[
https://issues.apache.org/jira/browse/HBASE-18880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16181322#comment-16181322
]
stack commented on HBASE-18880:
-------------------------------
I was going to mark this as minor since the .out has the exception that caused
the server crash -- and .out is where the 'unexpected' emissions go -- but I
notice that you say that the server stays up. Are there threads that should be
marked daemon threads that currently are not? The patch works? +1 if it does;
i.e. the server is shudown if unexpected exception out of start (or join).
> Failed to start rest server if the value of hbase.rest.threads.max is too
> small.
> --------------------------------------------------------------------------------
>
> Key: HBASE-18880
> URL: https://issues.apache.org/jira/browse/HBASE-18880
> Project: HBase
> Issue Type: Bug
> Components: REST
> Affects Versions: 3.0.0, 2.0.0-alpha-4
> Reporter: Guangxu Cheng
> Assignee: Guangxu Cheng
> Priority: Critical
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-18880.master.v0.patch
>
>
> After HBASE-18224, Jetty has be updated to 9.4.6, and it requires more
> threads to start up.
> If the value of hbase.rest.threads.max is too small, the rest server will
> fail to start.
> What I observed was as follows:
> 1. The process did not exit. (At the beginning, I thought the rest server has
> been start normally because of the process exists.)
> 2. Can't connect to the rest server and I didn't found any exception log in
> ***.log.
> 3. the main thread has exited (jstack log).
> 4. Found the exception information from ***.out.
> {code}
> java.lang.IllegalStateException: Insufficient threads: max=5 <
> needed(acceptors=1 + selectors=8 + request=1)
> at org.eclipse.jetty.server.Server.doStart(Server.java:414)
> at
> org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68)
> at org.apache.hadoop.hbase.rest.RESTServer.main(RESTServer.java:360)
> {code}
> I think the process should exit and log the information in ***.log when it
> happens.
> So that the user can directly discover that the rest server is abnormal.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
