lock file used.
+# metaServer.checkpoint.lockFileName =
+
+# Max consecutive checkpoint write failures.
+# Meta server will exit if checkpoint write fails
+# metaServer.checkpoint.maxFailedCount times in the row for any reason (not
+# enough disk space for example).
+# Default is 2.
+# metaServer.checkpoint.maxFailedCount = 2
+
+# Checkpoint write timeout. Max time the checkpoint write can take before
+# declaring write failure.
+# Default is 3600 sec.
+# metaServer.checkpoint.writeTimeoutSec = 3600
+
+# Use synchronous mode to write checkpoint, i.e. tell host os to flush all data
+# to disk prior to write system call return.
+# The main purpose is to reduce the number of "dirty" / unwritten pages in the
+# host os vm subsystem / file system buffer cache, therefore reducing memory
+# contention and lowering the chances of paging out meta server and other
+# processes with no memory locking.
+# Default is on.
+# metaServer.checkpoint.writeSync = 1
+
+# Checkpoint write buffer size.
+# The buffer size should be adequate with synchrounous write mode enabled,
+# especially if journal and data of host's file system are on the same spinning
+# media device, in order to minimize the number of seeks.
+# Default is 16MB.
+# metaServer.checkpoint.writeBufferSize = 16777216
+
+# ---------------------------------- Audit log. --------------------------------
+
+# All request headers and response status are logged.
+# The audit log records are null ('\0') separated.
+# The log could be useful for debugging and audit purposes.
+# The logging require some cpu, but the main resource consumption is disk io.
+# Default is off.
+# metaServer.clientSM.auditLogging = 0
+
+# Colon (:) separated file name prefixes to store log segments.
+# Default is empty list.
+# metaServer.auditLogWriter.logFilePrefixes =
+
+# Maximum log segment size.
+# Default is -1 -- unlimited.
+# metaServer.auditLogWriter.maxLogFileSize = -1
+
+# Maximum number of log segments.
+# Default is -1 -- unlimited.
+# metaServer.auditLogWriter.maxLogFiles = -1
+
+# Max. time to wait for the log buffer to become available.
+# When wait is enabled the request processing thread will wait for the log
+# buffer disk io to complete. If the disk subsystem cannot keep up with the
+# logging it will slow down the meta server request processing.
+# Default is -1. Do not wait, drop log record instead.
+# metaServer.auditLogWriter.waitMicroSec = -1
+
+#-------------------------------------------------------------------------------
+
+# ---------------------------------- Message log. ------------------------------
+
+# Message log level FATAL, ALERT, CRIT, ERROR, WARN, NOTICE, INFO, DEBUG
+# Defaul is DEBUG, except for non debug builds with NDEBUG defined INFO is
+# default.
+metaServer.msgLogWriter.logLevel = INFO
+
+# Colon (:) separated file name prefixes to store log segments.
+# Default is empty list. The default is to use file name from the command line
+# or if none specified write into file descriptor 2 -- stderror.
+# metaServer.msgLogWriter.logFilePrefixes =
+
+# Maximum log segment size.
+# Default is -1 -- unlimited.
+# metaServer.msgLogWriter.maxLogFileSize = -1
+
+# Maximum number of log segments.
+# Default is -1 -- unlimited.
+# metaServer.msgLogWriter.maxLogFiles = -1
+
+# Max. time to wait for the log buffer to become available.
+# When wait is enabled the request processing thread will wait for the log
+# buffer disk io to complete. If the disk subsystem cannot keep up with the
+# logging it will slow down the meta server request processing.
+# Default is -1. Do not wait, drop log record instead.
+# metaServer.msgLogWriter.waitMicroSec = -1
+
+#-------------------------------------------------------------------------------
+
+# -------------------- Chunk servers authentication. ---------------------------
+#
+# Authentication is off by default. Both X509 (ssl) and Kerberos authentication
+# methods can be enabled at the same time. Chunk server can negotiate
+# authentication method. If both Kerberos and X509 are configured on the chunk
+# server and meta server then Kerberos authentication is used.
+# Chunk and meta servers perform mutual authentication with authentication
+# enabled.
+#
+# Use of X509 authentication is recommended in order to avoid KDC dependency.
+# Chunk servers have to periodically request Kerberos tickets from KDC. The meta
+# server enforces Kerberos ticket expiration time, by asking chunk server to
+# re-authenticate when its ticket expires. Therefore KDC unavailability for any
+# reason, including network communication outage, might result in chunk servers
+# disconnects. Long enough KDC unavailability might result in unrecoverable data
+# loss, due to the file system unability to perform replication and recovery
+# in response to disk and node failures.
+#
+# Please see OpenSSL documentation for detailed description about X509
+# authentication configuration.
+# src/test-scripts/qfsmkcerts.sh might be used as a simple example how to create
+# and use certificate authority, and X509 certificates.
+
+# Maximum authenticated session lifetime. This limits authenticated session time
+# for all authentication methods. In other words, the session [connection] must
+# be re-authenticated if the authentication token (Kerberos ticket, or x509
+# certificate) "end time" is reached or authenticated session exists longer than
+# the value of this parameter.
+# Default is 24 hours.
+# metaServer.clientAuthentication.maxAuthenticationValidTimeSec = 86400
+
+# Check chunk server authenticated name against the user and group database.
+# If enabled then the authenticated name must be present in the user database in
+# order for chunk server to be accepted.
+# Default is 0 (off), use only black and white lists, if configured, see below.
+# metaServer.CSAuthentication.useUserAndGrupDb = 0
+
+# ================= X509 authentication ========================================
+# Meta server's X509 certificate file in PEM format.
+# metaServer.CSAuthentication.X509.X509PemFile =
+
+# Password if X509 PEM file is encrypted.
+# metaServer.CSAuthentication.X509.X509Password =
+
+# Meta server's private key file.
+# metaServer.CSAuthentication.X509.PKeyPemFile =
+
+# Password if private key PEM file is encrypted.
+# metaServer.CSAuthentication.X509.PKeyPassword =
+
+# Certificate authorities file. Used for both chunk server certificate
+# validation and to create certificate chain with meta server's X509
+# certificate.
+# metaServer.CSAuthentication.X509.CAFile =
+
+# Certificate authorities directory can be used in addition to CAFile.
+# For more detailed information please see SSL_CTX_load_verify_locations manual
+# page. CAFile/CADir corresponds to CAfile/CApath in the man page.
+# metaServer.CSAuthentication.X509.CADir =
+
+# If set (the default) verify peer certificate, and declare error if peer, i.e.
+# chunk server, does not preset "trusted" valid X509 certificate.
+# Default is on.
+# metaServer.CSAuthentication.X509.verifyPeer = 1
+
+# OpenSSL cipher configuration.
+# metaServer.CSAuthentication.X509.cipher = !ADH:!AECDH:!MD5:HIGH:@STRENGTH
+
+# SSL/TLS session cache timeout. Session cache is only used with X509
+# authentication method, with non default client or server side openssl options
+# that turns off use of tls session tickets.
