Skip to content

Condor improvements

What?

This merge request introduces improvements to the interface between asimov and htcondor. The intention behind these is to substantially improve runtimes of asimov when checking job status while also reducing strain on the htcondor schedulers.

Why?

The approach we used to checking the status of condor jobs in O3 relied on querying the scheduller each time the status needed to be determined. This became problematic when the scheduller was busy, and completing the update of asimov's database from condor reguarly created race conditions.

How?

We now use a slightly less intensive approach to querying the scheduller, only asking for a small subset of the parameters of each job. In addition, the results are now cached for a short period of time.

Who?

This work was done by Daniel Williams (daniel.williams@ligo.org)

Checklist

  • New code
  • Tests
  • Documentation & examples
  • Release target
Edited by Daniel Williams

Merge request reports