lkml.org 
[lkml]   [2003]   [Apr]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: loopback behaviour under high load ...
On Wed, 2 Apr 2003, Martin Josefsson wrote:

> On Wed, 2003-04-02 at 18:23, Davide Libenzi wrote:
> > On Tue, 1 Apr 2003, Davide Libenzi wrote:
> >
> > > And netstat shows Recv-Q=0 for the server socket, and Send-Q=N for the
> > > client socket. This has been tested on 2.5.66 vanilla.
> >
> > An update on this. Kernel 2.4.20+epoll does not have this problem.
>
> I see something similar with 2.5.64-bk10 (seen with earlier 2.5 kernels
> as well).
>
> I use distcc on my machine which is running 2.5.64-bk10 and using two
> remote servers running 2.4.20 to do the compiling.
>
> Sometimes I see a ~120s stall in the middle of a connection with
> Recv-Q=0 on the server (2.4.20) and Send-Q=N on the client (2.5.64-bk10)
> (it's the client that's sending data to the server)
>
> x.x.x.x = client
> y.y.y.y = server
>
> As you can see there's a long delay between 16:02:09 and 16:04:24
>
> There's some packetloss here (probably a congested interface in the
> router on the way). The dump was made on the client.

Hmm, we don't see any dropped packet at interface ( loopback ) level here:

$ cat /proc/net/softnet_stat
002a0c14 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
$ ifconfig
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:1349417 errors:0 dropped:0 overruns:0 frame:0
TX packets:1349417 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0




- Davide

-
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/

\
 
 \ /
  Last update: 2005-03-22 13:34    [from the cache]
©2003-2011 Jasper Spaans