| commit | f2775f7f34cc8898abd0caff12e868b8d34b8ad6 | [log] [tgz] |
|---|---|---|
| author | Mike Walch <mwalch@gmail.com> | Thu Aug 11 15:00:26 2016 -0400 |
| committer | Mike Walch <mwalch@gmail.com> | Fri Aug 12 11:17:28 2016 -0400 |
| tree | f37f59f224f414735fc49e021e2a3fada7457c3c | |
| parent | 1becb112906eab1114c37eceaee2ff2fe345fba8 [diff] |
Several minor updates to Muchos * Upgrade Hadoop and Spark versions * Starting using Spark dist without Hadoop binaries * Cleaned up accumulo configuration
Muchos supports deploying Apache Fluo and its dependencies to a cluster that can be optionally launched in Amazon EC2. Apache Fluo depends on Apache Accumulo, Apache Zookeeper, and Apache Hadoop. Setting up these dependencies is time consuming. Muchos provides a set of helper scripts to automate setting up these dependencies on a cluster. This makes it quick for a developer to experiment with Fluo in a realistic environment.
Muchos is intended to help developers experiment with and test Fluo and Fluo applications in a realistic environment. Muchos is not recommended for setting up production environments as it has no support for updating and upgrading dependencies. Also, Muchos has a wipe command thats great for testing and very dangerous in production environments.
Muchos is structured into two high level components :
Checkout Uno for setting up Fluo's dependencies on a single machine.
First clone the Muchos repo:
git clone https://github.com/astralway/muchos.git
Now, create and modify your configuration file for Muchos:
cd muchos/ cp conf/muchos.props.example conf/muchos.props
In order to run the muchos command, your AWS credentials need to be set in muchos.props like this:
[ec2] aws.access.key_id=AKIAIOSFODNN7EXAMPLE aws.secret.key=wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY
See AWS Key ID Documentation for more information.
You will need to upload your public key to the AWS management console and set key.name in muchos.props to the name of your key pair. If you want to give others access to your cluster, add their public keys to a file named keys in your conf/ directory. During the setup of your cluster, this file will be appended on each node to the ~/.ssh/authorized_keys file for the user set by the cluster.username property.
When Muchos launches a cluster, it uses a free CentOS 7 image that is hosted in the AWS marketplace but managed by the CentOS orginization. If you have never used this image in EC2 before, you will need to go to the CentOS 7 product page to accept the software terms under the ‘Manual Launch’ tab. If this is not done, you will get an error when you try to launch your cluster.
The CentOS organization periodically updates AMIs and deprecates older AMIs which makes them unavailable to new users. This can also cause an error when you try to launch your cluster. If this occurs, you will need to find the AMI ID for your EC2 region on the CentOS 7 product page and set the ‘aws_ami’ property in your ‘muchos.props’ file to override the default AMIs used by Muchos.
Run the following command to launch an EC2 cluster called mycluster:
muchos launch -c mycluster
After your cluster has launched, you do not have to specify a cluster anymore using -c (unless you have multiple clusters running).
Run the following command to confirm that you can ssh to the leader node:
muchos ssh
You can check the status of the nodes using the EC2 Dashboard or by running the following command:
muchos status
The muchos setup command will set up your cluster and start Hadoop, Zookeeper, & Accumulo. It will download release tarballs of Fluo, Accumulo, Hadoop, etc. The release versions of these tarballs are specified in muchos.props.
Optionally, you can have Muchos use a snapshot version (rather than a released version) of Accumulo or Fluo by building a snapshot tarball and placing it in the conf/upload directory before running muchos setup. This option is only necessary if you want to run the latest unreleased version of Fluo or Accumulo.
