Viraj Jasani created PHOENIX-8041:
-------------------------------------
Summary: Client thread pools stop working after ThreadGroup
destruction
Key: PHOENIX-8041
URL: https://issues.apache.org/jira/browse/PHOENIX-8041
Project: Phoenix
Issue Type: Bug
Reporter: Viraj Jasani
PhoenixDriver.INSTANCE has one QueryServicesImpl per JVM, and all connections
made through this driver share its thread pool. The pool's thread factory comes
from Executors.defaultThreadFactory(), which puts every new thread in the
ThreadGroup of the thread that created the factory. The pool allows core thread
timeout, so when it is idle, all its threads exit. If that ThreadGroup is
destroyed after this, the pool can no longer create threads, and every task
submitted to it fails with IllegalThreadStateException.
The pools in PhoenixHAExecutorServiceProvider have similar problem.
While this is real concern, it is surfaced by DropColumnIT intermittent
failures.
{code:java}
org.apache.phoenix.exception.PhoenixIOException
at
org.apache.phoenix.iterate.BaseResultIterators.getIterators(BaseResultIterators.java:1641)
...
Caused by: java.lang.IllegalThreadStateException
at java.lang.ThreadGroup.addUnstarted(ThreadGroup.java:865)
at
java.util.concurrent.Executors$DefaultThreadFactory.newThread(Executors.java:629)
at
org.apache.phoenix.job.JobManager$ContextClassLoaderThreadFactory.newThread(JobManager.java:230)
at
java.util.concurrent.ThreadPoolExecutor.addWorker(ThreadPoolExecutor.java:912)
at
org.apache.phoenix.iterate.ParallelIterators.submitWork(ParallelIterators.java:130)
{code}
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
