Tools » teuthology

teuthology.front has to be rebooted pretty often these days due to cephfs issues. One of the negative side-effects is that when running jobs are killed by the reboot, they don't have a way to tell paddles they're dying. Cleaning them up manually is more time-consuming than it needs to be.

For now, implementing a pecan command that checks for running jobs with an updated timestamp that is past a certain age should be good enough.

It would be nice, though, to figure out a sane way to automatically execute that periodically.