serialization of jobs?
|Assignee:||Alex Norton||% Done:||
Not even sure this is a defect. While running a number of jobs on a system (snape and 4privet.ostt) using recent software (svn version 5443) a large job was running (a debian iso) and nomos was processing entries.... I deleted an upload. When I checked the jobs, the delete job was scheduled, but never seemed to run. It sat there for over 1/2 hour. When I checked the systems this morning, everything had run to completion.
It appears that if all the jobs are at the same priority, and something like nomos is running, none of the other jobs get run.... is this how the jobs are supposed to run, or is there a problem here? I expected the delete job to get some cycles even if other jobs are running, or are they run in order now?
Updated by Alex Norton over 1 year ago
The secret here is that it was a delete job. The del agent is special and if you look in its configuration file you'll find that it uses the exclusive flag. This flag indicates that no other agent can be run at the same time as the del agent. So del agents are serialized with everything else. Nomos is not, I'm sure that if you start a large Nomos job, and a Copyright job, they will both get run at the same time.