Re: 2.613: network write socket problems

[Bernd Schubert <bernd.schubert@pci.uni-heidelberg.de>]

On Monday 12 September 2005 17:39, Bernd Schubert wrote:
> Hello,
>
> on last Friday  we switched on our server to 2.6.13 and today we are
> experiencing problems with our nfs clients.
> In particular I'm talking about the unfs3 daemon, not the kernel nfs
> daemon. Both are running on the server but on different ports, of course. 
> Both are also serving to the same clients, but different directories.
>
> Today it already several times happend that the unfs3 daemon stalled.
> Ethereal showed no network packages on the unfs3 daemon port during this
> time. A strace to the proc-id of the daemon clearly shows that *some*
> writes to some network sockets will take ages to finish
>
> write(37, "\200\0\0x\203\326(\5\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 124)
> = 124

Sorry for the noise, its not a kernel problem. Switching back to 2.6.11 didn't 
help, so we investigated further. It turned out, that one of our clients was 
in a kind of a zombie state and asking for filehandles, but not answering 
request from the server. Since unfs3 is only single threaded, all other 
clients had to wait for timeouts.

Bernd
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/