[Freeswitch-users] CPU availability?

[Craig Stevenson]

Perhaps an issue in packet throughput (packets-per-second) on the NIC or
any virtualized driver levels?

-- Craig


On Sat, Jul 11, 2020 at 10:57 AM Joel Serrano <joel at textplus.com> wrote:

> Have you checked the max-sessions (switch.conf.xml) and the ulimits etc
> (OS level)?
>
>
> On Sat, Jul 11, 2020 at 03:45 Sergey Safarov <s.safarov at gmail.com> wrote:
>
>> Need looks FreeSwitch logs.
>>
>> On Sat, Jul 11, 2020 at 12:53 PM David Villasmil <
>> david.villasmil.work at gmail.com> wrote:
>>
>>> I get to 2000 channels fine. My question is why can’t I higher without
>>> calls failing? FS never uses more than 40% CPU and 20 gb RAM.
>>> Config is basically vanilla, I unloaded most unnecessary modules, too.
>>>
>>> On Sat, 11 Jul 2020 at 04:59, Sergey Safarov <s.safarov at gmail.com>
>>> wrote:
>>>
>>>> > Just for info. Can you try on windows?
>>>>
>>>> That does not matter.
>>>> On Linux, I get results with 50 calls per second, every call continues
>>>> for 24 seconds, 1000 concurrent calls, and 10 000 total calls with only 3
>>>> failed calls. Tested kazoo cluster, more complex cluster then standalone
>>>> freeswitch.
>>>> So must work on Linux
>>>>
>>>>
>>>> On Sat, Jul 11, 2020 at 6:09 AM Sergey Safarov <s.safarov at gmail.com>
>>>> wrote:
>>>>
>>>>> SIPP can record failed call ID
>>>>> Using this call-id you search for FreeSwitch logs.
>>>>> Also, you can record all network traffic using dumpcap or tcpdump
>>>>> utility to disk. Later you can open these pcap files using sngrep utility
>>>>> and check waht is wrong with failed calls.
>>>>>
>>>>>
>>>>> On Sat, Jul 11, 2020 at 5:07 AM Alex Sibyakin <alex at freeswitch.com>
>>>>> wrote:
>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> What is your OS? What is your Sofia configuration?
>>>>>>
>>>>>> What the vmstat -w -S M 1 10 says?
>>>>>>
>>>>>> Can you try without Docker?
>>>>>>
>>>>>> Regards,
>>>>>> Alex
>>>>>>
>>>>>>
>>>>>> On Fri, 2020-07-10 at 19:53 +0100, David Villasmil wrote:
>>>>>>
>>>>>> Hello all,
>>>>>>
>>>>>> Let's assume i have a 28 CPU (56 cores) with 64gb RAM.
>>>>>> I'm load-testing that box with something like this:
>>>>>>
>>>>>> SIPp->--\       /----> FS1
>>>>>>          ---->FS
>>>>>> SIPp->--/       \----> FS2
>>>>>>
>>>>>> SIPp is simply sending calls with rtp to FS.
>>>>>> FS just forward the calls to FS1 and FS2 based on some very simple
>>>>>> decision.
>>>>>> FS1 and FS2 simply answer and execute echo.
>>>>>> At ~2000 channels FS fails to reply to some SIP messages, basically
>>>>>> calls are timing out, answer ratio goes down.
>>>>>> BUT CPU never goes over 45% on the FS and memory is ample. I don't
>>>>>> see packe drops on the interface.
>>>>>> FS is running in a docker.
>>>>>>
>>>>>> I'm wondering whether FS would use all of the available CPU or am I
>>>>>> hitting some other limit somewhere.
>>>>>>
>>>>>> Any insight is welcomed!
>>>>>>
>>>>>>
>>>>>> Regards,
>>>>>>
>>>>>> David Villasmil
>>>>>> email: david.villasmil.work at gmail.com
>>>>>> phone: +34669448337
>>>>>>
>>>>>>
>>>>>> _________________________________________________________________________
>>>>>>
>>>>>>
>>>>>>
>>>> _________________________________________________________________________
>>>>
>>>>
>>>> --
>>> Regards,
>>>
>>> David Villasmil
>>> email: david.villasmil.work at gmail.com
>>> phone: +34669448337
>>> _________________________________________________________________________
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20200712/9aba408a/attachment.html>