+# Default is 4 hours.
+# metaServer.CSAuthentication.X509.session.timeout = 14400
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to SSL_OP_NO_COMPRESSION
+# metaServer.clientAuthentication.X509.options =
+
+# ================= Kerberos authentication =====================================
+# Kerberos principal: service/host@realm
+
+# Meta server's Kerberos principal [service/host@realm] service name part.
+# metaServer.CSAuthentication.krb5.service =
+
+# Meta server's Kerberos principal [service/host@realm] host name part.
+# metaServer.CSAuthentication.krb5.host =
+
+# Kerberos keytab file with the key(s) that corresponds to the meta server's
+# principal.
+# metaServer.CSAuthentication.krb5.keytab =
+
+# Copy keytab into memory keytab, if supported by the kerberos versions, to
+# improve performance, and avoid disk access.
+# Default is on.
+# metaServer.CSAuthentication.krb5.copyToMemKeytab = 1
+
+# Client's (chunk server) principal "unparse" mode.
+# Can be set to space separated combination of the following modes:
+# short noRealm display
+# The result of the principal conversion to string is used as client's
+# (chunk server's) "authenticated name".
+# The default is fullly qualified principal name. For chunk servers it
+# would typically be in the form of service/host@realm.
+# The "unparsed" chunk server name is checked against "black" and "white" chunk
+# server list names as described below.
+# metaServer.CSAuthentication.krb5.princUnparseMode =
+
+# OpenSSL cipher configuration for TLS-PSK authentication method. This method
+# is used with TLS-PSK and with Kerberos authentication.
+# metaServer.CSAuthentication.psk.cipherpsk = !ADH:!AECDH:!MD5:!3DES:PSK:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to the logical OR of
+# SSL_OP_NO_COMPRESSION and SSL_OP_NO_TICKET
+# metaServer.CSAuthentication.psk.options =
+
+# ================= PSK authentication =========================================
+# PSK chunk server authentication is intended only for testing and possibly for
+# small [test] clusters with very few chunk servers, where the same
+# authentication credentials [PSK "key"] are used for for all chunk servers.
+
+# Chunk server PSK key id. This string sent to the chunk as TLS PSK "hint", and
+# also used as chunk server "authenticated name".
+# This effectively overrides chunk server key id.
+# If chunk server key id set to non empty string, then it can be left empty.
+# In such case chunk server key id is used as authenticated name. The chunk
+# server key id sent as "clear text" as part of ssl handshake, and is not
+# "tied" in any way known to the meta server logic to the key id, therefore any
+# "name" can be used. In other words the key is the only real security
+# "credential" with this authentication scheme.
+# The resulting chunk server name must not be empty, and pass "black" and
+# "white" list check, see below.
+# metaServer.CSAuthentication.psk.keyId =
+
+# Chunk server PSK key (the "pre-shared-key"). The same key must be used on the
+# chunk server side in order for psk authentication to work.
+# The default is empty key -- PSK authentication is not enabled.
+# The key must be base 64 encoded, i.e. it must be valid base 64 sequence.
+# metaServer.CSAuthentication.psk.key =
+
+# ================= Chunk servers's "black" and "white" lists ==================
+
+# Chunk server's X509 common names and/or kerberos names, "black" ("revocation")
+# list. If chunk server's authenticated name matches one of the name in this
+# list the authentication will fail. The names in the list are must be
+# separated by spaces. Names with white space symbols are not supported.
+# metaServer.CSAuthentication.blackList =
+
+# Chunk server's X509 common names and/or kerberos names, "white list". Unless
+# the list is empty the chunk server's authenticated name must match one of the
+# names in the list.
+# metaServer.CSAuthentication.whiteList =
+
+#-------------------------------------------------------------------------------
+
+# -------------------- User / "client" authentication. -------------------------
+# Client X509 and kerberos authentication parameters only differ from chunk
+# server's authentication parameters by metaServer.clientAuthentication prefix.
+# The defaults are identical to chunk server authentication.
+
+# Maximum authenticated session lifetime. This limits authenticated session time
+# for all authentication methods. In other words, the session [connection] must
+# be re-authenticated if the authentication token (delegation token, Kerberos
+# ticket, or x509 certificate) "end time" is reached or authenticated session
+# exists longer than the value of this parameter.
+# Default is 24 hours.
+# metaServer.clientAuthentication.maxAuthenticationValidTimeSec = 86400
+
+# ================= X509 authentication ========================================
+# Meta server's X509 certificate file in PEM format.
+# metaServer.clientAuthentication.X509.X509PemFile =
+
+# Password if X509 PEM file is encrypted.
+# metaServer.clientAuthentication.X509.X509Password =
+
+# Meta server's private key file.
+# metaServer.clientAuthentication.X509.PKeyPemFile =
+
+# Password if private key PEM file is encrypted.
+# metaServer.clientAuthentication.X509.PKeyPassword =
+
+# Certificate authorities file. Used for both chunk server certificate
+# validation and to create certificate chain with meta server's X509
+# certificate.
+# metaServer.clientAuthentication.X509.CAFile =
+
+# Certificate authorities directory can be used in addition to CAFile.
+# For more detailed information please see SSL_CTX_load_verify_locations manual
+# page. CAFile/CADir corresponds to CAfile/CApath in the manual page.
+# metaServer.clientAuthentication.X509.CADir =
+
+# If set (the default) verify peer certificate, and declare error if peer, i.e.
+# QFS client, does not preset certificate.
+# Default is on.
+# metaServer.clientAuthentication.X509.verifyPeer = 1
+
+# OpenSSL cipher configuration for X509 authentication method.
+# metaServer.clientAuthentication.X509.cipher = !ADH:!AECDH:!MD5:HIGH:@STRENGTH
+
+# SSL/TLS session cache timeout. Session cache is only used with X509
+# authentication method, with non default client or server side openssl options
+# that turns off use of tls session tickets.
+# Default is 4 hours.
+# metaServer.clientAuthentication.X509.session.timeout = 14400
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to SSL_OP_NO_COMPRESSION
+# metaServer.clientAuthentication.X509.options =
+
+# ================= Kerberos authentication =====================================
+# Kerberos principal: service/host@realm
+
+# Meta server's Kerberos principal [service/host@realm] service name part.
+# metaServer.clientAuthentication.krb5.service =
+
+# Meta server's Kerberos principal [service/host@realm] host name part.
+# metaServer.clientAuthentication.krb5.host =
+
+# Kerberos keytab file with the key(s) that corresponds to the meta server's
+# principal.
+# metaServer.clientAuthentication.krb5.keytab =
+
+# Copy keytab into memory keytab, if supported by the kerberos versions, to
+# improve performance, and avoid disk access.
+# Default is on.
+# metaServer.clientAuthentication.krb5.copyToMemKeytab = 1
+
+# Client's principal "unparse" mode.
