HC loses connection to its server at long GC pauses

[tair.sabirgaliev]

I noticed that HC loses connection to its server and subsequently unregisters it while server survives a long GC pause (20-30 seconds). What is expected behavior in such cases?

Steps to reproduce:

0. install stock AS7

1. start domain.sh and wait until all servers are started

2. identify PID of any running server process

3. kill -SIGSTOP <PID>

4. wait for 20-30 seconds

5. observe "Unregistering server" in console log

6. kill -SIGCONT <PID>

7. verify the server is up and running but not available for management.

I used SIGSTOP to simulate a GC pause, though not sure if it is appropriate, had no other means of having an artificial GC pause long enough to reproduce the issue.

[jaikiran]

I think there's a configurable timeout for this. I'll check with someone who has more knowledge about this.

[tair.sabirgaliev]

Why there are no attempts to reconnect? Presumably HC knows that the server was not scheduled for shutdown and should be running. I observed this phenomenon also when my dev machine was put to sleep and woken up after some period -- I don't think this can be solved adjusting timeouts.

[ctomc]

Hi,

i am assuming that you are seeing this on 7.1.1.Final right?

There have been changes in this area in current master, so can you try to reproduce this with nightly build https://community.jboss.org/thread/167590

--

tomaz

[emuckenhuber]

Yeah, this should be fixed in upstream. The related jira issue is: https://issues.jboss.org/browse/AS7-5887

[tair.sabirgaliev]

Thanks, I should further develop my JIRA searching skills

[tair.sabirgaliev]

Thanks Tomaz! latest from jenkins handles it gracefully