On Mon, Jul 3, 2017 at 12:02 PM, Stefan Richter <s.rich...@data-artisans.com> wrote: > Another thing that could be really helpful, if possible, can you attach a > profiler/sampling to your job manager and figure out the hotspot methods > where most time is spend? This would be very helpful as a starting point > where the problem is potentially caused.
A stack trace will also be helpful to see whether some threads are stuck. If it is possible to run on Mesos without HA mode (@Till: is that possible?), it might be worthwhile to re-run this without HA to get a hint whether it is related to HA mode. – Ufuk
