On Tue, 30 Mar 2010 10:44:38 -0400, J. Bruce Fields wrote:
> On Fri, Mar 19, 2010 at 06:10:38PM -0400, Dennis Nezic wrote:
> > On Fri, 19 Mar 2010 00:27:20 -0400, Dennis Nezic wrote:
> > >
> > > Both the patches mentioned in that thread already seem to have
> > > been applied to my kernels. So, although the problem seems
> > > related, it wasn't that bug in particular. The person in that
> > > thread was talking about mounts dying after 5-15minutes, which
> > > doesn't happen with me -- my problem only seems to occur under
> > > intense activity.
> >
> > Hrm. I just noticed that my scp transfers are stalling -- which also
> > didn't used to happen before with my old kernel. No error messages.
> > Ftp transfers work fine. Eek. :S. (Despite the freezing/stalling, my
> > *actual* network connection works perfectly.)
>
> That also suggests some network problem.... Is the scp problem
> reproduceable? Are packets getting dropped?
The scp problem is quite reproduceable -- when it decides to act up, it
quite consistently freezes/stalls at the same point (+/- a few (dozen)
MB) -- at least when testing roughly at the same time. I usually give
up after about 10 attempts. Sometimes, when it's a full moon, it will
work after ten attempts. (Restarting sshd has no effect.)
I haven't done any tcpdump yet.
I suspect the nfs stalls are also similarly reproduceable -- except
it's harder to tell since it doesn't display the progress as nicely as
scp :b. However, I did notice that nfs transfers very often stall at the
very beginning (I'm not sure if it's at byte 0, or a few MB in) -- as
well as at various points in the middle. (It "feels" like a bursting
buffer problem -- I remember maany times mplayer playing songs over
nfs, and having it stall before beginning the next song -- it buffered
about 32KB, but was waiting for a few more before it could start
playing. Very annoying :|.)
>
> Also: is the kernel dumping any backtraces into the server's logs?
Nothing suspicious in the system logs, nor in verbose mode.
I'll try a different wifi driver soon, and see if the problem persists.