search

apache / fluo-muchos

Apache Fluo Muchos
https://fluo.apache.org
Apache License 2.0
26 stars 37 forks source link
accumulo ansible aws azure big-data fluo hacktoberfest

readme

Muchos

Build Status Apache License

Muchos automates setting up Apache Accumulo or Apache Fluo (and their dependencies) on a cluster

Muchos makes it easy to launch a cluster in Amazon's EC2 or Microsoft Azure and deploy Accumulo or Fluo to it. Muchos enables developers to experiment with Accumulo or Fluo in a realistic, distributed environment. Muchos installs all software using tarball distributions which makes its easy to experiment with the latest versions of Accumulo, Hadoop, Zookeeper, etc without waiting for downstream packaging.

Muchos is not recommended at this time for production environments as it has no support for updating and upgrading dependencies. It also has a wipe command that is great for testing but dangerous for production environments.

Muchos is structured into two high level components:

Check out Uno for setting up Accumulo or Fluo on a single machine.

Requirements

Common

Muchos requires the following common components for installation and setup:

EC2

Muchos requires the following for EC2 installations:

Azure

Muchos requires the following for Azure installations:

When running Muchos under Ubuntu 18.04, checkout these tips.

Quickstart

The following commands will install Muchos, launch a cluster, and setup/run Accumulo:

git clone https://github.com/apache/fluo-muchos.git

cd fluo-muchos/
cp conf/muchos.props.example conf/muchos.props
vim conf/muchos.props                                   # Edit to configure Muchos cluster
./bin/muchos launch -c mycluster                        # Launches Muchos cluster in EC2 or Azure
./bin/muchos setup                                      # Set up cluster and start Accumulo

The launch command will create a cluster with the name specified in the command (e.g. 'mycluster'). The setup command can be run repeatedly to fix any failures and will not repeat successful operations.

After your cluster is launched, SSH to it using the following command:

./bin/muchos ssh

Run the following command to terminate your cluster. WARNING: All cluster data will be lost.

./bin/muchos terminate

Please continue reading for more detailed Muchos instructions.

Launching an EC2 cluster

Before launching a cluster, you will need to complete the requirements above, clone the Muchos repo, and create muchos.props. 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.

Configuring the AMI

You might also need to configure the aws_ami property in muchos.props. Muchos by default uses a Fedora 35 image for EC2. By default, the aws_ami property is set to this Fedora 35 AMI in us-east-1. You will need to change this value if a newer image has been released or if you are running in different region than us-east-1.

Launching the cluster

After following the steps above, run the following command to launch an EC2 cluster called mycluster:

./bin/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:

./bin/muchos ssh

You can check the status of the nodes using the EC2 Dashboard or by running the following command:

./bin/muchos status

Launching an Azure cluster

Before launching a cluster, you will need to complete the requirements for Azure above, clone the Muchos repo, and create your conf/muchos.props file by making a copy of the muchos.props example. 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. You will also need to ensure you have authenticated to Azure and set the target subscription using the Azure CLI.

Muchos by default uses an AlmaLinux 9 image that is hosted in the Azure marketplace. The Azure Linux Agent is already pre-installed on the Azure Marketplace images and is typically available from the distribution's package repository.

Edit the values in the sections within muchos.props as below Under the general section, edit following values as per your configuration

Under the azure section, edit following values as per your configuration:

Please refer to the muchos.props example for the full list of Azure-specific configurations - some of which have supplementary comments.

Within Azure the nodes section is auto populated with the hostnames and their default roles.

After following the steps above, run the following command to launch an Azure VMSS cluster called mycluster (where 'mycluster' is the name assigned to your cluster):

.bin/muchos launch -c `mycluster` # Launches Muchos cluster in Azure

Set up the cluster

Once your cluster is built in EC2 or Azure, the ./bin/muchos setup command will set up your cluster and start Hadoop, Zookeeper & Accumulo. It will download release tarballs of Fluo, Accumulo, Hadoop, etc. The versions of these tarballs are specified in muchos.props and can be changed if desired.

