[Freeswitch-users] Memory leak

[Michael Jerris]

I’ve fixed a few memory leaks in the last 60 days (now in the 1.6.10 release).  If you can capture ones with a specific set of call flows in valgrind, that would be the path to try to fix them if there is indeed a leak.  Just a graph showing memory use growth isn’t a particularly useful metric to find a specific issue, or often to even know if there is an issue.  If you suspect a problem, capturing the exact allocation that is not being freed using valgrind or clang memory tools, and coming up with a reproducible scenario to create the issue, is the only way we can address probes like these.

Mike


> On Sep 13, 2016, at 2:12 PM, David Villasmil <david.villasmil.work at gmail.com> wrote:
> 
> These are just a couple of them:
> Note the sessions and uptime
> <IMG_9202.PNG><IMG_9200.PNG>
> 
> On Tue, Sep 13, 2016 at 8:07 PM David Villasmil <david.villasmil.work at gmail.com <mailto:david.villasmil.work at gmail.com>> wrote:
> I don't have the data with me now, but i have servers that wave been running fs for literally months, last time i checked it had processed like 700k+ calls
> 
> On Tue, Sep 13, 2016 at 7:04 PM Anthony Minessale <anthony.minessale at gmail.com <mailto:anthony.minessale at gmail.com>> wrote:
> Your prize for 47 days uptime is probably a new version! 
> That is the best way to check for memory leaks, update whenever you can.
> 
> 
> 
> On Tue, Sep 13, 2016 at 11:41 AM, Florent Krieg <flokrrr at gmail.com <mailto:flokrrr at gmail.com>> wrote:
> Hello,
> 
> We experience the same behaviour here.
> 
> UP 0 years, 47 days, 19 hours, 27 minutes, 31 seconds, 208 milliseconds, 341 microseconds
> 2172632 session(s) since startup
> 
> cat /proc/11818/status shows:
> VmPeak:  6829064 kB
> VmSize:  6778648 kB
> VmLck:         0 kB
> VmPin:         0 kB
> VmHWM:    442120 kB
> VmRSS:    428764 kB
> VmData:  6561024 kB
> VmStk:       136 kB
> VmExe:        24 kB
> VmLib:     29644 kB
> VmPTE:      1828 kB
> VmSwap:        0 kB
> Threads:        507
> 
> And free says:
>              total       used       free     shared    buffers     cached
> Mem:           15G        15G       181M       188M       170M        14G
> -/+ buffers/cache:       1.2G        14G
> Swap:          15G        28M        15G
> 
> 
> Dunno if it helps, but we don't appear to meet any issue with calls (no crash or nothing, on other servers we have it has been running for months or even years with full memory if I'm not mistaken).
> 
> 
> Regards
> Florent
> 
> 2016-09-13 18:11 GMT+02:00 Anthony Minessale <anthony.minessale at gmail.com <mailto:anthony.minessale at gmail.com>>:
> That looks normal to me.
> FreeSWITCH needs a minimum of 2 GB dedicated ram for prolonged use.
> If you chart goes past 2 to 2.5 gigs, you may have a problem then.
> 
> you can run valgrind but you can only run 1 call at a time testing typical callflow.
> valgrind --tool=memcheck --log-file=vg.log --leak-check=full --leak-resolution=high --show-reachable=yes /path/to/freeswitch -vg
> 
> 
> 
> Also when those months are over, hopefully you update cos you can keep running the same FS for many months anyway ;)
> 
> 
> 
> 
> 
> On Tue, Sep 13, 2016 at 9:29 AM, Sergey Safarov <s.safarov at gmail.com <mailto:s.safarov at gmail.com>> wrote:
> I has configured FreeSwitch process memory usage graph.
> According this graph durring 24 days size of used memory ingrezed about two times.
> <FS-memory-chart.png>
> Pastebin of FreeSwitch process memory map is placed at https://pastebin.freeswitch.org/view/9e66572a <https://pastebin.freeswitch.org/view/9e66572a> 
> Are you have any suggestion how to find memory leak.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20160914/c5ee5789/attachment.html