[
https://issues.apache.org/jira/browse/DISPATCH-1502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16992715#comment-16992715
]
Gordon Sim commented on DISPATCH-1502:
--------------------------------------
The improved handling of MAUs helps a bit with this (I think because when the
general work is not backed up then things get repooled fatser so less new
objects are needed).
> large scale link disconnects can push memory up
> -----------------------------------------------
>
> Key: DISPATCH-1502
> URL: https://issues.apache.org/jira/browse/DISPATCH-1502
> Project: Qpid Dispatch
> Issue Type: Improvement
> Reporter: Gordon Sim
> Assignee: Ted Ross
> Priority: Minor
> Fix For: 1.11.0
>
>
> If a router has some connections with large numbers of links on them, and
> these connections get suddenly disconnected, the memory for the router can
> grow quite a bit in processing the link disconnects (>30% in some cases).
> The cause is that there are a large number of qdr_action_t and
> qdr_general_work_t instances used as the work is passed from worker thread to
> core and back to a worker thread again. As this happens very rapidly
> instances are not returned to the pool before more are needed so the pool
> grows (in my experiments it was growing by close to the number of links
> involved).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
