[collectd] memory leak

[Richard Elling]

Hi Dennis,

> On Jan 28, 2015, at 7:00 AM, Dennis Jacobfeuerborn <dennisml at conversis.de> wrote:
> 
> Hi,
> it seems collectd is suffering from a memory leak as on some of my
> systems the process has ballooned to 800+MB.
> I found a reference to a write_graphite leak here:
> https://github.com/collectd/collectd/issues/359
> 
> Has this been resolved in a release? What is the recommended way to get
> around this issue?

We've got a change-to-be-tested for this issue that, well, still needs to be tested.
The basic idea is to change the write_graphite backlog to a circular buffer and
cap its size. If you'd like to volunteer to test, the changes are available in the
'circular-buffer' branch at:
https://github.com/Coraid/coraid-collectd/tree/circular-buffer <https://github.com/Coraid/coraid-collectd/tree/circular-buffer>

more specifically:
https://github.com/Coraid/coraid-collectd/commits/circular-buffer/src/daemon <https://github.com/Coraid/coraid-collectd/commits/circular-buffer/src/daemon>

Note: many of the agents we use collect derive-type information, so it is fine
to keep just the most recent. Other use cases might have different requirements,
but in any case it isn't worth running out of memory.
 -- richard

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.verplant.org/pipermail/collectd/attachments/20150129/7b80758e/attachment.html>