[
https://issues.apache.org/jira/browse/HTTPCLIENT-1108?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jon Moore resolved HTTPCLIENT-1108.
-----------------------------------
Resolution: Invalid
Fix Version/s: (was: 4.2 Alpha1)
Upon further review, the implementation already does the right thing; I was
confused by the free connection management between ConnPoolPerRoute and
RouteSpecificPool. The RouteSpecificPool *does* manage the free connections in
LIFO order already (and there's even a comment that this is to promote unneeded
connections becoming idle); on the other hand, the ConnPoolPerRoute uses FIFO
when it "steals" a connection from one route-specific pool to give to another,
as that's the connection that is least likely to be missed/needed for that
route.
> different connection reuse strategy would reduce number of open connections
> ---------------------------------------------------------------------------
>
> Key: HTTPCLIENT-1108
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1108
> Project: HttpComponents HttpClient
> Issue Type: Improvement
> Components: HttpClient
> Affects Versions: 4.1.1
> Reporter: Jon Moore
> Assignee: Jon Moore
> Priority: Minor
>
> Currently, the ThreadSafeClientConnManager reuses persistent connections in a
> round-robin fashion: when the connection is checked back into the pool, it is
> added to a queue of available connections for later use. This has the perhaps
> unintended effect that the client may keep open more connections than needed,
> because the "last used" time keeps getting updated and none of the
> connections can get reclaimed via closeIdleConnections().
> Exchanging the queue (FIFO) for a stack (LIFO) would result in extra
> connections actually becoming idle for long enough to be reclaimed.
> I have a working patch for this; just need to get my act together to get it
> in.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
