cbatch

README

cbatch - Containerized Torque

Run your batch jobs in containers, created dynamically at run time. Without changes to your jobs.

https://asciinema.org/a/e9gfcdnbrt5e6hb1jor5lp8x8

---

Build Requirements

• Go https://golang.org/
• Glide https://github.com/Masterminds/glide

Build

You can download and compile the project with:

$ go get github.com/cbatchx/cbatch
$ cd $GOPATH/src/github.com/cbatchx/cbatch
$ glide install
$ go build

After building you can run it on a torque cluster provided by vagrant.

Install

Move config/config.toml to /etc/cbatch.toml and edit to fit your system.

$ sudo cp config.toml /etc/cbatch.toml

Move cbatch to some bin folder.

$ sudo cp cbatch /usr/bin/cbatch

Add $jobstarter option to your Torque mom config.

$ echo '$job_starter /usr/bin/cbatch' >> /var/spool/torque/mom_priv/config
$ echo '$job_starter_run_privileged true' >> /var/spool/torque/mom_priv/config

Development Requirements without Torque

• Docker installed locally

In this mode you simply give cbatch some environment variables to "mock" a submitted torque job.

Run without Torque

To run cbatch without Torque some environment variables must be "mocked".

• PBS_O_LOGNAME must be set to a user on the system.
• PBS_O_HOME must be set to the home directory of the user.
• PBS_JOBID must be set to the name of one of the jobs in test/jobs without .SC

The config.toml file must also be configured to make it run. See the config/config.toml

You must also supply which shell to run the job as and the job as parameters to cbatch as Torque normally supplies cbatch with these.

$ go build
$ PBS_O_LOGNAME=$USER PBS_O_HOME=$HOME PBS_JOBID=hello ./cbatch /bin/bash `pwd`/test/jobs/hello.SC

You can also make the job interactive by running (you will end up with a shell in the container): $ PBS_O_LOGNAME=$USER PBS_O_HOME=$HOME PBS_JOBID=hello PBS_ENVIRONMENT=PBS_INTERACTIVE ./cbatch /bin/bash

Development Requirements with Torque

• Vagrant https://www.vagrantup.com/

Run with Torque

The following command builds a three part virtual Torque cluster with 1 master host and 2 slaves.

$ NODES=2 vagrant up

NODES defines the number of slave nodes that will be created for the cluster.

Access

$ vagrant ssh master
$ echo 'echo "Hello Docker"' | qsub

You can also ssh into slave1, slave2 etc. to look at the logs.

Suspend

$ NODES=2 vagrant suspend

Take Down

$ NODES=2 vagrant destroy

Vagrant setup is built upon crcollins/torquecluster