+# Can be set to space separated combination of the following modes:
+# short noRealm display
+# The result of the principal conversion to string is used as client's
+# (client's) "authenticated name".
+# The default is fully qualified principal name. For users this typically would
+# it would be in the form user@realm
+# The resulting authentication name should match password database, the meta
+# server host uses. The recommended value is "short', discard realm if it
+# matches the default the kerberos configuration's default realm.
+# metaServer.clientAuthentication.krb5.princUnparseMode =
+
+# OpenSSL cipher configuration for TLS-PSK authentication method. This method
+# is used with delegation and with Kerberos authentication.
+# metaServer.clientAuthentication.psk.cipherpsk = !ADH:!AECDH:!MD5:!3DES:PSK:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to the logical OR of
+# SSL_OP_NO_COMPRESSION and SSL_OP_NO_TICKET
+# metaServer.clientAuthentication.psk.options =
+
+# The following two parameters and respective defaults are intended to allow
+# non authenticated access for the meta server web UI from the local host.
+
+# Space separated list of host ips that RPC listed in the next parameter are
+# permitted with no authentication.
+# Default is 127.0.0.1
+# metaServer.clientAuthentication.noAuthOpsHostIps = 127.0.0.1
+
+# Space separated list of RPC names that allowed with no authentication, if the
+# client's host ip obtained with getpeername() call matches one of the ips in
+# the preceding list.
+# Default is RPCs used by the meta server web UI.
+# metaServer.clientAuthentication.noAuthOps = PING GET_CHUNK_SERVERS_COUNTERS GET_CHUNK_SERVER_DIRS_COUNTERS GET_REQUEST_COUNTERS DISCONNECT
+
+# ================= Client's "black" and "white" lists =========================
+
+# Client's (user) X509 common names and/or kerberos names, "black"
+# ("revocation") list. If client's authenticated name matches one of the name in
+# this list the authentication will fail. The names in the list are must be
+# separated by spaces. Names with white space symbols are not supported.
+# metaServer.clientAuthentication.blackList =
+
+# Client's X509 common names and/or kerberos names, "white list". Unless
+# the list is empty the client's authenticated name must match one of the
+# names in the list.
+# metaServer.clientAuthentication.whiteList =
+
+# ================== Delegation ================================================
+#
+# Delegation token expiration time limit.
+# Default is 24 hours.
+# metaServer.clientAuthentication.maxDelegationValidForTimeSec = 86400
+
+# Do not limit delegation token end time to the meta server session credentials
+# (Kerberos ticket or X509 certificate) end time.
+# Default is 0.
+# metaServer.clientAuthentication.delegationIgnoreCredEndTime = 0
+
+================================================================================
+
+# Allow to use "clear text" communication mode by performing SSL/TLS shutdown
+# immediately after successful authentication completion. If enabled, the QFS
+# client's corresponding setting defines the communication mode between the
+# client and the write master. The "clear text" communication mode between chunk
+# servers (synchronous replication, re-replication, and chunk recovery) will be
+# used if this parameter set to "on".
+# Using this mode might make sense in order reduce chunk server CPU utilization
+# and/or possibly increase IO throughput, in the cases where chunk server
+# communication channel is considered to have adequate security for the purpose
+# at hands.
+# Default if "off" / "no"
+# metaServer.clientCSAllowClearText = 0
+
+# Chunk server access token maximum lifetime.
+# Chunk server access token time defines chunk access time limit.
+# Chunk access tokens have 10 min time limit -- twice chunk lease time. The
+# chunk server access token effectively defines maximum client and chunk server
+# to chunk server connections lifetimes. The client and chunk servers attempt
+# to obtain and use a new chunk server access token before the current token
+# expires, and re-open connection with the newly obtained token.
+# Default is 2 hours.
+# metaServer.CSAccessValidForTimeSec = 7200
+
+# The meta server limits the write lease end time to the max of the current time
+# plus the value of the following parameter, and the authentication end time.
+# The parameter is intended primarily for testing, to avoid spurious write
+# retries with authentication maximum life time set to very small value -- 5sec.
+# (The short authentication lifetime is used in order to test the
+# re-authentication logic.)
+# metaServer.minWriteLeaseTimeSec = 600
+
+#-------------------------------------------------------------------------------
+
+# -------------------- User and group configuration. ---------------------------
+# User and group database parameters.
+# The meta server host's user and group configuration used for QFS file system
+# when QFS authentication is enabled. The user and group database is used to map
+# "authenticated names" obtained with Kerberos and X509 authentication methods
+# to user and group ids, and to establish group membership. Authenticated names
+# that have no corresponding user id, or user id that have no corresponding
+# "user name" are considered invalid, and as the result the authentication
+# fails.
+# User and group id with value 4294967295 have special treatment. Access always
+# denied for users with such id.
+# Root user entry with name "root" and id 0 added if not present in the the user
+# database, unless expliclity excluded with metaServer.userAndGroup.excludeUser
+# parameter.
+# With authentication enabled QFS client library does not use host's local user
+# and group database, the meta server's host database is effectively used by all
+# QFS clients.
+
+# Minimal user id to include in user name to id mapping.
+# Default is 0.
+# metaServer.userAndGroup.minUserId = 0
+
+# Maximum user id to include in user name to id mapping.
+# Default is 4294967295.
+# metaServer.userAndGroup.maxUserId = 4294967295
+
+# Minimal group id to include in group name to group id mapping.
+# Default is 0.
+# metaServer.userAndGroup.minGroupId = 0
+
+# Maximum group id to include in group name to group id mapping.
+# Default is 4294967295.
+# metaServer.userAndGroup.maxGroupId = 4294967295
+
+# Omit entries with user names if it has one of the specified prefixes.
+# metaServer.userAndGroup.omitUserPrefix =
+
+# Omit entries with group names if it has one of the specified prefixes.
+# Default is empty list.
+# metaServer.userAndGroup.omitGroupPrefix =
+
+# Update / re-read user and group to id mappings with every N seconds.
+# By default period updates are effectively disabled. The parameter reload with
+# HUP signal can be used to trigger user and group information update.
+# Default is 315360000.
+# metaServer.userAndGroup.updatePeriodSec = 315360000
+
+# Disable user and group initial loading and/or reloading.
+# Default is enabled.
+# metaServer.userAndGroup.disable = 0
+
+# Space separated list of the user names to exclude when loading or updating
+# user database.
+# Default is empty list.
+# metaServer.userAndGroup.excludeUser =
+
+# Space separated list of the group names to exclude when loading or updating
+# group database.
+# Default is empty list.
+# metaServer.userAndGroup.excludeGroup =
+
+# Space separated list of the group names, where members of these groups
+# have effective user id 0 -- root.
+# Default is empty list.
+# metaServer.userAndGroup.rootGroups =
+
+# Space separated list of the user names, where such users have effective user
+# id 0 -- root.
+# User with name root and id 0 always added, even if it isn't presnet or
+# excluded from the user database.
