Andrey created FLINK-6278:
-----------------------------
Summary: After "FATAL" message in logs taskmanager still working
Key: FLINK-6278
URL: https://issues.apache.org/jira/browse/FLINK-6278
Project: Flink
Issue Type: Bug
Affects Versions: 1.2.0
Reporter: Andrey
Steps to reproduce:
* create job which cause OOM. For example for each incoming message create byte
array and store in memory:
{code}
}).flatMap(new RichFlatMapFunction<String, AggregatedHash>() {
private List<byte[]> oomList = new ArrayList<>();
@Override
public void flatMap(String value, Collector<AggregatedHash> out)
throws Exception {
if (oomHost != null && oomHost.equals(host)) {
//multiply speed towards oom
oomList.add(new byte[100 * 1024]);
}
}
}
{code}
* after some time task manager hangs
* according to logs task manager will be disconnected from zookeeper (ha mode
was configured) and job manager
* then big log entry:
{code}
2017-04-07 09:13:32,893 ERROR org.apache.flink.runtime.taskmanager.TaskManager
-
==============================================================
====================== FATAL =======================
==============================================================
A fatal error occurred, forcing the TaskManager to shut down: Task 'Flat Map ->
Sink: Unnamed (2/2)' did not react to cancelling signal in the last 30 seconds,
but is stuck in method:
org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:182)
org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:63)
org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:272)
org.apache.flink.runtime.taskmanager.Task.run(Task.java:655)
java.lang.Thread.run(Thread.java:745)
{code}
* TM still running.
Thread dump attached.
Expected:
* shutdown task manager
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
