刘方奇 created FLINK-24053:
---------------------------
Summary: stop with savepoint timeout
Key: FLINK-24053
URL: https://issues.apache.org/jira/browse/FLINK-24053
Project: Flink
Issue Type: Bug
Components: Runtime / Checkpointing, Runtime / REST
Affects Versions: 1.13.0, 1.12.0, 1.11.0
Reporter: 刘方奇
Hello, when we use the "stop with savepoint" feature, we always meet a bug.
We will always cost 5 mins waiting the application to end, then the application
will throw a timeout exception.
{code:java}
//代码占位符
java.util.concurrent.TimeoutException:
nulljava.util.concurrent.TimeoutException: null at
org.apache.flink.runtime.concurrent.FutureUtils$Timeout.run(FutureUtils.java:1036)
~[classes/:?] at
org.apache.flink.runtime.concurrent.DirectExecutorService.execute(DirectExecutorService.java:211)
~[classes/:?] at
org.apache.flink.runtime.concurrent.FutureUtils.lambda$orTimeout$14(FutureUtils.java:445)
~[classes/:?] at
java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
~[?:1.8.0_251] at java.util.concurrent.FutureTask.run(FutureTask.java:266)
~[?:1.8.0_251] at
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180)
~[?:1.8.0_251] at
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
~[?:1.8.0_251] at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
~[?:1.8.0_251] at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
~[?:1.8.0_251] at java.lang.Thread.run(Thread.java:748) ~[?:1.8.0_251]
{code}
And we found there was always the function called
org.apache.flink.runtime.rest.handler.job.savepoints.SavepointHandlers.SavepointStatusHandler.closeHandlerAsync()
run timeout, and its timeout setting is 5mins.
There was a question that the handler 's close may be not important, cause the
handler serves other handler called
org.apache.flink.runtime.rest.handler.job.savepoints.SavepointHandlers.StopWithSavepointHandler
which was already closed.So should we skip this close ?
PS : There was no problem when we test the code that skip the handler 's close.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
