[Icecast] icecast crash (segv)

[Laurent CORBES]

Hi all,

I'm experiencing a systematic crash with icecast2 when I goes in this case
(tested with debian 2.3.1-5 2.3.1-6 and latest trunk)

When there is more than ~1400 clients connected and a mountpoint that was not
relaying goes online it crash.
Without this I can go up to 5000 clients without problems. Changing the number
of threadpool does'nt change anything.

here is the term log:
stream-04:/etc/icecast2# icecast2 -c /etc/icecast2/icecast.xml
Changed groupid to 105.
Changed userid to 102.
Segmentation fault
stream-04:/etc/icecast2# 


error.log in debug:

[2008-01-31  16:38:58] DBUG source/source_main Client added
[2008-01-31  16:38:58] DBUG source/source_main Client added
[2008-01-31  16:38:58] DBUG source/source_main Client added
[2008-01-31  16:38:58] INFO source/source_main listener count on /fbidf/all/fbidf-32k.mp3 now 966
[2008-01-31  16:38:58] DBUG format/format_check_http_buffer processing pending client headers
[2008-01-31  16:38:58] DBUG format/format_check_http_buffer processing pending client headers
[2008-01-31  16:38:58] DBUG format/format_check_http_buffer processing pending client headers
[2008-01-31  16:38:58] DBUG auth/add_listener_to_source max on /fbidf/all/fbidf-32k.mp3 is -1 (cur 966)
[2008-01-31  16:38:58] DBUG auth/add_listener_to_source Added client to /fbidf/all/fbidf-32k.mp3
[2008-01-31  16:38:58] DBUG auth/add_authenticated_listener client authenticated, passed to source
[2008-01-31  16:38:58] DBUG auth/add_listener_to_source max on /centpourcent/all/centpourcent-128k.mp3 is -1 (cur 910)
[2008-01-31  16:38:58] DBUG auth/add_listener_to_source Added client to /centpourcent/all/centpourcent-128k.mp3
[2008-01-31  16:38:58] DBUG auth/add_authenticated_listener client authenticated, passed to source
[2008-01-31  16:38:58] DBUG source/source_main Client added
[2008-01-31  16:38:58] INFO source/source_main listener count on /centpourcent/all/centpourcent-128k.mp3 now 911
[2008-01-31  16:38:58] DBUG format/format_check_http_buffer processing pending client headers
[2008-01-31  16:38:58] INFO slave/start_relay_stream Starting relayed source at mountpoint "/telerama/all/telerama-128k.mp3"
[2008-01-31  16:38:58] INFO slave/open_relay_connection connecting to 10.10.30.49:9042

And nothing more.

Here is my config files general options:
<icecast>
    <limits>
        <clients>8192</clients>
        <sources>1024</sources>
        <threadpool>128</threadpool>
        <queue-size>524288</queue-size>
        <client-timeout>1</client-timeout>
        <header-timeout>15</header-timeout>
        <source-timeout>60</source-timeout>
        <!-- If enabled, this will provide a burst of data when a client 
             first connects, thereby significantly reducing the startup 
             time for listeners that do substantial buffering. However,
             it also significantly increases latency between the source
             client and listening client.  For low-latency setups, you
             might want to disable this. -->
        <burst-on-connect>no</burst-on-connect>
        <!-- same as burst-on-connect, but this allows for being more
             specific on how much to burst. Most people won't need to
             change from the default 64k. Applies to all mountpoints  -->
        <burst-size>65536</burst-size>
    </limits>

And my mountpoints are all defined like this:

   <relay>
          <max-listeners>10000</max-listeners>
          <server>10.10.30.49</server>
          <port>9036</port>
          <mount>/</mount>
          <local-mount>/XXX/all/XXX-32k.mp3</local-mount>
          <on-demand>0</on-demand>
          <relay-shoutcast-metadata>1</relay-shoutcast-metadata> 
          <no-yp>1</no-yp>
   </relay>

The relaying source server is a shoutcast.

I don't think this is a file descriptor limitation as I can have more than 5000
clients and 65536 open fd. 

Thanks in advance for any explanation.

-- 
Laurent Corbes - laurent.corbes at smartjog.com
+33 (0)1 4996 6325
Smartjog SA - http://www.smartjog.com/