[Icecast] Frequent disconnects when relaying

[Raj Mathur]

Hi,

We're running primary server (A) that has two local streams from ices2 
0.2 and a secondary server (B) that relays the two streams from A and 
has one local stream from ices2 in addition.  Both A and B are in the 
same building, though not on the same LAN.

Recently we've started having B disconnect one channel or the other 
frequently from A.  All listeners on that channel on B get 
disconnected, and then B resumes the channel again after a while.  This 
always happens during our peak time around 15:00 UTC and continues 
while the peak time lasts (a few hours).  The setup was stable for 
about a year before that, and this is a new phenomenon.

A is running Icecast 2.3.1-5
B is running Icecast 2.3.2-2 (upgraded from 2.3.1 earlier today)
Both machines are on Debian Stable.

When the disconnect happens, the messages in the primary server's (A) 
log are:

[2009-02-27  18:03:28] INFO source/send_to_listener Client 662536 
(nn.nn.nn.nn) has fallen too far behind, removing
[2009-02-27  18:03:28] DBUG source/source_main Client removed
[2009-02-27  18:03:28] INFO source/source_main listener count on /XX now 
33
[2009-02-27  18:03:28] DBUG stats/modify_node_event update node 
slow_listeners (11626)

The messages on B are:

[2009-02-27  18:03:36] INFO source/get_next_buffer End of Stream /XX
[2009-02-27  18:03:36] INFO source/source_shutdown Source "/XX" exiting
[2009-02-27  18:03:36] DBUG source/source_clear_source clearing 
source "/XX"
[2009-02-27  18:03:36] DBUG stats/process_source_event delete source 
node /XX

[Sorry, lines have wrapped].

Appreciate any help in tracing and solving this problem.  Can provide 
pcap files if you let me know how much data would be useful.

Regards,

-- Raju
-- 
Raj Mathur                raju at kandalaya.org      http://kandalaya.org/
       GPG: 78D4 FC67 367F 40E2 0DD5  0FEF C968 D0EF CC68 D17F
PsyTrance & Chill: http://schizoid.in/   ||   It is the mind that moves