# optional, example commands to build a snapshot version of Fluo cd /path/to/fluo mvn package cp modules/distribution/target/fluo-1.0.0-beta-3-SNAPSHOT-bin.tar.gz /path/to/muchos/conf/upload/
The muchos setup command will install and start Accumulo, Hadoop, and Zookeeper. The optional services below will only be set up if configured in the [nodes] section of muchos.props:
fluo - Fluo only needs to be installed and configured on a single node in your cluster as Fluo applications are run in YARN. If set as a service, muchos setup will install and partially configure Fluo but not start it. To finish setup, follow the steps in the ‘Run a Fluo application’ section below.
metrics - The Metrics service installs and configures collectd, InfluxDB and Grafana. Cluster metrics are sent to InfluxDB using collectd and are viewable in Grafana. If Fluo is running, its metrics will also be viewable in Grafana.
mesosmaster - If specified, a Mesos master will be started on this node and Mesos slaves will be started on all workers nodes. The Mesos status page will be viewable at http://<MESOS_MASTER_NODE>:5050/. Marathon will also be started on this node and will be viewable at http://<MESOS_MASTER_NODE>:8080/.
If you run the muchos setup command and a failure occurs, you can run the command again with no issues. Any cluster setup that was successfully completed will not be repeated. While some setup steps can take over a minute, you can use Ctrl-C to stop setup if it hangs for a long time. Just remember to run muchos setup again to finish setup.
The setup command is idempotent. It can be run again on a working cluster. It will not change the cluster if everything is configured and running correctly. If a process has stopped, the setup command will restart the process.
The muchos wipe command can be used to wipe all data from the cluster and kill any running processes. After running the wipe command, run the setup command to start a fresh cluster.
If you set proxy_socks_port in your muchos.props, a SOCKS proxy will be created on that port when you use muchos ssh to connect to your cluster. If you add a proxy managment tool to your browser and whitelist http://leader*, http://worker* and http://metrics* to redirect traffic to your proxy, you can view the monitoring & status pages below in your browser. Please note - The hosts in the URLs below match the configuration in [nodes] of muchos.prop.example and may be different for your cluster.
mesosmaster configured on leader1)mesosmaster configured on leader1)Running an example Fluo application like Webindex, Phrasecount, or Stresso is easy with Muchos as it configures your shell with common environment variables. To run an example application, SSH to to a node on cluster where Fluo is installed and clone the example repo:
muchos ssh # SSH to cluster proxy node ssh <node where Fluo is installed> # Nodes with Fluo installed is determined by Muchos config hub clone astralway/webindex # Clone repo of example application. Press enter for user/password.
Start the example application using its provided scripts. To show how simple this can be, commands to run the Webindex application are shown below. Read the Webindex README to learn more before running these commands.
cd webindex/ ./bin/webindex init # Initialize and start webindex Fluo application ./bin/webindex getpaths 2015-18 # Retrieves CommonCrawl paths file for 2015-18 crawl ./bin/webindex load-s3 2015-18 0-9 # Load 10 files into Fluo in the 0-9 range of 2015-18 crawl ./bin/webindex ui # Runs the WebIndex UI
If you have your own application to run, you can follow the instructions starting at the Configure a Fluo application section of the Fluo production setup instructions to configure, initialize, and start your application. To automate these steps, you can mimic the scripts of example Fluo applicaitons above.
After muchos setup is run, users can install additional software on the cluster using their own Ansible playbooks. In their Ansible playbooks, users can reference any configuration in the Ansible inventory file at /etc/ansible/hosts which is set up by Muchos on the proxy node. The inventory file lists the hosts for services on the cluster such as the Zookeeper nodes, Namenode, Accumulo master, etc. It also has variables in the [all:vars] section that contain settings that may be useful in user playbooks. It is recommended that any user-defined Ansible playbooks should be managed in their own git repository (see mikewalch/muchos-custom for an example).
If you launched your cluster on EC2, run the following command terminate your cluster. WARNING - All data on your cluster will be lost:
muchos terminate
The config command allows you to retrieve cluster configuration for your own scripts:
$ muchos config -p leader.public.ip 10.10.10.10
Muchos is powered by the following projects:
muchos launch to start a cluster in AWS EC2.muchos setup to install, configure, and start Fluo, Accumulo, Hadoop, etc on an existing EC2 or bare metal cluster.Install nose using pip:
pip install nose
The following command runs the unit tests:
nosetests -w bin/impl