I noticed just now that this job was queued in Jenkins for over 10 minutes before it eventually started. This seems like a bug somewhere in Jenkins, Gearman or Zuul.
The build was triggered in Jenkins, but Jenkins did not start it because it says it is waiting for an executor on integration-slave-precise-1013. But why did it not simply use another slave instead? There was plenty of room.
The mwext-*-testextension jobs are throttled by a plugin so there is only one instance of them per node. That is to prevent the slaves disk space to be filled with multiple copy of mediawiki/core
There might be a bug in that plugin :(
jjb/mediawiki-extensions.yaml:182: - throttle-one-per-node
That is the Jenkins extension.
This is intentionally configured this way via throttle-on-per-node to avoid multiple workspace on one slave for the same extension (both of which would clone mediawiki-core).
While we'll still have copies of mediawiki-core for each extension on each slave, this helps a little bit.
Blocked by: