<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hello,<br>
<br>
On 8/17/11 1:41 PM, Sebastian Damm wrote:
<blockquote
cite="mid:CABkWSFyVUkjEG--WZsUER0O_=_3dpO6wCxZ+JER1sRc4AAZ10A@mail.gmail.com"
type="cite">Hi Henning,<br>
<br>
<div class="gmail_quote">On Wed, Aug 17, 2011 at 11:33 AM, Henning
Westerholt <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:henning.westerholt@1und1.de">henning.westerholt@1und1.de</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On Wednesday 17 August 2011, Sebastian Damm
wrote:<br>
> we have a script running on our Kamailio servers which
dumps all online<br>
> users with kamctl ul show, parses it and then sends it
to database. (For<br>
> explanation, we started doing it this way because we
noticed at least with<br>
> an older (open)ser release that SER would completely
stop working for a<br>
> couple of seconds while writing registered users to
database every 5<br>
> minutes. When dumping it, we don't see any disruption.)<br>
<br>
</div>
what about fixing this bug instead of using this workaround?
This sounds<br>
really inefficient.<br>
</blockquote>
<div><br>
Actually, piping the output through awk and generating a file
for LOAD DATA INFILE" is not too bad. Additionally, we need to
have that data in more than one DB, which prevents us from
using one of the built-in db modes. We use an own DB mode
which reads only on startup and then never touches the DB
again.<br>
</div>
<blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt
0.8ex; border-left: 1px solid rgb(204, 204, 204);
padding-left: 1ex;">I don't think there is a hard limit. Do
you see any errors (e.g. memory<br>
allocation issues) in the server logs? Another idea, if you're
somehow stuck<br>
to the old/ buggy version of the server, what about using
DB_ONLY mode?<br>
</blockquote>
<div><br>
I don't see any errors in the log. And actually, that's the
only bug left preventing our migration from 1.5 to 3.1. And
I'd say, DB_ONLY is not an option, should be a lot less
performant than if Kamailio keeps the state in memory
(although I haven't benchmarked it). <br>
<br>
Other ideas?<br>
</div>
</div>
</blockquote>
not sure what was the default in latest 1.x, maybe MI was using the
system memory manager. Since location records can need quite some
memory if there are lot of entries, it may be better to use the
system memory directly in MI to build the response to the commands.<br>
<br>
In 3.1, the option is to edit lib/kmi/Makefile and turn on the
define of MI_SYSTEM_MALLOC<br>
<br>
Then recompile/reinstall and try again. Let us know the results.<br>
<br>
As alternative, in 3.1, you can do RPC command with sercmd through
ctl module in order to dump location records. There you can control
the size of the buffers used to build the output:<br>
<br>
*
<a class="moz-txt-link-freetext" href="http://kamailio.org/docs/modules/stable/modules/ctl.html#binrpc_max_body_size">http://kamailio.org/docs/modules/stable/modules/ctl.html#binrpc_max_body_size</a><br>
<br>
Cheers,<br>
Daniel
<pre class="moz-signature" cols="72">--
Daniel-Constantin Mierla -- <a class="moz-txt-link-freetext" href="http://www.asipto.com">http://www.asipto.com</a>
Kamailio Advanced Training, Oct 10-13, Berlin: <a class="moz-txt-link-freetext" href="http://asipto.com/u/kat">http://asipto.com/u/kat</a>
<a class="moz-txt-link-freetext" href="http://linkedin.com/in/miconda">http://linkedin.com/in/miconda</a> -- <a class="moz-txt-link-freetext" href="http://twitter.com/miconda">http://twitter.com/miconda</a></pre>
</body>
</html>