storm-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ethan Li <>
Subject Re: Start command not propagated by Nimbus
Date Tue, 04 Jun 2019 20:56:31 GMT
Hi Mitchell,

Does the UI show that this topology is fully scheduled? Are you using ResourceAwareScheduler?
Because it’s possible that the scheduler cannot find enough resources to schedule this topology.

Also after starting the topology, you could login to zookeeper and check if there is a assignment
belong to this topology. You can also read the content of it to get some better idea. But
you will need some coding to do it since it’s serialized. 

It’s really hard to tell the root cause without more information. Is it possible for you
to provide all the related nimbus.log, supervisor.log so I can take a look?


> On May 17, 2019, at 11:38 AM, Mitchell Rathbun (BLOOMBERG/ 731 LEX) <>
> We have a topology that was never started, even though nimbus received the start command.
Supervisor never received a command to start this topology, so the issue wasn't in our topology
code. In the logs, I see:
> 2019-05-11 10:07:28,087 INFO  nimbus [pool-14-thread-16] Activating WingmanTopology4246:
> There were a bunch of topologies started around the same time, and most of them had the
following message occur next:
> [timer] Setting new assignment for topology id <Topology Name>:................
> However, we did not see this logged for the topology that wasn't started. When the cluster
was stopped, we saw:
> 2019-05-11 10:36:04,447 INFO  nimbus [pool-14-thread-4] Delaying event :remove for 5
secs for WingmanTopology4246-251-1557583643
> 2019-05-11 10:36:04,457 INFO  nimbus [pool-14-thread-4] Adding topo to history log: WingmanTopology4246-251-1557583643
> What could have caused this? There were 16 topologies submitted to be run in total, our
storm.yaml file allocates more than enough slots under supervisor.slots.ports.

View raw message