+# Default is empty list.
+# metaServer.userAndGroup.rootUsers =
+
+# Space separated list of the user names. Specified users are allowed to
+# perform meta server administrative requests: fsck, chunk server retire,
+# toggle worm, recompute directory sizes, dump to chunk to servers map,
+# dump replication candidates, check chunk leases, list open files.
+# Default is root user.
+# metaServer.userAndGroup.metaServerAdminUsers = root
+
+# Space separated list of group names. Members of these groups are allowed to
+# perform meta server aministration described in the previous parameter's
+# section.
+# Default is empty list.
+# metaServer.userAndGroup.metaServerAdminGroups =
+
+# Space separated list of the user names. Specified users are allowed to
+# perform meta server status inquiry requests: ping, up servers, meta stats, get
+# chunk servers counters, get chunk directory counters, get meta server request
+# counters.
+# Default is root user.
+# metaServer.userAndGroup.metaServerStatsUsers = root
+
+# Space separated list of group names. Members of these groups are allowed to
+# perform meta server status requests described in the previous parameter's
+# section.
+# Default is empty list.
+# metaServer.userAndGroup.metaServerStatsGroups =
+
+# Space separated list of the user names. Specified users are allowed to
+# renew and cancel delegation tokens that belong to other users. No delegation
+# can be used, to perform delegation renew or cancel, i.e. Kerberos or X509
+# authentication methods must be used.
+# Default is empty list.
+# metaServer.userAndGroup.delegationRenewAndCancelUsers =
+
+# Space separated list of group names. Members of these groups are allowed to
+# renew and cancel delegation tokens that belong to other users, as described
+# in the previous parameter's section.
+# metaServer.userAndGroup.delegationRenewAndCancelGruops =
+
+# -------------------- Meta server cryptographic keys. -------------------------
+
+# Key lifetime. The value defines maximum time before the delegation tokens
+# issued by the meta server expire, and have to be renewed.
+# Default is 4 hours.
+# metaServer.cryptoKeys.keyValidTimeSec = 14400
+
+# Keys change period.
+# Key lifetime minus key change period is the minim time before delegation token
+# must be renewed.
+# metaServer.cryptoKeys.keyChangePeriodSec = 7200
+
+# Meta server crypto keys file name.
+# File name to save meta server keys.
+# Specify file name to save keys, in order to ensure that delegation tokens are
+# persistent across meta server restarts.
+# Default is none, no keys are not persistent across meta server restarts.
+# metaServer.cryptoKeys.keysFileName =
+
+# ------------------- Meta server authentication override ----------------------
+#
+# Setting the following 3 parameters to the values specified below, will
+# effectively disable client authentication. The chunk server authentication can
+# still be enabled.
+# Like the other parameters, removing or commenting out these parameters will not
+# turn back QFS client authentication on, until chunk and meta servers restart.
+# To turn back the QFS client authentication the parameters should be explicitly
+# set back to the original / default values.
+#
+# Overriding the default behavior might be useful for initial authentication
+# setup and/or debugging. Only two sets of values, the default and the set of
+# inverted values work, any other combinations will not though could be useful
+# for testing, i.e. 1 0 0 0 and 0 1 1 1
+#
+# Default is 1 if QFS client authentication *not* configured, 0 otherwise.
+# metaServer.clientAuthentication.authNone = 1
+#
+# Default is 0 if QFS client authentication *not* configured, 1 otherwise.
+# metaServer.clientCSAuthRequired          = 0
+#
+# Default is 0 if chunk and meta server server authentication is *not*
+# configured, 1 otherwise.
+# chunkServer.client.auth.enabled          = 0
+#
+# Default is 0 if chunk and meta server server authentication is *not*
+# configured, 1 otherwise.
+# chunkServer.remoteSync.auth.enabled      = 0
+
+# -------------------- File system ID ------------------------------------------
+#
+# File system id is 64 bit file system identifier generated by the meta server
+# at the time the file system is created. The file system id is used to protect
+# against accidental use of chunk files that belong to a different file system.
+# For example file system id should prevent use of stale chunk inventory that
+# belongs to "old"/different file system, in the case where a new file file
+# file system was created, and the same "cluster key" is used.
+
+# Require file system id in the chunk server hello.
+# Default is 0, to maintain backward compatibility with file systems crated with
+# no file system id.
+# metaServer.fileSystemIdRequired = 0
+
+# The following parameter might be used for temporary file systems, that might
+# be intentionally re-created from scratch on meta server restart, or in the
+# case of loss of transaction log and/or checkpoint. Use with extra caution.
+# Default is 0, Do not delete chunks on file system id mismatch. Do not use
+# chunk directories if chunk directory file system does not match.
+# metaServer.deleteChunkOnFsIdMismatch = 0
+
+#-------------------------------------------------------------------------------
+
+# ===================== Chunk servers configuration parameters. ================
+# Configuration parameters in the meta server configuration file take precedence
+# over the chunk server configuration files.
+
+# ---------------------------------- Message log. ------------------------------
+
+# Chunk server log level.
+chunkServer.msgLogWriter.logLevel = NOTICE
+
+# Colon (:) separated file name prefixes to store log segments.
+# Default is empty list. The default is to use file name from the command line
+# or if none specified write into file descriptor 2 -- stderror.
+# chunkServer.msgLogWriter.logFilePrefixes =
+
+# Maximum log segment size.
+# Default is -1 -- unlimited.
+# chunkServer.msgLogWriter.maxLogFileSize = -1
+
+# Maximum number of log segments.
+# Default is -1 -- unlimited.
+# chunkServer.msgLogWriter.maxLogFiles = -1
+
+# Max. time to wait for the log buffer to become available.
+# When wait is enabled the request processing thread will wait for the log
+# buffer disk io to complete. If the disk subsystem cannot keep up with the
+# logging it will slow down the request processing.
+# For chunk servers keeping the default is strongly recommended to minimize
+# dependency on the host's disk subsystem reliability and performance.
+# Default is -1. Do not wait, drop log record instead.
+# chunkServer.msgLogWriter.waitMicroSec = -1
+
+#-------------------------------------------------------------------------------
+
+# Disk io request timeout.
+# Default is 270 sec. Production value is 40 sec.
+# chunkServer.diskIo.maxIoTimeSec = 270
+
+# Synchronous replication timeouts.
+# Record append synchrounous replicaiton timeout.
+# Default is 180 sec. Production value is 20 sec.
+# chunkServer.recAppender.replicationTimeoutSec = 180
+# Write replication timeout.
+# Default is 300 sec. Production value is 20 sec.
+# chunkServer.remoteSync.responseTimeoutSec = 300
+
+# Controls buffered io -- use os file system cache, instead of direct io on the
+# os / file systems that support direct io (most file systems on linux).
+# Default is off.
+# It is conceivable that enabling buffered io might help with short reads for
+# the "broadcast" / "web server" type of loads. For the "typical" large io (1MB)
+# requests sequential type loads enabling caching will likely lower cluster
+# performance due to higher system (os) cpu overhead, and memory contention.
