DoIT HTCondor Pool

Overview of HTCondor pools on DoIT resources.

1.0 HTCondor headend and puppet master is the headend of the DoIT HTCondor pool. It manages the resources made available for jobs and schedules them. This machine primarily takes linux jobs that overflow from CHTC. In addition, it pushes out linux VM images to windows hosts in the pool.

The linux machines in the pool are managed via puppet. condor0 is the puppet master for these nodes, making it easier to maintain the distributed infrastructure.

2.0 Linux Pool

2.1 Linux Pool dedicated worker nodes

  • (dynes project)
  • (dynes project)
  • (lab captivator)

2.2 Linux Pool virtual machine worker nodes

In order to prevent starvation to other vm's on the host, the VM worker nodes are typically provisioned with 1 or 2 less cpu cores than are usually available. In addition, the CPU shares are configured to low.
  • on
  • on moxi-server
  • on
  • on
  • on
  • tbd on
  • tbd on

2.3 Staff workstation VM's

Some staff in Network Services run linux virtual machines on their apple workstations. These resources can also be added into the DoIT HTCondor pool. Optionally, these workstations can be configured to only run jobs at specific times of the day.

See Joining the DoIT HTCondor Pool for specific instructions.

3.0 Windows Pool

Kiosks and Infolabs around campus managed by the DoIT RaDS group are running the windows version of HTCondor. These machines then run a linux vm image (via the HTCondor vm "universe" feature) as their job when the machines are idle. These linux VM's then call home to a submit node at CHTC to find their work. So, effectively, we are running HTCondor inside of HTCondor.

Keywords:htcondor condor pool condor0   Doc ID:39707
Owner:Michael H.Group:Network Services
Created:2014-05-01 10:50 CDTUpdated:2017-08-08 14:10 CDT
Sites:Network Services
Feedback:  0   0