Let's stop having our alerting be people complaining of test failures.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
Or, as @thcipriani just suggested, let's make a cronjob that deletes the workspaces before it becomes an issue, as that's what we do when it is an issue.
Comment Actions
Looking at the RelEng SAL it seems like all we've been doing is`rm -rf /srv/jenkins-workspace/workspace/*`.
(( $(df --output=pcent /srv | awk -F '%' '!/Use/ {print $1}') > 95 )) && \
rm -rf /srv/jenkins-workspace/workspace/*We could cron that, or have a workspace cleanup job, or the jobs could clean-up after themselves (which might be the Right Thing). Adding @hashar to see if he has any preferences among the options here.
Comment Actions
Done as part of T201224 . We now automatically depool the faulty slaves and have an IRC notification.