[
https://issues.apache.org/jira/browse/ACCUMULO-4335?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15320850#comment-15320850
]
Sean Busbey commented on ACCUMULO-4335:
---------------------------------------
There are several in examples and tests that I'm not worried about. There are a
couple in the server code. They look like they're mostly in early init code
(like when the GC is first checking for a lock, or when Master is doing an
upgrade) or fail-safe (like if we somehow get a bare exception in
TabletServer.main) and they all are non-zero exit.
The only issue I can see is this one in TabletServer
{code}
} catch (Exception ex) {
log.error("Uncaught exception in TabletServer.main, exiting", ex);
System.exit(1);
} finally {
DistributedTrace.disable();
}
{code}
That finally block will never happen if we hit the exception path. But HTrace
itself has a shutdown hook to ensure things are closed off, so probably fine?
Since these are essentially an unrelated problem I'd prefer to defer correcting
them, if you don't mind.
> Error conditions that result in a Halt call should ensure non-zero process
> exit code
> ------------------------------------------------------------------------------------
>
> Key: ACCUMULO-4335
> URL: https://issues.apache.org/jira/browse/ACCUMULO-4335
> Project: Accumulo
> Issue Type: Bug
> Components: gc, tserver
> Affects Versions: 1.6.5, 1.7.0
> Reporter: Sean Busbey
> Assignee: Sean Busbey
> Fix For: 1.8.0, 2.0.0
>
> Attachments: ACCUMULO-4335.1.patch
>
>
> There are several places where we get some fatal error and exit with a 0
> code, indicating no error.
> e.g. when the TabletServer loses the ability to monitor zk.
> (this impacts 1.6.0+, but I can only select 1.6.5 from 1.6.z)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
