[Spce-user] CPU spike when navigating UI?

Alex Lutay alutay at sipwise.com
Fri Apr 10 04:41:58 EDT 2015


Dear Kalen,

It is hard to say anything without access to server/logs.

For sure new releases gets new functionality/components, like Redis,
mod_perl migrated to fast-cgi, kamailio got new functions,
as a result requires more memory.. etc.
As an example see: 
/usr/share/ngcp-cfg-schema/cfg_scripts/config/15066_increase_proxy_pkgmem.up 
which increases 16->24 Mb
kamailio RAM usage (per thread!).

So, it is possible that mr3.3.1 hardware was enough but
close to limit while new version jumper over the limit.

Can you please check swap usage on your system,
maybe memory is not enough now. CPU/IO usage is also should be checked.

Finally you can reboot your server into previous kernel 3.2.0-4 (which 
was used in mr3.3.1). It should be still available/installed.
Please do this as a final resort.

Also you can switch your system to low performance mode:
https://www.sipwise.com/doc/mr3.8.1/spce/ar01s14.html#requirements-and-performance 
(see script ngcp-toggle-performance-config there).

NOTE: it is really LOW PERFORMANCE, which works well in Sipwise QA env 
on 1 core 1,5Gb RAM. but you can check the values it decreases and 
decrease them manually/partially. Usually ngcp-panel eats a lot of 
memory, so you can safe a lot there.

Please share your results here.
Also please share your current server details.
Thank you!

P.S. I am worried about your phrase:
 > ... a few weeks ago this problem began...

Maybe new virtual servers were started on the same hardware and
NGCP have to share physical resources now? Normally "after-upgrade" 
issues start on next day(s), but not weeks.

On 10/04/15 09:52, Kalen Krueger wrote:
> We've been running sip:wise since mr3.3.1 on the same virtual
> hardware for almost a year and have never had the problem... as soon
> as we upgraded from 3.5.1 -> 3.7.2 a few weeks ago this problem
> began... and it appears to be intermittent.   Previously we thought
> it was only occurring if we were in the GUI making adjustments, which
> is when I started this thread.   In this particular case we weren't
> doing anything in the GUI... it was actually after business hours for
> us.  We had a subscriber contact us to report the problem with
> calls... and our only short term fix was the restart the system.
>
> Also, we haven't added any new subscribers of significance since the
> recent upgrade.   It certainly could be that the new version requires
> more resources, but I looked through our performance monitoring over
> for the last several weeks, and it appears that the CPU requirements
> have generally gone down since the upgrade.   So a bit stumped right
> now....

-- 
Alexander Lutay
Head of Quality Assurance



More information about the Spce-user mailing list