[Icecast] high CPU load for large # sources?

[Klaas Jan Wierenga]

Ok Karl,

Thanks for your help, I'll send the strace output after the next 'peak time'
:-).
You reply did ring a bell with me. I have fileserving on and it is also not
unlikely that someone is downloading a recorded broadcast.... silly me. I'll
investigate further when the situation arises again.

Cheers,
KJ

-----Oorspronkelijk bericht-----
Van: Karl Heyes [mailto:karl at xiph.org]
Verzonden: maandag 7 maart 2005 23:31
Aan: Klaas Jan Wierenga
CC: icecast
Onderwerp: Re: [Icecast] high CPU load for large # sources?


On Mon, 2005-03-07 at 22:01, Klaas Jan Wierenga wrote:
> Hi all,
>
> I have an icecast setup with 20+ sources. During peak times some 20
sources
> will be connected with a total of some 250 listeners more-or-less equally
> divided over the 20 sources. All streams are running at a measly 16 kbps.
> There is enough bandwidth to/from the server. During these peak times I
see
> very high CPU usage for icecast 98-99%. The system I'm running is an Intel
> Celeron 2.4GHz running Fedora Linux 1. It does appear that the server is
> happily serverving all clients, but what could cause the high CPU load?

I've only seen a busy CPU report on OS X which was relating to the file
serving.

> Is every source (/mount) serviced by a different thread? How many threads
> would I have then? My threadpool is set at 20. Could it be that the
timeout
> on poll/select of 250 in each thread causes Icecast to effectively
busywait
> since with 20 threads the actual time between the poll/select system calls
> is reduced to ~10ms, the timeslice of Linux!

Each source is handled by a separate thread, but there's no relationship
between those and threadpool, it's unlikely that a thread pool of more
than 2 or 3 is ever useful.

If poll/select are returning continuously before 250 ms then it's either
signal related (unlikely) or some fd has triggered, the poll timeout can
also drop to 0 on client bursts, but those are short lived.

I would determine which thread(s) are busy from ps, then use

strace -o output.txt -p <thread of busy icecast>

ctrl-C it after a few seconds, then email the output.txt file to me.

karl.