timeout await for agent: driver-conductor. Retrying...

[patriknw]

After updating Akka from 1.2.3 to 1.2.4 one of our tests went haywire. Endless loop of "timeout await for agent: driver-conductor. Retrying...".

It's not reproducible every time but failed several times in our nightly tests.

Here is the test and the problem is in the shutdown after the actual successful tests: https://github.com/akka/akka/blob/master/akka-remote-tests/src/multi-jvm/scala/akka/remote/artery/AeronStreamConcistencySpec.scala#L89-L91

[mjpt777]

I've tried to recreate this but with no luck. Seems like there might be a race condition in the start stop logic. 1.2.4 is more efficient for startup and shutdown. I'm made a change to Agrona that should avoid any potential race conditions in this area.

[patriknw]

stacks.zip

Here are some thread dumps, 2 different processes, 3 dumps for each.

[mjpt777]

This is very helpful. There is a race on shutdown which I have addressed.

[patriknw]

great, thanks