[Icecast-dev] Ices host hangup with icecast2.2kh9 on server host

Karl Heyes karl at xiph.org
Sat Jul 9 04:01:43 PDT 2005


On Sat, 2005-07-09 at 09:02, Iceuse - Kris wrote:
> Hi,
> 
> I have the following config;
> 4 ices hosts, ices-kh60
> 
> 2 are streaming to icecast2.1kh5 server.
> 2 are streaming to icecast2.2kh9 server.
> 
> The 2 hosts streaming to icecast2.2kh9 are crashing (machine crash...) 
> twice a week.

> I was thinking of hardware problem, but it think it's not the case.

depends on the nature of the crash, a complete crash is probably because
of a device driver which may of been triggered by ices/rawrec. The only
other sort of thing that user apps can do to cause problems like this
are massive memory leaks.

> To feed ices, I use:

> rawrec | ices
> 
> Which I found very sensitive and may cause machine handup if any trouble.
> Is it possible that under certain circumstances, icecast2.2kh9 is 
> blocking something, causing ices freeze, then rawrec crashing the 
> machine while trying to do access sound card without being able to send 
> audio to ices ?

it's possible, in which case there's a few bugs showing up. icecast
should not have a problem with a stalling listener and drops them when
they are too far behind, ices-kh60 uses non-blocking sockets to connect
to icecast so should be processing the input and the device driver
should be able to handle a stall without crashing.

> This is the only reasonnable explanation to the problem I found.
> 
> What other solution may I use to read sound card input rather than using 
> rawrec ?

ices 2 has been able to read oss/alsa/sun input for some time, is their
a particular reason for rawrec?

> I saw in icecast2.2kh12e that there is some socket changes. Can this may 
> correct the above problem?
> (I will take several days to test for me)

12e has many changes in over kh9, the connection changes are related to
the processing incoming client requests using non-blocking sockets, so I
don't think it will affect you if the problem happened mid-stream. I
have fixed an odd case in the ogg handler which may cause it to loop
instead of go back to reading though.

Even a busy loop should not cause a crash though, so check for updated
drivers.

karl.




More information about the Icecast-dev mailing list