[
https://issues.apache.org/jira/browse/HTTPCORE-429?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Oleg Kalnichevski resolved HTTPCORE-429.
----------------------------------------
Resolution: Fixed
Fix Version/s: 5.0-alpha2
4.4.6
This one turned out to be not so easy. There were two options how to fix the
problem for individual routes: by making {{#getStats}} method iterate through
the entire global request queue thus making it expensive under load or by
maintaining an extra structure per each route to keep track of route specific
requests and making every lease / release operation a bit more expensive but
keeping {{#getStats}} method cheap. I opted for the former.
Fixed in SVN trunk and 4.4.x branch. Please review / re-test.
Oleg
> pool statistics don't expose all 'pending' requests
> ---------------------------------------------------
>
> Key: HTTPCORE-429
> URL: https://issues.apache.org/jira/browse/HTTPCORE-429
> Project: HttpComponents HttpCore
> Issue Type: Bug
> Components: HttpCore NIO
> Affects Versions: 4.4.5
> Reporter: gui
> Fix For: 4.4.6, 5.0-alpha2
>
>
> During a performance run i noticed that the async http client was causing OOM
> exceptions.
> After enabling pool statistics, i did not see a big 'queue' in the pool (the
> 'org.apache.http.pool.PoolStats#getPending' method always returns a small
> number)
> However, after some memory profiling it noticed that the pool has a lot of
> pending leaseRequests, but those requests are not exposed in the poolstats.
> I think the org.apache.http.pool.PoolStats#getPending should also include the
> leaseRequest count.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
