Query blocked/Running for a long time

[Somasundara, Ashwin]

Hi guys –

   We are facing some issues while running queries. If we run 10 sessions (or 10 queries), 9 of them seem to run fine, however, one of them gets stuck for about 4-5minutes. It does complete eventually. When we looked at the thread dump, we see the threads get stuck at this point. This happens say 3 out of 5 times. Any ideas why this might be? Anybody has run into this kind of issues earlier?

 

"http-worker-312-acceptor-1@7459a21e-http@35c9a231{HTTP/1.1,[http/1.1, h2c, h2c-17, h2c-16, h2c-15, h2c-14]}{0.0.0.0:8081}" #312 prio=3 os_prio=0 tid=0x00007f5c88287000 nid=0x17a13 waiting for monitor entry [0x00007f5b1dca1000]
     java.lang.Thread.State: BLOCKED (on object monitor)
     at sun.nio.ch.ServerSocketChannelImpl.accept(ServerSocketChannelImpl.java:234)
     - waiting to lock <0x00007f6a2f02faf0> (a java.lang.Object)
     at org.eclipse.jetty.server.ServerConnector.accept(ServerConnector.java:373)
     at org.eclipse.jetty.server.AbstractConnector$Acceptor.run(AbstractConnector.java:593)
     at org.eclipse.jetty.util.thread.QueuedThreadPool.runJob(QueuedThreadPool.java:671)
     at org.eclipse.jetty.util.thread.QueuedThreadPool$2.run(QueuedThreadPool.java:589)
     at java.lang.Thread.run(Thread.java:748)

 

 

Thanks,

Ashwin

[Somasundara, Ashwin]

Edit: We are running 0.184 Version of presto. Data is on S3 and in ORC format. Machine is r4.4xlarge and 4 nodes and 1 co-ordinator.