VoIP Mailing List Archives

[kawarod at laposte.net]

Hi list,

I'm testing mod_limit like this:
<action application="set" data="max_calls=60"/>
<action application="set" data="GW_IP=10.10.20.100"/>
<action application="limit" data="${AREA} ${GW_IP} ${max_calls}"/>

where AREA could be any value like: US/EUROPE/AFRICA... that has been
set for the call

then I use sipp for load testing with 4 cps and max 65 calls so that I
will be limited by the limit of 60 in the dialplan.

I use "limit_usage EUROPE 10.10.20.100" in cli and I see the limit value
growing up to 60.

Sipp still tries to establish new call (up to 65 calls) at 4cps and for
each new cps in excess, FS sends a 503.
I wait for 10 seconds and stop sipp, but the limit value is never
decreasing even when there is no more channels used (show channels), the
limit value is stuck to 60.


If I limit Sipp to 55 calls (below the limit value), the limit value
increase and decrease depending on load, and the pbm doesn't appear.

Does anybody is using mod_limit and have encountered the same pbm.

I'm using latest svn: 12580.

regards,
rod


_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

[mrene_lists at avgs.ca]

Were you doing transfers (action="transfer" in the dialplan) ?

If yes, retry with revision 12581, and open a JIRA if it still is an
issue (also include a full debug log (delete your freeswitch.log file
before doing your test and attach it after) in your JIRA)

Math

On 12-Mar-09, at 10:12 AM, rod wrote:

[cstomi.levlist at gmai...]

Hello,

we had the same problem. we couldn't test r12581 for sure yet, but we will.
This fix is only for the limit (db version), right?
Would be limit_hash a better choice to increase performance anyway?
Rod, do yoo have maybe experiences with limit_hash with your sipp?

Thanks in advance,
Tamas


Mathieu Rene írta:

[kawarod at laposte.net]

Hello,

I'm now running r12590, the pbm was still there, but this was because of
a broken dialplan.

I'm using this for exceeded limit:
<extension name="TOO_MANY_CALLS">
<condition field="destination_number" expression="^limit_exceeded$">
<action application="respond" data="503"/>
<action application="hangup"/>
</condition>
</extension>

but this extension was at the end of my dialplan and I matched an other
extension before reaching the limit extension.
What was odd was that in ngrep I saw FS sending a 503, I will
investigate on this.

I will rerun long term test and let you know if all is ok, so I have to
do for a previous mail related to ghost sessions in CLI.

@Tamas:
I just give a try to limit_hash, and didn't make many tests with it.
It's on my todo list. Limit_hash is not a better choice than limit,
their usage are different:
- limit_hash is a good way to rate limit a specific gateway for
example so that your switch won't be flooded by a misconfigured peer gw
- limit is for limiting concurrent call to an extension/gateway, eg
you have a peer that provides you with 30channels and you want to allow
15 channels for mobile (PLMN) and 15 for PSTN without relying

If you still have pbm with limit and svn, pay attention to you dialplan :p

@Mathieu
I hope you didn't work on a virtual pbm, cause it seems to be a dialplan
misconfiguration. I'll let you know if I still have pbm. Thanks for your
help.

regards.
rod

Tamas Cseke wrote:

[mike at jerris.com]

What is a pbm?

On Mar 13, 2009, at 4:51 AM, rod <kawarod@laposte.net> wrote:

[jalsot at gmail.com]

My guess is: pbm = problem :)

T.

Michael Jerris ¨ªrta:

[intralanman at freeswi...]

Tamas wrote:

[cstomi.levlist at gmai...]

Hello,

Thank you for your help.

limit,Limit,<realm> <id> [<max> [number [dialplan [context]]]]
limit_hash,Limit (hash),<realm> <id> [<max>[/interval]] [number
[dialplan [context]]]

I diged into mod_limit deeplier and as far as I understand, limit and
limit_hash can be used for the same thing:
limiting concurrent calls. please correct me, if I'm wrong...

but limit_hash has another feature /interval this is that make the rate
limitation you mentioned.

Our dialplan is good that is out of question. I see hangup in console.
we do <action application="hangup" data="BEARERCAPABILITY_NOTAVAIL"/>

I tested both of them with sipp and I wasn't able to reproduce the problem

r12581 did fix the case when limit is called more times, right?

Thanks,
Tamas


rod írta:

[mrene_lists at avgs.ca]

Yeah, you should be using limit_hash unless you need data replication
on multiple FS boxes since its a lot faster.

You can also do <action application="limit_hash realm id max/rate !
bearercapability_notavail" />
On another note, normal_circuit_congestion looks like a more
appropriate cause.

If you have problems, please open a jira including logs and a detailed
description of the required steps to reproduce the problem.

Math

On 13-Mar-09, at 12:19 PM, Tamas Cseke wrote:

[cstomi.levlist at gmai...]

Hello,

Thank you for your help.

limit,Limit,<realm> <id> [<max> [number [dialplan [context]]]]
limit_hash,Limit (hash),<realm> <id> [<max>[/interval]] [number
[dialplan [context]]]

I diged into mod_limit deeplier and as far as I understand, limit and
limit_hash can be used for the same thing:
limiting concurrent calls. please correct me, if I'm wrong...

but limit_hash has another feature /interval this is that make the rate
limitation you mentioned.

Our dialplan is good that is out of question. I see hangup in console.
we do <action application="hangup" data="BEARERCAPABILITY_NOTAVAIL"/>

I tested both of them with sipp and I wasn't able to reproduce the problem

r12581 did fix the case when limit is called more times, right?

Thanks,
Tamas


rod írta:

[kawarod at laposte.net]

Hi,

not too hard :p
but it's just a bad habit when I write in my native language (french). I
guess that this spelling is not too common for english speaker.

I'll do my best next time to write it correctly.

@tamas
you are right, we could use limit_hash the same way as limit when not
specifying the /rate

@Mathieu
did you suggest limit_hash is more scalable than limit? But I don't
understand why limit_hash is not suitable for data replication (DB
lookup for limit and memory for limit_hash??), even if I don't know how
to do it with limit.

regards.

Raymond Chandler wrote:

[mrene_lists at avgs.ca]

limit_hash uses a faster data structure then limit but works the same
way for tne end-user.

viens sur IRC si t'as des questions en francais =)

Math

On 17-Mar-09, at 3:06 AM, rod wrote:

[kawarod at laposte.net]

thanks Mathieu.

I setup an IRC account to give it a try.

Comme ça je pourrais t'embeter avec mes "pbms" :p

rod

Mathieu Rene wrote: