[
https://issues.apache.org/jira/browse/TS-1580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13503570#comment-13503570
]
Alan M. Carroll commented on TS-1580:
-------------------------------------
Having looked at this myself and reconsidering the data above I no longer think
that netVCs are being leaked. The allocation numbers are total bytes, not
elements, so even 629K of netVC is only 915 netVCs "in use" which is not a
large number. Given that each event thread can accumulate 512 netVCs in its
thread local cache (which count as "in use" to this dump) that's not even
noticeable.
On the other hand, almost 7G of memory allocated to mutexes seems quite
unreasonable. That's far far more than could be held by a few thousand netVCs.
> mutexAllocator and netVCAllocator in-use count continually growing
> ------------------------------------------------------------------
>
> Key: TS-1580
> URL: https://issues.apache.org/jira/browse/TS-1580
> Project: Traffic Server
> Issue Type: Bug
> Components: Core
> Affects Versions: 3.2.0
> Reporter: Shaun McGinnity
> Fix For: 3.3.3
>
>
> Running HTTP traffic through 3.2.0 we see the in-use count from the
> mutexAllocator and netVCAllocators continually growing.
> An example allocator dump before a test:
> 88064 | 2064 | 688 | memory/netVCAllocator
> 1566720 | 1564800 | 80 | memory/mutexAllocator
> After the test has completed (about 15 minutes of traffic):
> 13914112 | 629520 | 688 | memory/netVCAllocator
> 2570240 | 2376480 | 80 | memory/mutexAllocator
> We never see the in-use counters fall back to the value set before the test.
> If we start the test again then we get further growth. Running for longer or
> under heavy load the growth is much clearer:
> 6791976960 | 6791970080 | 80 | memory/mutexAllocator
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
