This issue seem to match what I am observing. I ended up changing yarn.nodemanager.vmem-check-enabled to false. Is this a bug? is there a side effect?
On Fri, Feb 24, 2017 at 2:54 PM, Ted Yu <yuzhih...@gmail.com> wrote: > Can you give more detail ? > > I found: > https://issues.apache.org/jira/browse/YARN-4714 > > Cheers > > On Fri, Feb 24, 2017 at 2:41 PM, Mohit Anchlia <mohitanch...@gmail.com> > wrote: > >> Figured out. It looks like there is a virtual memory limit check enforced >> in yarn which just surfaced with java 8 >> >> On Fri, Feb 24, 2017 at 2:09 PM, Mohit Anchlia <mohitanch...@gmail.com> >> wrote: >> >>> I recently upgraded the cluster from java 7 to java 8. Now when I run >>> flink on a yarn cluster I see errors: Eventually application gives up and >>> terminates. Any suggestions? >>> >>> Association with remote system [akka.tcp://flink@slave:35543] has >>> failed, address is now gated for [5000] ms. Reason: [Disassociated] >>> >> >> >