+# Default is off.
+# chunkServer.bufferedIo = 0
+
+# If sparse files, and in particular chunks aren't used (sequential write only
+# for example) the following parameter can be set to 0.
+# Default is 1 -- enabled.
+# chunkServer.allowSparseChunks = 1
+
+# The minimal amount of space in bytes that must be available in order for the
+# chunk directory to be used for chunk placement (considered as "writable").
+# Default is chunk size -- 64MB plus chunk header size 16KB.
+# chunkServer.minFsAvailableSpace = 67125248
+
+# The minimal amount of space that must be available in order for the chunk
+# directory to be used for chunk placement (considered as "writable"), expressed
+# as part total host file system space.
+# Default is 0.05 or 5%, or in other words stop using chunk directory when the
+# host file system where the chunk directory resides reaches 95% space
+# utilization.
+# chunkServer.maxSpaceUtilizationThreshold = 0.05
+
+# The "weight" of pending disk io in chunk placement.
+# If set to 0 or less the pending io (number of io bytes in the disk queue) has
+# no no effect on the placement (choosing chunk directory where to create chunk).
+# If weight set to greater than 0, then the average pending io per chunk
+# directory (host file system / disk) is calculated, as
+# (total_pending_read_bytes * total_pending_read_weight +
+# total_pending_write_bytes * total_pending_write_weight) / chunk_directory_count
+# Chunk directories with pending_read + pending_write that exceed the value the
+# above are taken out of the consideration for placement.
+# Default is 0. Typical production value is 1.3
+# chunkServer.chunkPlacementPendingReadWeight  = 0
+# chunkServer.chunkPlacementPendingWriteWeight = 0
+
+# Averaging interval for calculating the average time the incoming "client's"
+# requests spend in io buffer wait queue. The "average wait time" value used by
+# the meta server for chunk placement. The average exponentially decays (IIR
+# filter).
+# Default is 20 sec. Typical production value is 8.
+# chunkServer.bufferManager.waitingAvgInterval  = 20
+
+# "Not available" directories rescan interval in seconds. Default is 180 sec.
+# (see comment in chunk server configuration file).
+# chunkServer.dirRecheckInterval = 60
+
+# The following parameter has effect only if client threads enabled, i.e. if
+# chunkServer.clientThreadCount parameter set to a value greater than 0 in the
+# chunk server configuration.
+# If the value is less than chunkServer.clientThreadCount, then
+# threads in range
+# [chunkServer.client.firstClientThreadIndex, chunkServer.clientThreadCount)
+# will be used to service all requests, except RS chunk recovery, otherwise the
+# "main" thread will be used.
+# chunkServer.client.firstClientThreadIndex = 0
+
+# The following parameter has effect only if client threads enabled, i.e. if
+# chunkServer.clientThreadCount parameter set to a value greater than 0 in the
+# chunk server configuration.
+# Limit number of client threads used for RS recovery. Each thread uses
+# single dedicated connection to the meta server to perform RS recovery.
+# If set to 0 or less, then the "main" thread is used to perform RS recovery.
+# The client threads in the range
+# [0, min(chunkServer.rsReader.maxRecoveryThreads, chunkServer.clientThreadCount)
+# are used to preform RS recovery.
+# Default is 5. The same value as
+# metaServer.maxConcurrentWriteReplicationsPerNode default.
+# chunkServer.rsReader.maxRecoveryThreads = 5
diff --git a/bigtop-deploy/puppet/modules/qfs/templates/QfsClient.prp b/bigtop-deploy/puppet/modules/qfs/templates/QfsClient.prp
new file mode 100644
index 00000000..94ae962b
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/templates/QfsClient.prp
@@ -0,0 +1,137 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# The meta server configuration.
+
+# Where is the metaserver
+metaServer.name = <%= scope['qfs::common::metaserver_host'] %>
+metaServer.port = <%= scope['qfs::common::metaserver_port'] %>
+
+# -------------------- Client and meta server authentication. ------------------
+# By default QFS client and meta server authentication (client and chunk server
+# authentication as a consequence) is off.
+#
+# If any of the following meta authentication method is configured then QFS
+# client and the meta server perform mutual authentication.
+#
+# The QFS client configuration parameters can be specified also via environment
+# variables: QFS_CLIENT_CONFIG and QFS_CLIENT_CONFIG and
+# QFS_CLIENT_CONFIG_meta_server_ip_port. The later variable takes precedence.
+# The dots in the meta server ip (or host name) are replaced with _ (underscore)
+# symbols. The underscore symbol also used to separate meta server ip and port.
+# The later, longer form allows to use configuration specific to a
+# particular meta server, and mainly intended to be used with the QFS
+# delegation where both the delegation token and the key can be passed via
+# environment variables (see PSK authentication section below)),
+#
+# The two from environment vairable values are supported:
+# 1. FILE:configuration_file_name
+# 2. parameter_name1=parameter_value1 parameter_name2=parameter_value2...
+# The second space separated key value pairs can be used to pass delegation
+# token and the corresponding key. Both these must be obtained from the meta
+# server via "delegate" request. See qfs tool help.
+# For example:
+# QFS_CLIENT_CONFIG_127_0_0_1_20000='client.auth.psk.keyId=AAAB9dYIWfKBXhXCI1jJ9gAAU0XunwAAAACMoK0z30ztT5S7k9slRuRdzy9CXmi1 client.auth.psk.keyId=P+4XRIBLLBvkICXWO+1aXBPUTMghEakkTk1T+RVsifR9NQ71E32KVd27y+2DbyC2'
+# export QFS_CLIENT_CONFIG_127_0_0_1_20000
+
+
+# ================= X509 authentication ========================================
+#
+# QFS client's X509 certificate file in PEM format.
+# client.auth.X509.X509PemFile =
+
+# Password if X509 PEM file is encrypted.
+# client.auth.X509.X509Password =
+
+# QFS client's private key file.
+# client.auth.X509.PKeyPemFile =
+
+# Password if private key PEM file is encrypted.
+# client.auth.X509.PKeyPassword =
+
+# Certificate authorities file. Used for both meta server certificate
+# validation and to create certificate chain with QFS client's X509
+# certificate.
+# client.auth.X509.CAFile =
+
+# Certificate authorities directory can be used in addition to CAFile.
+# For more detailed information please see SSL_CTX_load_verify_locations manual
+# page. CAFile/CADir corresponds to CAfile/CApath in the man page.
+# client.auth.X509.CADir =
+
+# If set (the default) verify peer certificate, and declare error if peer, i.e.
+# meta server, does not preset "trusted" valid X509 certificate.
+# Default is on.
+# client.auth.X509.verifyPeer = 1
+
+# OpenSSL cipher configuration.
