#2983
  • Created on
    2013-01-29 17:41
  • Reported by
    Patrik Nordwall
  • Status
    Fixed
  • Priority
    Normal (3)
  • Assigned to
    Patrik Nordwall
  • Milestone
    2.2-M1
  • Plan Level
    -
  • Due Date
    -

Attachments

Regression in watch?

StressSpec is now failing in "end routers that are running while nodes are removed" and that is not same problem that initially was reported in #2952, which Björn is investigating.
https://jenkins.akka.io:8498/job/akka-local/189

I can see from the logs that the WorkResult is received and "jobs/s, retried" is logged, i.e. the actor is stopped, but then the watch followed by expectMsg Terminated times out.

I'm thinking if this can be something new introduced by the remote deadletter fix earlier today.
User picture
  on 2013-01-29 18:03 *
By drewhk
I cannot exclude that possibility, but the fix only unwraps remote message send envelopes when they arrive to deadletters.
User picture
  on 2013-01-30 06:42 *
By Rich Dougherty
I can replicate the failure locally. It appears to start happening on commit e0a9dd7, Dead letters containing remote envelopes handled correctly #2959.
User picture
  on 2013-01-30 07:14 *
By Patrik Nordwall
Me too, but only when running the full StressSpec. I have tried to minimize without luck.

Note that it is a local actor that is watched via testActor.

It's a concurrent watch and stop scenario. A related ticket to that is #2915. How sure are we that it was really fixed?
User picture
  on 2013-01-30 08:36 *
By Patrik Nordwall
Assigned to set to Patrik Nordwall
Status changed from New to Accepted
User picture
  on 2013-01-31 10:36 *
By Patrik Nordwall
Status changed from Accepted to Test
User picture
  on 2013-01-31 10:38 *
By Patrik Nordwall
The full story is here: https://github.com/akka/akka/pull/1094
and follow up ticket #2993 is created.
User picture
  on 2013-01-31 12:38 *
By Patrik Nordwall
Status changed from Test to Fixed
User picture
  on 2013-02-15 13:03 *
By Patrik Nordwall
Milestone changed from Current to 2.2-M1