Re: conserver uses 99% CPU

On Mon, Nov 21, 2005 at 01:45:35PM +0300, Alexandra N. Kossovsky wrote:
> After using home-made library, conserver sometimes runs with 99% CPU.

is there any way you can reliably reproduce this?  this sounds similar
to another issue i've heard of (but can't reproduce) - sometimes i
wonder if it's something with a particular linux kernel or gcc version.
also, is it a single conserver process that uses all the cpu, or do all
processes spike to maximum?

> Please note, that home-made library sometimes fails to close
> connection to conserver propely, but server should not behave in such a way
> if a client failed to do the things properly.

that should be ok.  the server is *supposed* to handle that.  but
perhaps there is a bug somewhere.

> home-made library is very simple: it connects, sends "^Ec;", logs all
> incoming data, tries to send "^Ec.", closes the socket.
> I'm ready to provide any necessary information.

if you can provide a way to reproduce this, that would be best.  aside
from that, if you run conserver with the -DDD flag, you'll get a ton of
output that will hopefully shed some light on what's going on when it
fails (but you'll need to get it to fail).

that's where i'd like to start...feel free to email me directly, and we
can post any findings (like a patch if we find the bug) to the list.

Bryan