+# client.auth.X509.cipher = !ADH:!AECDH:!MD5:HIGH:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to SSL_OP_NO_COMPRESSION
+# client.auth.X509.options =
+
+# ================= Kerberos authentication ====================================
+#
+# Kerberos service principal: service/host@realm
+
+# Meta server's Kerberos principal [service/host@realm] service name part.
+# client.auth.krb5.service =
+
+# Meta server's Kerberos principal [service/host@realm] host name part.
+# client.auth.krb5.host =
+
+# Normally kinit is sufficient for the user authentication.
+# The following Kerberos parameters might be used in the case when another
+# "service" acts as QFS client.
+
+# Kerberos keytab file with the key(s) that corresponds to the QFS client's
+# principal, if used. Key table is typically used for service.
+# client.auth.krb5.keytab =
+
+# QFS client's kerberos principal. krb5_parse_name() used to convert the name
+# into the Kerberos 5 internal principal representation.
+# client.auth.krb5.clientName =
+
+# Force Kerberos client cache initialization during intialization.
+# Default is off.
+# client.auth.krb5.initClientCache = 0
+
+# OpenSSL cipher configuration for TLS-PSK authentication method. This method
+# is used with delegation and with Kerberos authentication.
+# client.auth.psk.cipherpsk = !ADH:!AECDH:!MD5:!3DES:PSK:@STRENGTH
+
+# The long integer value passed to SSL_CTX_set_options() call.
+# See open ssl documentation for details.
+# Default is the integer value that corresponds to the logical OR of
+# SSL_OP_NO_COMPRESSION and SSL_OP_NO_TICKET
+# metaServer.clientAuthentication.psk.options =
+
+# ================= PSK / delegation authentication ============================
+#
+# Both delegation token and delegation key are expected to be valid base 64
+# encoded binary blobs -- the exact string representation returned by the
+# delegation request.
+
+# QFS client delegation token, The token must be obtained via delegation request
+# the meta server. Both the token and the corresponding key must be specified.
+# client.auth.psk.keyId =
+
+# QFS client delegation key, The key must be obtained via delegation request to
+# the meta server.
+# client.auth.psk.key =
+
+#-------------------------------------------------------------------------------
diff --git a/bigtop-deploy/puppet/modules/qfs/templates/hadoop-qfs b/bigtop-deploy/puppet/modules/qfs/templates/hadoop-qfs
new file mode 100644
index 00000000..7041daf2
--- /dev/null
+++ b/bigtop-deploy/puppet/modules/qfs/templates/hadoop-qfs
@@ -0,0 +1,79 @@
+#!/bin/bash
+#
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# A simple wrapper script around hadoop which enables the use of qfs easily.
+# Using the hadoop command with qfs requires users to specify overrides for
+# various parameters (e.g. fs.defaultFS) which can become cumbersome to do on
+# the command line every time. The goal here is to provide a clean user
+# experience in using hadoop and qfs.
+
+export JAVA_LIBRARY_PATH=/usr/lib/qfs
+
+QFS_OPTS=" -Dfs.qfs.impl=com.quantcast.qfs.hadoop.QuantcastFileSystem"
+QFS_OPTS="$QFS_OPTS -Dfs.defaultFS=qfs://<%= scope['qfs::common::metaserver_host'] %>:<%= scope['qfs::common::metaserver_client_port'] %>"
+QFS_OPTS="$QFS_OPTS -Dfs.qfs.metaServerHost=<%= scope['qfs::common::metaserver_host'] %>"
+QFS_OPTS="$QFS_OPTS -Dfs.qfs.metaServerPort=<%= scope['qfs::common::metaserver_port'] %>"
+
+usage() {
+  echo "
+usage: $0 <options>
+  $0 is a simple wrapper around the standard hadoop command that allows for an
+  easy interface with qfs. A subset of the standard hadoop commands are
+  supported (where it makes sense). For all other commands, use the hadoop
+  command directly.
+
+  Commands:
+     fs                             query the qfs filesystem
+     jar <jar> <main class>         submit a job to hadoop that uses qfs
+                                        Note: the main class is required
+     distcp <srcurl> <desturl>      copy file or directories recursively
+                                        Note: supports copy into qfs from hdfs
+  "
+  exit 1
+}
+
+if [ $# == 0 ]; then
+    usage
+fi
+
+COMMAND=$1
+shift
+
+case $COMMAND in
+    # usage flags
+    --help|-h)
+        usage
+        ;;
+    fs|distcp)
+        hadoop $COMMAND $QFS_OPTS "$@"
+        ;;
+    jar)
+        if [ $# -lt 2 ]; then
+            echo "$COMMAND: jar file and main class are required"
+            usage
+        fi
+
+        JAR_FILE=$1
+        MAIN_CLASS=$2
+        shift 2
+        hadoop $COMMAND $JAR_FILE $MAIN_CLASS $QFS_OPTS "$@"
+        ;;
+    *)
+        echo "$COMMAND: unsupported -- try using the hadoop command directly"
+        usage
+        ;;
+esac
diff --git a/bigtop-deploy/puppet/modules/tachyon/manifests/init.pp b/bigtop-deploy/puppet/modules/tachyon/manifests/init.pp
deleted file mode 100644
index ef7e5dfa..00000000
--- a/bigtop-deploy/puppet/modules/tachyon/manifests/init.pp
+++ /dev/null
@@ -1,79 +0,0 @@
-# The ASF licenses this file to You under the Apache License, Version 2.0
-# (the "License"); you may not use this file except in compliance with
-# the License.  You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-class tachyon {
-
-  class deploy ($roles) {
-    if ("tachyon-master" in $roles) {
-      include tachyon::master
-    }
-
-    if ("tachyon-worker" in $roles) {
-      include tachyon::worker
-    }
-  }
-
-  class common ($master_host){
-    package { "tachyon-tfs":
-      ensure => latest,
-    }
-
-    # add logging into /var/log/..
-    file {
-        "/etc/tachyon/conf/log4j.properties":
-        content => template("tachyon/log4j.properties"),
-        require => [Package["tachyon-tfs"]]
-    }
-
-    # add tachyon-env.sh to point to tachyon master
-    file { "/etc/tachyon/conf/tachyon-env.sh":
-        content => template("tachyon/tachyon-env.sh"),
-        require => [Package["tachyon-tfs"]]
-    }
-  }
-
-  class master {
-    include common
-
-   exec {
-        "tachyon formatting":
-           command => "/usr/lib/tachyon/bin/tachyon format",
-           require => [ Package["tachyon-tfs"], File["/etc/tachyon/conf/log4j.properties"], File["/etc/tachyon/conf/tachyon-env.sh"] ]
-    }
-
-    if ( $fqdn == $tachyon::common::master_host ) {
-      service { "tachyon-master":
-        ensure => running,
-        require => [ Package["tachyon-tfs"], Exec["tachyon formatting"] ],
-        hasrestart => true,
-        hasstatus => true,
-      }
-    }
-
-  }
-
-  class worker {
-    include common
-
-   if ( $fqdn == $tachyon::common::master_host ) {
-      notice("tachyon ---> master host")
-      # We want master to run first in all cases
-      Service["tachyon-master"] ~> Service["tachyon-worker"]
-   }
-
-    service { "tachyon-worker":
-      ensure => running,
-      require => [ Package["tachyon-tfs"], File["/etc/tachyon/conf/log4j.properties"], File["/etc/tachyon/conf/tachyon-env.sh"] ],
-      hasrestart => true,
-      hasstatus => true,
-    }
-  }
-}
diff --git a/bigtop-deploy/puppet/modules/tachyon/templates/tachyon-env.sh b/bigtop-deploy/puppet/modules/tachyon/templates/tachyon-env.sh
deleted file mode 100755
index e3a5fb1b..00000000
--- a/bigtop-deploy/puppet/modules/tachyon/templates/tachyon-env.sh
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env bash
-# Licensed to the Apache Software Foundation (ASF) under one or more
-# contributor license agreements.  See the NOTICE file distributed with
-# this work for additional information regarding copyright ownership.
