On Fri, Jan 27, 2012 at 10:46 AM, Robert Muir <[email protected]> wrote:
> On Fri, Jan 27, 2012 at 10:22 AM, Yonik Seeley
> <[email protected]> wrote:
>>
>> FWIW, these aren't real failures of these particular tests, but
>> another test bleeding over into these
>> caused by a still-running thread later closing a searcher (and it
>> get's counted during whatever test is running at the time).
>>
>
> Hmm, we had this confusion before and thats why we track resource
> leaks from tests that still leave threads running
> after all afterClass() methods have finished. In lucene we actually
> check this on a per-method basis (after tearDown):
> because its annoying for a thread to confusingly cause a different
> method even in the same test class to fail.
>
> That's why its important for tests to not leave a lot of threads
> running (there are a lot of warnings in solr tests about this):

Yep, I've been planning on moving the facet executor to the Core or
CoreContainer so it can be shut down.
We'll still have some that aren't in our control though - I've started
shutting down the  hhtp connection manager used to talk to other solr
instances in distrib, but now what I see is
"MultiThreadedHttpConnectionManager cleanup".

Just like Solr ignores some expected exceptions, we may want a way to
downgrade the severity of certain expexted threads still running at
the end of tests.

-Yonik
http://www.lucidimagination.com

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to