vanilla_improvements notes are at the bottom.
|StarCluster:||Cluster Computing Toolkit for the Cloud|
|Author (trunk):||Justin Riley <firstname.lastname@example.org>|
|Author (vanilla improvements):||François-Michel L'Heureux <email@example.com>|
|Team:||Software Tools for Academics and Researchers (http://star.mit.edu)|
StarCluster is a utility for creating and managing computing clusters hosted on Amazon's Elastic Compute Cloud (EC2). StarCluster utilizes Amazon's EC2 web service to create and destroy clusters of Linux virtual machines on demand.
All that's needed to create your own cluster(s) on Amazon EC2 is an AWS account and StarCluster. StarCluster features:
- Simple configuration - with examples ready to go out-of-the-box
- Create/Manage Clusters - simple start command to automatically launch and configure one or more clusters on EC2
- Automated Cluster Setup - includes NFS-sharing, Open Grid Scheduler queuing system, Condor, password-less ssh between machines, and more
- Scientific Computing AMI - comes with Ubuntu 11.10-based EBS-backed AMI that contains Hadoop, OpenMPI, ATLAS, LAPACK, NumPy, SciPy, IPython, and other useful libraries
- EBS Volume Sharing - easily NFS-share Amazon Elastic Block Storage (EBS) volumes across a cluster for persistent storage
- EBS-Backed Clusters - start and stop EBS-backed clusters on EC2
- Cluster Compute Instances - support for "cluster compute" instance types
- Expand/Shrink Clusters - scale a cluster by adding or removing nodes
- Elastic Load Balancing - automatically shrink or expand a cluster based on Open Grid Scheduler queue statistics
- Plugin Support - allows users to run additional setup routines on the cluster after StarCluster's defaults. Comes with plugins for IPython parallel+notebook, Condor, Hadoop, MPICH2, MySQL cluster, installing Ubuntu packages, and more.
Interested? See the getting started section for more details.
Install StarCluster using easy_install:
$ easy_install StarCluster
or using pip:
$ pip install StarCluster
$ (Download StarCluster from http://star.mit.edu/cluster/downloads.html) $ tar xvzf starcluster-X.X.X.tar.gz (where x.x.x is a version number) $ cd starcluster-X.X.X $ sudo python setup.py install
After the software has been installed, the next step is to setup the configuration file:
$ starcluster help StarCluster - (http://star.mit.edu/cluster) Software Tools for Academics and Researchers (STAR) Please submit bug reports to firstname.lastname@example.org !!! ERROR - config file /home/user/.starcluster/config does not exist Options: --------  Show the StarCluster config template  Write config template to /home/user/.starcluster/config [q] Quit Please enter your selection:
Select the second option by typing 2 and pressing enter. This will give you a template to use to create a configuration file containing your AWS credentials, cluster settings, etc. The next step is to customize this file using your favorite text-editor:
$ vi ~/.starcluster/config
This file is commented with example "cluster templates". A cluster template defines a set of configuration settings used to start a new cluster. The example config provides a smallcluster template that is ready to go out-of-the-box. However, first, you must fill in your AWS credentials and keypair info:
[aws info] aws_access_key_id = #your aws access key id here aws_secret_access_key = #your secret aws access key here aws_user_id = #your 12-digit aws user id here
The next step is to fill in your keypair information. If you don't already have a keypair you can create one from StarCluster using:
$ starcluster createkey mykey -o ~/.ssh/mykey.rsa
This will create a keypair called mykey on Amazon EC2 and save the private key to ~/.ssh/mykey.rsa. Once you have a key the next step is to fill-in your keypair info in the StarCluster config file:
[key key-name-here] key_location = /path/to/your/keypair.rsa
For example, the section for the keypair created above using the createkey command would look like:
[key mykey] key_location = ~/.ssh/mykey.rsa
After defining your keypair in the config, the next step is to update the default cluster template smallcluster with the name of your keypair on EC2:
[cluster smallcluster] keyname = key-name-here
For example, the smallcluster template would be updated to look like:
[cluster smallcluster] keyname = mykey
Now that the config file has been set up we're ready to start using StarCluster. Next we start a cluster named "mycluster" using the default cluster template smallcluster in the example config:
$ starcluster start mycluster
The default_template setting in the [global] section of the config specifies the default cluster template and is automatically set to smallcluster in the example config.
