sshfs/nfs cause server lockup - resolved

[Chris]

On 19/12/06, Kris Kennaway <kris at obsecurity.org> wrote:
> On Tue, Dec 19, 2006 at 08:20:21PM +0000, Chris wrote:
> > On 18/12/06, Kris Kennaway <kris at obsecurity.org> wrote:
> > >On Mon, Dec 18, 2006 at 12:39:13AM +0000, Chris wrote:
> > >> On 14/12/06, Kris Kennaway <kris at obsecurity.org> wrote:
> > >> >On Thu, Dec 14, 2006 at 01:28:48AM +0000, Chris wrote:
> > >> >
> > >> >> It does make sense if thats the problem since the entire server even
> > >> >> locally stops working properly, and it always follows a unexpected
> > >> >> nfs/sshfs disconnection ie. network timeout.
> > >> >>
> > >> >> I am now running 6.2-RC that has the new file and currently at 1 day
> > >> >> 11hrs uptime.
> > >> >
> > >> >OK, thanks for following part of the advice I gave a month ago ;) Let
> > >> >us know if the problems persist.
> > >> >
> > >> >Kris
> > >> >
> > >> >
> > >> >
> > >>
> > >> Early today the nfs hub was rebooted so had a unexpected disconnection
> > >> also noted by the sshfs timeout prompt waiting for me in the terminal
> > >> , was able to remount fine and no server lockup or other probolems.
> > >>
> > >> Current uptime is 5 days, 10:48
> > >
> > >OK, good to know.
> > >
> > >Thanks,
> > >Kris
> > >
> > >
> > >
> > >
> >
> > Some bad news, I was offline for a day here, then I logged in today
> > reattached to screen, and was greeted with a timeout message to the
> > sshfs server, at this point server still functioning fine.  When I ran
> > the sshfs command again it locked, with only pings responding and had
> > to hard reboot it.
> >
> > I will setup my local machne now so I can do proper debugging for you.
>
> OK, it's (still) probably an sshfs bug though.
>
> Kris
>
>
>

Ok how to repeat the bug everytime.  Works on sshfs and nfs.

First.

The server died again (hub having its own problems so causing lots of timeouts).
This time instead of remounting I tried to ls the 2 mounts simply list
empty dirs, first dir worked and 2nd dir caused lockup, so its some
kind of problem with the filesystem nodes or something.

With this in mind on my local box I yanked out the network cable
causing a unexpected timeout, box hung, tried to do the ddb procedure
but didnt work, I may have been doing it wrong.

Booted local box again mounted nfs over internet and tried same thing
yanked out network cable, same thing accessing the dir where nfs mount
to hung server hard reboot needed.

Local box using 6.2-RC as well.  GENERIC kernel default make.conf.

Chris