Optionally, Muchos can setup the cluster using an Accumulo or Fluo tarball that is placed in the conf/upload directory of Muchos. This option is only necessary if you want to use an unreleased version of Fluo or Accumulo. Before running the muchos setup command, you should confirm that the hash (typically SHA-512 or SHA-256) of your tarball matches what is set in conf/checksums. Run the command shasum -a 512 /path/to/tarball on your tarball to determine its hash. The entry in conf/checksums can optionally include the algorithm as a prefix. If the algorithm is not specified then Muchos will infer the algorithm based on the length of the hash. Currently Muchos supports using sha512 / sha384 / sha256 / sha224 / sha1 / md5 hashes for the checksum.

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:

  1. 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.

  2. 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.

  3. spark - If specified on a node, Apache Spark will be installed on all nodes and the Spark History server will be run on this node.

  4. 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/.

  5. client - Used to specify a client node where no services are run but libraries are installed to run Accumulo/Hadoop clients.

  6. swarmmanager - Sets up Docker swarm with the manager on this node and joins all worker nodes to this swarm. When this is set, docker will be installed on all nodes of the cluster. It is recommended that the swarm manager is specified on a worker node as it runs docker containers. Check out Portainer if you want to run a management UI for your swarm cluster.

  7. elkserver - Sets up the Elasticsearch, Logstash, and Kibana stack. This allows logging data to be search, analyzed, and visualized in real time.

If you run the muchos setup command and a failure occurs, you can repeat the command until setup completes. Any work that was successfully completed will not be repeated. While some setup steps can take over a minute, use ctrl-c to stop setup if it hangs for a long time. Just remember to run muchos setup again to finish setup.

Manage the cluster

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 ./bin/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 management 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.

Run a Fluo application

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 a node on cluster where Fluo is installed and clone the example repo:

./bin/muchos ssh                      # SSH to cluster proxy node
ssh <node where Fluo is installed>    # Nodes with Fluo installed is determined by Muchos config
hub clone apache/fluo-examples        # Clone repo of Fluo example applications. 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 fluo-examples/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 Fluo application instructions to configure, initialize, and start your application. To automate these steps, you can mimic the scripts of example Fluo applications above.

Customize your cluster

After ./bin/muchos setup is run, users can install additional software on the cluster using their own Ansible playbooks. In their own 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).

High-Availability (optional)

Additionally, Muchos can be configured to provide High-Availability for HDFS & Accumulo components. By default, this feature is off, however it can be turned on by editing the following settings in muchos.props under the general section as shown below:

hdfs_ha = True                        # default is False
nameservice_id = muchoshacluster      # Logical name for the cluster, no special characters

Before enabling HA, it is strongly recommended you read the Apache doc for HDFS HA & Accumulo HA

Also in the [nodes] section of muchos.props ensure the journalnode and zkfc service are configured to run.

When hdfs_ha is True it also enables the ability to have HA resource managers for YARN. To utilize this feature, specify resourcemanager for multiple leader nodes in the [nodes] section.

Terminating your cluster

If you launched your cluster, run the following command to terminate your cluster. WARNING - All data on your cluster will be lost:

./bin/muchos terminate

Automatic shutdown of clusters

With the default configuration, clusters will not shutdown automatically after a delay and the default shutdown behavior will be stopping the node. If you would like your cluster to terminate after 8 hours, set the following configuration in muchos.props:

shutdown_delay_minutes = 480
shutdown_behavior = terminate

If you decide later to cancel the shutdown, run muchos cancel_shutdown.

Retrieving cluster configuration

The config command allows you to retrieve cluster configuration for your own scripts:

$ ./bin/muchos config -p leader.public.ip
10.10.10.10

Contributions

We welcome contributions to the project. These notes should be helpful.

Powered by

Muchos is powered by the following projects: