<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<pre>Hello all,
I'm having some interesting problems with our Freeswitch system.
I have found Tim's posting at <a class="moz-txt-link-freetext" href="https://lists.freeswitch.org/pipermail/freeswitch-users/2012-February/080422.html" moz-do-not-send="true">https://lists.freeswitch.org/pipermail/freeswitch-users/2012-February/080422.html</a>
It seems very similar but does not help me... Wish it did.
Hardware: Dell R610, dual 6 core Xeon X5680 @ 3.3GHz with 10 Gig of RAM
Network: Dual 10Gig Fiber to 40Gig fabric
Databases: also Dell R610 with dual 10Gig and Mariadb 10.5.9 NO ISSUES - VERY FAST - ULTRA LOW LATENCY
We have 2 different FS systems
1) One acting as an SBC of sorts - process SIP/RTP
2) One acting as a PBX - process SIP/RTP
Providers are connected to SBC (public - context)
PBX (sbc - context) connects to SBC (internal - context)
End Points connect to PBX (device - context)
picographically:
ITSP --- SBC --- PBX --- EndPoints
The machine running FreeSWITCH as an SBC is fine - 50 calls (100 sessions) and near zero load average - CPU idle 98% - ALL GREAT
About 100 devices registered, all NAT'd, approx 50 subscription via BLF buttons
Registration period is maxed to 60 minutes and most devices use the default of 60 minutes
Typically, We see only a few CPS, maybe 5 if it's really busy
All our trunks are SIP. Only using PCMU/8000 codec.
We use MARIADB in the core - as well as for sofia, voicemail, and anything else that can use it.
Config, Directory and dialplan are done using CURL and handled via apache2 on the same server over 127.0.0.1.
I keep hearing about FS systems handling thousands of users on one box.
I'm nowhere near that and it seems to be maxed out.
Freeswitch on both systems is Version 1.10.6-release git 1ff9d0a 2021-03-25 13:16:09Z 64bit
So... The problem
The machine running FreeSWITCH as a PBX seems to be struggling during periods of any usage - even only 1 call (2 channels).
Just a moment ago, load average is 4.5 - CPU Idle is 65% - 8 sessions active - 315 sessions since startup - 45 threads - 8 hours running
The audio seems fine - no complaints reported..
There is a very noticable chocking on the BLFs though... Every few minutes, the BLFs just freeze - yes - freeze... Then they catch up
for example, for testing this, I have programmed a BLF on one phone.
I run manually "luarun /etc/freeswitch/scripts/on.lua *8896@domain.com" - the lights turns red
then after a few seconds,
I run manually "luarun /etc/freeswitch/scripts/off.lua *8896@domain.com" - the light turns green
Sometimes, when I run these manually, the lights take 30 seconds to change - sometimes - usually, under a second.
They change the state of the subscribe target
Here is the LUA script to turn the light ON(RED)
##############################################################
#!/lua
local random = math.random
math.randomseed(os.time())
local function uuid()
local template ='xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'
return string.gsub(template, '[xy]', function (c)
local v = (c == 'x') and random(0, 0xf) or random(8, 0xb)
return string.format('%x', v)
end)
end
local function SendCallDirectorBLFevent()
Uuid = uuid()
local event = freeswitch.Event("PRESENCE_IN")
event:addHeader("proto", "sip")
event:addHeader("event_type", "presence")
event:addHeader("alt_event_type", "dialog")
event:addHeader("Presence-Call-Direction", "outbound")
event:addHeader("from", argv[1])
event:addHeader("login", argv[1])
event:addHeader("unique-id", Uuid )
event:addHeader("answer-state", "confirmed")
event:fire();
end
SendCallDirectorBLFevent()
##############################################################
The issue is the choking on the BLFs and the HIGH load avg for such small amounts of calls on the PBX - the SBC seems fine.
SO.....
I have been trying to resolve this for about a month - looking everywhere I could.
I have monitored the network stats, disk IO stats, cpu usage, processes, etc.... I don't see anything different between the SBC and PBX stats...
I would like the BLFs to show the correct state a little faster (more like half a second, not 10-60 seconds after the event) and would I like to know how to fix the high CPU load on the PBX.
I have confirmed 2 things today:
1) it's not using mod_mariadb as I put back the ODBC and not help
2) this is seems most noticeable when new calls come in. The BLFs pause to 5-8 seconds and continue - there is a noticeable spike in load avg when calls keep coming in for 5 minutes...
Any information that would be helpful?
Thanks,
Jerry
</pre>
</body>
</html>