Hi Arvid, I suspect that the JobManager considered the TaskManager as broken (because somehow it fails to report heartbeats back, when a job has failed). I recently pushed code that logs if a Jobmanager unregisters a Taskmanager. This is going to help us debugging those problems.
Robert On Wed, Aug 13, 2014 at 3:22 PM, Arvid Heise <[email protected]> wrote: > Hi Flinker, > > I'm receiving strange messages with higher DOPs. During scheduling the job > manager forgets about registered task managers. They are still up however. > > I attached the logs from jm and tm. > > Best, > > Arvid >
