After restarting Kamailio on September 3rd, the problem has again gotten to the point of no call processing.

Here are the memory statistics; additionally, I can see the virtual memory has ballooned up to 400mb again.

shmem:fragments = 2378

shmem:free_size = 209840

shmem:max_used_size = 268393216

shmem:real_used_size = 268225616

shmem:total_size = 268435456

shmem:used_size = 220534792

-Tim

From: sr-users-bounces@lists.sip-router.org [mailto:sr-users-bounces@lists.sip-router.org] On Behalf Of Heenan, Timothy Steven
Sent: Wednesday, September 03, 2014 3:28 PM
To: 'miconda@gmail.com'; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

I have about 460 destination sets. In each set I could have from 1 address up to 6.

When I took that initial memory log, Kamailio had been restarted shortly beforehand. I just took another one after the system had decayed so Kamailio was responding to Invites with “SIP/2.0 500 No error (2/SL).” (exhibiting the issue after a few days) and the log is about 400Mb! I’ll attach a small snippet.

Seems to be going on and on about the dispatcher module.

-Tim

From: Daniel-Constantin Mierla [mailto:miconda@gmail.com]
Sent: Thursday, August 28, 2014 5:03 AM
To: Heenan, Timothy Steven; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

Looking at the logs, there are 1336 URIs, which seems ok given your numbers, because the module keeps also the previous set of records (needed because at reload time there can be kamailio worker processes using the records -- perhaps we can improve a bit here, I will review the current reload code a bit, since it was a contribution from another developer quite long time ago).

Some more questions:
- how many destination sets (groups) do you have? Can you estimate the minimum and maximum addresses in a set?
- when did you take the memory log? In other words, for how long was kamailio running? Did you wait enough to notice the steady increase of memory usage?

Cheers,
Daniel

On 28/08/14 01:56, Heenan, Timothy Steven wrote:

Thank you for the help.

For dispatcher, I’m using a database that contains about 700 records.

A reload is performed via cronjob every 5 minutes. The command being used is:

kamctl dispatcher reload

Thanks,

-Tim

From: Daniel-Constantin Mierla [mailto:miconda@gmail.com]
Sent: Wednesday, August 27, 2014 5:22 PM
To: Heenan, Timothy Steven; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

The logs suggest some leaks in dispatcher module -- it allocates shared memory only for caching routing records. Few more questions I need to get the answers in order to troubleshoot further.

How many records do you have in dispatcher table? Is it in database or the text file?

How ofter do you reload? What is the command you use for reload?

Cheers,
Daniel

On 26/08/14 22:42, Heenan, Timothy Steven wrote:

I have attached is the output of “kamcmd cfg.set_now_int core mem_dump_shm 1” to this email. I had to restart Kamailio to load ctl and ctl_rpc, so let me know if I need to do another one after some time passes and calls are processed.

Also, here is the output of “kamctl mi get_statistics shmem:”

shmem:fragments = 398

shmem:free_size = 259170720

shmem:max_used_size = 9265184

shmem:real_used_size = 9264736

shmem:total_size = 268435456

shmem:used_size = 8559768

Thanks!

-Tim

From: Daniel-Constantin Mierla [mailto:miconda@gmail.com]
Sent: Tuesday, August 26, 2014 10:46 AM
To: Heenan, Timothy Steven; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

It is not needed to see the memory operations logs. Just to dump the chunks in memory.

You would need ctl and ctl_rpc modules (iirc -- they should be in the default config file), then run:

kamcmd cfg.set_now_int core mem_dump_shm 1

Extract the logs of the dump from the shared memory and send them over.

Send also the output for:

kamctl mi get_statistics shmem:

Cheers,
Daniel

On 26/08/14 17:34, Heenan, Timothy Steven wrote:

Thank you.

I have recompiled with MEMDBG, however I am not observing any additional output in my log files.

I think I have something set incorrectly in kamailio.cfg.

Here are my log settings from Kamailio.cfg

debug=3

log_stderror=no

log_facility=LOG_LOCAL0

memdbg=2

memlog=2

…

modparam("debugger", "cfgtrace", 1)

modparam("debugger", "log_level", 2)

modparam("debugger", "cfgpkgcheck", 1)

modparam("debugger", "mod_level", "core=2")

modparam("debugger", "mod_level", "tm=2")

modparam("debugger", "mod_level", "dispatcher=2")

modparam("debugger", "mod_level", "siptrace=2")

From: Daniel-Constantin Mierla [mailto:miconda@gmail.com]
Sent: Tuesday, August 26, 2014 3:54 AM
To: Heenan, Timothy Steven; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

Hello,

I quickly looked over the modules that use shared memory and I couldn't spot a commit related to any leak.

Have you had the chance to compile with MEMDBG?

Cheers,
Daniel

On 19/08/14 19:14, Heenan, Timothy Steven wrote:

Do you think any of these modules possibly causing this issue?

-Tim

From: sr-users-bounces@lists.sip-router.org [mailto:sr-users-bounces@lists.sip-router.org] On Behalf Of Heenan, Timothy Steven
Sent: Tuesday, August 12, 2014 2:37 PM
To: 'miconda@gmail.com'; Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

Hi Daniel,

Here’s a list of the modules I’m running:

mi_fifo

db_mysql

sl

kex

tm

tmx

rr

xlog

maxfwd

usrloc

registrar

textops

pv

acc

permissions

siputils

auth

lcr

dispatcher

sanity

debugger

siptrace

We’re setting the memory allocation to 256 ( -m 256 –M 64 ).

I’ll see about recompiling with MEMDBG=1.

Regards,

- Tim

From: sr-users-bounces@lists.sip-router.org [mailto:sr-users-bounces@lists.sip-router.org] On Behalf Of Daniel-Constantin Mierla
Sent: Tuesday, August 12, 2014 4:17 AM
To: Kamailio (SER) - Users Mailing List
Subject: Re: [SR-Users] Kamailio 4.0.4 slow memory leak

Hello,

can you give the list of loaded modules? How much memory did you allocate (-m parameter value)?

It will narrow down searches to see if there was anything similar fixed since 4.0.4.

To troubleshoot easier, would be good to recompile with MEMDBG=1, then the details of chunks in memory can be dumped and analysed.

Cheers,
Daniel

On 12/08/14 01:40, Heenan, Timothy Steven wrote:

Greetings,

I’m running into a slow memory leak on my kamailio 4.0.4 SIP proxies. I’m observing a steady increase in the memory consumption until there is no more left. Kamailio then starts repeating this in the logs:

ERROR: dispatcher [dispatch.c:279]: add_dest2list(): no more memory.

What would be the best way to debug this kind of a memory leak? The proxy does not handle any registrations but does route a fair amount of calls.
-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Next Kamailio Advanced Trainings 2014 - http://www.asipto.com
Sep 22-25, Berlin, Germany ::: Oct 15-17, San Francisco, USA
This email message and any attachments are for the sole use of the intended recipient(s). Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message and any attachments.

--

Daniel-Constantin Mierla

http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda

Next Kamailio Advanced Trainings 2014 - http://www.asipto.com

Sep 22-25, Berlin, Germany ::: Oct 15-17, San Francisco, USA