-# The ASF licenses this file to You under the Apache License, Version 2.0
-# (the "License"); you may not use this file except in compliance with
-# the License.  You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-# This file contains environment variables required to run Tachyon. Copy it as tachyon-env.sh and
-# edit that to configure Tachyon for your site. At a minimum,
-# the following variables should be set:
-#
-# - JAVA_HOME, to point to your JAVA installation
-# - TACHYON_MASTER_ADDRESS, to bind the master to a different IP address or hostname
-# - TACHYON_UNDERFS_ADDRESS, to set the under filesystem address.
-# - TACHYON_WORKER_MEMORY_SIZE, to set how much memory to use (e.g. 1000mb, 2gb) per worker
-# - TACHYON_RAM_FOLDER, to set where worker stores in memory data
-# - TACHYON_UNDERFS_HDFS_IMPL, to set which HDFS implementation to use (e.g. com.mapr.fs.MapRFileSystem,
-#   org.apache.hadoop.hdfs.DistributedFileSystem)
-
-# The following gives an example:
-
-if [[ `uname -a` == Darwin* ]]; then
-  # Assuming Mac OS X
-  export JAVA_HOME=${JAVA_HOME:-$(/usr/libexec/java_home)}
-  export TACHYON_RAM_FOLDER=/Volumes/ramdisk
-  export TACHYON_JAVA_OPTS="-Djava.security.krb5.realm= -Djava.security.krb5.kdc="
-else
-  # Assuming Linux
-  if [ -z "$JAVA_HOME" ]; then
-    export JAVA_HOME=/usr/lib/jvm/java-7-oracle
-  fi
-  export TACHYON_RAM_FOLDER=/mnt/ramdisk
-fi
-
-export JAVA="$JAVA_HOME/bin/java"
-
-echo "Starting tachyon w/ java = $JAVA "
-
-export TACHYON_MASTER_ADDRESS=<%= @master_host %>
-export TACHYON_UNDERFS_ADDRESS=$TACHYON_HOME/underfs
-#export TACHYON_UNDERFS_ADDRESS=hdfs://localhost:9000
-export TACHYON_WORKER_MEMORY_SIZE=1GB
-export TACHYON_UNDERFS_HDFS_IMPL=org.apache.hadoop.hdfs.DistributedFileSystem
-
-echo "TACHYON master => $TACHYON_MASTER_ADDRESS "
-
-CONF_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
-
-export TACHYON_JAVA_OPTS+="
-  -Dlog4j.configuration=file:$CONF_DIR/log4j.properties
-  -Dtachyon.debug=false
-  -Dtachyon.underfs.address=$TACHYON_UNDERFS_ADDRESS
-  -Dtachyon.underfs.hdfs.impl=$TACHYON_UNDERFS_HDFS_IMPL
-  -Dtachyon.data.folder=$TACHYON_UNDERFS_ADDRESS/tmp/tachyon/data
-  -Dtachyon.workers.folder=$TACHYON_UNDERFS_ADDRESS/tmp/tachyon/workers
-  -Dtachyon.worker.memory.size=$TACHYON_WORKER_MEMORY_SIZE
-  -Dtachyon.worker.data.folder=$TACHYON_RAM_FOLDER/tachyonworker/
-  -Dtachyon.master.worker.timeout.ms=60000
-  -Dtachyon.master.hostname=$TACHYON_MASTER_ADDRESS
-  -Dtachyon.master.journal.folder=$TACHYON_HOME/journal/
-  -Dorg.apache.jasper.compiler.disablejsr199=true
-  -Djava.net.preferIPv4Stack=true
-"
-
-# Master specific parameters. Default to TACHYON_JAVA_OPTS.
-export TACHYON_MASTER_JAVA_OPTS="$TACHYON_JAVA_OPTS"
-
-# Worker specific parameters that will be shared to all workers. Default to TACHYON_JAVA_OPTS.
-export TACHYON_WORKER_JAVA_OPTS="$TACHYON_JAVA_OPTS"
diff --git a/bigtop-deploy/puppet/modules/zeppelin/templates/zeppelin-env.sh b/bigtop-deploy/puppet/modules/zeppelin/templates/zeppelin-env.sh
index 8095f010..ebcfc06d 100644
--- a/bigtop-deploy/puppet/modules/zeppelin/templates/zeppelin-env.sh
+++ b/bigtop-deploy/puppet/modules/zeppelin/templates/zeppelin-env.sh
@@ -18,6 +18,7 @@ export ZEPPELIN_WEBSOCKET_PORT=<%= @web_socket_port %>
 export ZEPPELIN_CONF_DIR=/etc/zeppelin/conf
 export ZEPPELIN_LOG_DIR=/var/log/zeppelin
 export ZEPPELIN_PID_DIR=/var/run/zeppelin
+export ZEPPELIN_WAR_TEMPDIR=/var/run/zeppelin/webapps
 export ZEPPELIN_NOTEBOOK_DIR=/var/lib/zeppelin/notebook
 export MASTER=<%= @spark_master_url %>
 export SPARK_HOME=/usr/lib/spark
diff --git a/bigtop-deploy/vm/utils/setup-env-debian.sh b/bigtop-deploy/vm/utils/setup-env-debian.sh
index 3d34fe12..ac92615d 100755
--- a/bigtop-deploy/vm/utils/setup-env-debian.sh
+++ b/bigtop-deploy/vm/utils/setup-env-debian.sh
@@ -29,6 +29,7 @@ if [ $enable_local_repo == "true" ]; then
     echo "deb file:///bigtop-home/output/apt bigtop contrib" > /etc/apt/sources.list.d/bigtop-home_output.list
     apt-get update
 else
+    apt-get install -y apt-transport-https
     echo "local apt = $enable_local_repo ; NOT Enabling local apt. Packages will be pulled from remote..."
