[syslog-ng] Do stats cause memory leak

[Evan Rempel]

Sorry that I mistyped the library function. It should be g_atomic_pointer_add (not g_atomic_pointer_and).


On 06/05/2017 10:33 AM, Scheidler, Balázs wrote:
> I'll check if that dependency can be removed.
>
> On Jun 4, 2017 9:04 AM, "Czanik, Péter" <peter.czanik at balabit.com <mailto:peter.czanik at balabit.com>> wrote:
>
>     SLES 11 is also affected: https://github.com/balabit/syslog-ng/issues/1403#issuecomment-290195851 <https://github.com/balabit/syslog-ng/issues/1403#issuecomment-290195851>
>
>     Both SLES 11 and RHEL 6 are ancient but still used at many places. Actually about a year ago both still had more installations than SLES 12 or RHEL 7.
>
>     Bye,
>
>     Peter Czanik (CzP) <peter.czanik at balabit.com <mailto:peter.czanik at balabit.com>>
>     Balabit / syslog-ng upstream
>     https://www.balabit.com/blog/author/peterczanik/ <https://www.balabit.com/blog/author/peterczanik/>
>     https://twitter.com/PCzanik
>
>     On Sun, Jun 4, 2017 at 1:07 AM, Evan Rempel <erempel at uvic.ca <mailto:erempel at uvic.ca>> wrote:
>
>         So it looks like the latest git head no longer compiles on Redhat 6. It is using g_atomic_pointer_and which requires glib2 2.30+ and Redhat 6 only has glib2 2.28
>
>         Looks like I will not be able to test this :-( nor get this issue fixed in our infrastructure.
>
>         Evan.
>
>
>         On 06/03/2017 02:08 PM, Scheidler, Balázs wrote:
>>         My guess that it is related to #1517 instead, albeit stats counters were worked on recently. If #1517 does not fix this, please report it again.
>>
>>
>>         -- 
>>         Bazsi
>>
>>         On Wed, May 31, 2017 at 9:02 AM, Fekete, Róbert <robert.fekete at balabit.com <mailto:robert.fekete at balabit.com>> wrote:
>>
>>             There is also this issue that might be related (don't know if it is solved by the above PR or not): https://github.com/balabit/syslog-ng/issues/1494 <https://github.com/balabit/syslog-ng/issues/1494>
>>
>>             On Wed, May 31, 2017 at 8:28 AM, Czanik, Péter <peter.czanik at balabit.com <mailto:peter.czanik at balabit.com>> wrote:
>>
>>                 Hi,
>>
>>                 https://github.com/balabit/syslog-ng/pull/1505 <https://github.com/balabit/syslog-ng/pull/1505> seems to address memory problems. If you use my git head packages, I can add this patch for you and see if it resolves your problems.
>>
>>                 Bye,
>>
>>                 Peter Czanik (CzP) <peter.czanik at balabit.com <mailto:peter.czanik at balabit.com>>
>>                 Balabit / syslog-ng upstream
>>                 https://www.balabit.com/blog/author/peterczanik/ <https://www.balabit.com/blog/author/peterczanik/>
>>                 https://twitter.com/PCzanik
>>
>>                 On Wed, May 31, 2017 at 4:28 AM, Evan Rempel <erempel at uvic.ca <mailto:erempel at uvic.ca>> wrote:
>>
>>                     We are running git head download post 3.9.1 and the memory footprint seems to continually grow.
>>
>>                     Looking at the statistics for queued messages I found none, but discovered that we have nearly 22,000
>>                     statistic sets for a total of 68195 tracked statistics.
>>                     % /usr/local/sbin/syslog-ng-ctl stats | wc -l
>>                     68195
>>
>>                     It it possible that all of these statistics take up a non-trivial amount of RAM?
>>
>>                     USER       PID %CPU %MEM VSZ   RSS TTY     STAT START   TIME COMMAND
>>                     root      1489 79.2 61.5 12134140 <tel:%281%29%20213%204140> 10139760 <tel:%281%29%20013%209760> ?   Ssl  Apr24 41581:22 /usr/local/sbin/syslog-ng
>>
>>                     That is a 12GB footprint after 1 month of uptime.
>>
>>                     Evan.
>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.balabit.hu/pipermail/syslog-ng/attachments/20170605/a13affad/attachment-0001.html>