[jira] [Created] (DRILL-3890) TPCH Concurrency test hit foremanException even when all the drillbits are alive

["Dechang Gu (JIRA)" <ji...@apache.org>]

Dechang Gu created DRILL-3890:
---------------------------------

             Summary: TPCH Concurrency test hit foremanException even when all the drillbits are alive
                 Key: DRILL-3890
                 URL: https://issues.apache.org/jira/browse/DRILL-3890
             Project: Apache Drill
          Issue Type: Bug
          Components: Functions - Drill
    Affects Versions: 1.2.0
         Environment: ucs-node 1 - node 11 (10+1 node cluster), RHEL 6.4 Linux 2.6.32-358.el6.x86_64, MapR 4.0.2.29870.GA, drill 1.2 master git commitID f78ab84183e73216b76732f66f87ccf48e2340d3
            Reporter: Dechang Gu


In TPCH Concurrency test, when number of query threads is 24 or more, many queries are terminated due to "FroemanException", for example:
SYSTEM ERROR: ForemanException: One more more nodes lost connectivity during query. Identified nodes were [ucs-node8.perf.lab:31010].
(org.apache.drill.exec.work.foreman.ForemanException) One more more nodes lost connectivity during query. Identified nodes were [ucs-node8.perf.lab:31010].
org.apache.drill.exec.work.foreman.QueryManager$2.drillbitUnregistered():527
org.apache.drill.exec.coord.ClusterCoordinator.drillbitUnregistered():68
org.apache.drill.exec.coord.zk.ZKClusterCoordinator.updateEndpoints():248
org.apache.drill.exec.coord.zk.ZKClusterCoordinator.access$300():58
org.apache.drill.exec.coord.zk.ZKClusterCoordinator$ZKListener.cacheChanged():154
org.apache.curator.x.discovery.details.ServiceCacheImpl$2.apply():164
org.apache.curator.x.discovery.details.ServiceCacheImpl$2.apply():160
org.apache.curator.framework.listen.ListenerContainer$1.run():92
com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute():293
org.apache.curator.framework.listen.ListenerContainer.forEach():83
org.apache.curator.x.discovery.details.ServiceCacheImpl.childEvent():157
org.apache.curator.framework.recipes.cache.PathChildrenCache$5.apply():509
org.apache.curator.framework.recipes.cache.PathChildrenCache$5.apply():503
org.apache.curator.framework.listen.ListenerContainer$1.run():92
com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute():293
org.apache.curator.framework.listen.ListenerContainer.forEach():83
org.apache.curator.framework.recipes.cache.PathChildrenCache.callListeners():500
org.apache.curator.framework.recipes.cache.EventOperation.invoke():35
org.apache.curator.framework.recipes.cache.PathChildrenCache$10.run():762
java.util.concurrent.Executors$RunnableAdapter.call():471
java.util.concurrent.FutureTask.run():262
java.util.concurrent.Executors$RunnableAdapter.call():471java.util.concurrent.FutureTask.run():262
java.util.concurrent.ThreadPoolExecutor.runWorker():1145
java.util.concurrent.ThreadPoolExecutor$Worker.run():615
java.lang.Thread.run():745


But at that time the drillbit on the identified node is still active. 
It works fine with 16 or fewer query threads.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)