 fi
 
diff --git a/bigtop-deploy/vm/utils/smoke-tests.sh b/bigtop-deploy/vm/utils/smoke-tests.sh
index 8dac31c0..93f795e6 100755
--- a/bigtop-deploy/vm/utils/smoke-tests.sh
+++ b/bigtop-deploy/vm/utils/smoke-tests.sh
@@ -36,9 +36,17 @@ export SQOOP_HOME=/usr/lib/sqoop/
 export HIVE_CONF_DIR=/etc/hive/conf/
 export MAHOUT_HOME="/usr/lib/mahout"
 
-su -s /bin/bash $HCFS_USER -c '/usr/bin/hadoop fs -mkdir /user/vagrant /user/root'
-su -s /bin/bash $HCFS_USER -c 'hadoop fs -chmod 777 /user/vagrant'
-su -s /bin/bash $HCFS_USER -c 'hadoop fs -chmod 777 /user/root'
+prep() {
+    HADOOP_COMMAND=$1
+    su -s /bin/bash $HCFS_USER -c "JAVA_LIBRARY_PATH=/usr/lib/qfs $HADOOP_COMMAND fs -mkdir /user/vagrant /user/root"
+    su -s /bin/bash $HCFS_USER -c "JAVA_LIBRARY_PATH=/usr/lib/qfs $HADOOP_COMMAND fs -chmod 777 /user/vagrant"
+    su -s /bin/bash $HCFS_USER -c "JAVA_LIBRARY_PATH=/usr/lib/qfs $HADOOP_COMMAND fs -chmod 777 /user/root"
+}
+
+prep hadoop
+if [[ $SMOKE_TESTS == *"qfs"* ]]; then
+    prep hadoop-qfs
+fi
 
 if [ -f /etc/debian_version ] ; then
     apt-get -y install pig hive flume mahout sqoop
diff --git a/bigtop-deploy/vm/vagrant-puppet-docker/README.md b/bigtop-deploy/vm/vagrant-puppet-docker/README.md
index 79e50f78..602b10b0 100644
--- a/bigtop-deploy/vm/vagrant-puppet-docker/README.md
+++ b/bigtop-deploy/vm/vagrant-puppet-docker/README.md
@@ -157,4 +157,4 @@ See `bigtop-deploy/puppet/config/site.csv.example` for more details.
 
 ##Notes
 
-* Users currently using vagrant 1.6+ is strongly recommanded to upgrade to 1.6.4+, otherwise you will encounter the [issue](https://github.com/mitchellh/vagrant/issues/3769) when installing plguins
+* Users currently using vagrant 1.6+ is strongly recommanded to upgrade to 1.6.4+, otherwise you will encounter the [issue](https://github.com/mitchellh/vagrant/issues/3769) when installing plugins
diff --git a/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_centos-7.yaml b/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_centos-7.yaml
index 49de5737..92d7468f 100644
--- a/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_centos-7.yaml
+++ b/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_centos-7.yaml
@@ -21,7 +21,7 @@ boot2docker:
         memory_size: "4096"
         number_cpus: "1"
 
-repo: "http://bigtop-repos.s3.amazonaws.com/releases/1.1.0/centos/6/x86_64"
+repo: "http://bigtop-repos.s3.amazonaws.com/releases/1.1.0/centos/7/x86_64"
 distro: centos
 components: [hadoop, yarn]
 namenode_ui_port: "50070"
diff --git a/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_ubuntu_ppc64le.yaml b/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_ubuntu_ppc64le.yaml
new file mode 100644
index 00000000..d34ba914
--- /dev/null
+++ b/bigtop-deploy/vm/vagrant-puppet-docker/vagrantconfig_ubuntu_ppc64le.yaml
@@ -0,0 +1,32 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+docker:
+        memory_size: "4096"
+        image:  "bigtop/deploy:ubuntu-15.04-ppc64le"
+
+boot2docker:
+        memory_size: "4096"
+        number_cpus: "1"
+
+repo: "http://bigtop-repos.s3.amazonaws.com/releases/1.1.0/ubuntu/vivid/ppc64el"
+distro: debian
+components: [hadoop, yarn]
+namenode_ui_port: "50070"
+yarn_ui_port: "8088"
+hbase_ui_port: "60010"
+enable_local_repo: false
+smoke_test_components: [mapreduce, pig]
+jdk: "openjdk-7-jdk"
diff --git a/bigtop-deploy/vm/vagrant-puppet-openstack/vagrantconfig.yaml b/bigtop-deploy/vm/vagrant-puppet-openstack/vagrantconfig.yaml
index 76633557..86077ddd 100644
--- a/bigtop-deploy/vm/vagrant-puppet-openstack/vagrantconfig.yaml
+++ b/bigtop-deploy/vm/vagrant-puppet-openstack/vagrantconfig.yaml
@@ -12,7 +12,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-repo: "http://bigtop01.cloudera.org:8080/view/Releases/job/Bigtop-0.8.0/label=centos6/6/artifact/output/"
+repo: "http://bigtop-repos.s3.amazonaws.com/releases/1.1.0/centos/6/x86_64"
 num_instances: 1
 distro: centos
 components: [hadoop, yarn]
diff --git a/bigtop-deploy/vm/vagrant-puppet-vm/README.md b/bigtop-deploy/vm/vagrant-puppet-vm/README.md
index 2b21a706..ed06eca0 100644
--- a/bigtop-deploy/vm/vagrant-puppet-vm/README.md
+++ b/bigtop-deploy/vm/vagrant-puppet-vm/README.md
@@ -59,7 +59,7 @@ num_instances: 5
 first, build up local yum repo
 
 ```
-cd bigtop; ./gradlew tachyon-yum
+cd bigtop; ./gradlew alluxio-yum
 ```
 
 and then enable local yum in vagrantconfig.yaml
diff --git a/bigtop-deploy/vm/vagrant-puppet-vm/Vagrantfile b/bigtop-deploy/vm/vagrant-puppet-vm/Vagrantfile
index 63aff201..b4017d0e 100755
--- a/bigtop-deploy/vm/vagrant-puppet-vm/Vagrantfile
+++ b/bigtop-deploy/vm/vagrant-puppet-vm/Vagrantfile
@@ -67,6 +67,7 @@ bigtop_master = "bigtop1.vagrant"
 
 $script = <<SCRIPT
 service iptables stop
+service firewalld stop
 chkconfig iptables off
 # Remove 127.0.0.1 entry since vagrant's hostname setting will map it to FQDN,
 # which miss leads some daemons to bind on 127.0.0.1 instead of public or private IP
author	ganeshraju <ganeshraju@gmail.com>	2016-06-21 16:10:43 -0500
committer	ganeshraju <ganeshraju@gmail.com>	2016-06-21 16:10:43 -0500
commit	9d62f51ec3684c75d0fa4b8d2da9301e0a18f661 (patch)
tree	2a255c6f11a6358f76e701b1357ee8a758d18998 /bigtop-deploy
parent	cff413decf877c03c670de67e002bf1d6d15a5fc (diff)