Hi, I had faced the same problem.. In my case, the workers were not able to send heartbeats because of long pauses due to garbage collection.
I configured storm workers to use a concurrent mark sweep GC and now it works fine Hope this helps -Palak On Oct 23, 2014 5:20 PM, "Vladi Feigin" <[email protected]> wrote: > Hi All, > > We observe that storm (nimbus) often moves workers to other > supervisors/nodes. > It' happens when the storm cluster is in good shape. We suspect it's > related to the heartbeats delays from supervisors to nimbus. Can be ? How > can we check,prove/disprove it? > I think it's better to avoid this since it's expensive operation .. > > Thank you in advance, > Vladi >
