ffs snapshot lockup
kostikbel at gmail.com
Thu Oct 5 08:30:36 UTC 2006
On Wed, Oct 04, 2006 at 05:16:53PM -0400, Vivek Khera wrote:
> On Oct 3, 2006, at 4:43 AM, Kostik Belousov wrote:
> >>Details are posted at http://vivek.khera.org/scratch/crashlogs/
> >>I have the crashdumps available to a kernel hacker upon request (i'd
> >>rather not make them generally available to the public...)
> >It seems that you have snapshotted fs exported by nfsd ? At least,
> >18a is
> >definitely the case. I have the patch (for current) that shall fix
> >the issue.
> >In fact, you need two patches:
> As per advice of Kris Kenneway, I turned off the software watchdog to
> rule out that as my problem. Then I ran a level 3 dump. Dump of root
> fs went fine, then it proceeded to do /usr. After a few minutes it
> locked up. Typescript 20 at the above URL shows the debugging info
> from the break into debugger of the locked up system. Since /usr was
> locked, nobody could log in at all.
> The network load was minimal at the time. I had everyone log out and
> close mail etc.
What were the symptoms of locked system ? Could you log in on console, or
do something at the shell prompt on console ?
Also, did the system respond to the pings ? Fs-related deadlocks (as
well as stalled disk io) usually do not prevent lowest levels of the
isr/network stack from working.
Again, I do not see the fs deadlock per se in the supplied script. Dump does
disk io, it seems that nfsd tries to serve some request. Sshd looks to be
ready to accept connections.
If console is available, but ping responses not arrive, this is definitely
network card problem.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20061005/6eabcb12/attachment.pgp
More information about the freebsd-stable