After the start command completes you should now have a working cluster. You can login to the master node as root by running:
$ starcluster sshmaster mycluster
You can also copy files to/from the cluster using the put and get commands. To copy a file or entire directory from your local computer to the cluster:
$ starcluster put mycluster /path/to/local/file/or/dir /remote/path/
To copy a file or an entire directory from the cluster to your local computer:
$ starcluster get mycluster /path/to/remote/file/or/dir /local/path/
Once you've finished using the cluster and wish to stop paying for it:
$ starcluster terminate mycluster
Have a look at the rest of StarCluster's available commands:
$ starcluster --help
- Amazon AWS Account
- Python 2.6+
- Boto 2.23.0+
- Paramiko 1.12.1+
- WorkerPool 0.9.2
- Jinja2 2.7
- decorator 3.4.0+
- iptools 0.6.1+
- optcomplete 1.2-devel+
- PyCrypto 2.5+
- scp 0.7.1+
- iso8601 0.1.8+
Watch an ~8 minute screencast @ http://star.mit.edu/cluster
To learn more have a look at the documentation: http://star.mit.edu/cluster/docs/latest
StarCluster has a mailing list for users and developers:
Join our IRC channel #starcluster on freenode. If you do not have an IRC client you can join the #starcluster channel using your web browser:
StarCluster is licensed under the LGPLv3 See COPYING.LESSER (LGPL) and COPYING (GPL) for LICENSE details
vanilla_improvements branch notes
This branch intends to be a mirror of https://github.com/jtriley/StarCluster develop with more features. Note that all the listed commits are only there for references and may contain issues that have been fixed in subsequent commits.
- Added commands
printconfig - To print your existing cluster configuration
- Will clean Open Grid Engine from dead nodes. (Eg.: Dead spot instances)
- Manages "impaired" nodes. (Reboots reserved instances, kills spot instances.)
(Useful with spot instances and used by the vanilla_improvements load balancer)
- If sge_qmaster crashed, restarts it.
- If a newly created instance failed to initialize (it's booted but not properly configured in OGS) the instance will be added back to the cluster.
- Fixes nodes with same alias. (Commit c58253)
- Improved load balancer
- More stable with spot instances with automatic cleaning, required when a spot instance dies. Note that stuck jobs resulting in a dead instance are killed by the clean command. You will need to relaunch your job.
- loadbalance new flags
- --ignore-grp Instances won't have the placement group constraint. When using spot instances, it makes it easier to get instances at a lower price.
- --reboot-interval - Delay in minutes beyond which a node is rebooted if it's still being unreachable via SSH. Defaults to 10.
- --num_reboot_restart - Number of reboots after which a node is restarted (stop/start). Helpful in case the issue comes from the hardware. If the node is a spot instance, it will be terminated instead since it cannot be stopped. Defaults to false.
Improved node addition
- Streaming the process by adding nodes as soon as they are ready instead of waiting for all of them. (Pull Request 434)
- Removed some remote read/writes (very slow) and replaced them get/edit/push.
- Cancels spot instances requests going to state "bid-too-low" or "capacity-oversubscribed", which avoids StarCluster to look frozen while waiting endlessly for them to become active. (Commit f4c4d0)
- Ability to configure various node instance types with a "selection factor". The call to add node will then pick the type to use in function of the spot market prices and the "selection factor". See the wiki page Various Node Types Support for instructions. (Pull Request 18)
Support for multiple subnets - Via the cluster template, allows to get spot instances in the cheapest zone. Dropped the --subnet-id start command flag. (Commit 0824e3)
Adds a mode where the cluster configuration is written to master:/etc/starcluster. To activate, simply add flag "--config-on-master" to the start command. Clusters in this mode have the following pros and cons. (Commit 4bc193)
- Allows to easily update the config by editing the file.
- No more obscure update config compressed/hashed data in metadata/tags and other "obscure" places.
- No longer possible to start a stopped cluster via StarCluster. (This is technically fixable, but not planned at the moment.)
Adds a --dns-suffix flag to the start command. (Commit 72f3bc)
The runplugin command supports additional arguments. Useful to create StarCluster related tools. (Commit c3e097)
Reuse node ids when adding nodes to avoid growing over 999 and crashing. (Commit a86c5a)
Reject unrecognized config parameters. (Commit 0fd2cb)
Running the tests
Once StarCluster is installed, run:
python setup.py test --coverage