Re: Problems with hangs under NetBSD-5.x

[Matthew Mondor <mm_lists%pulsar-zone.net@localhost>]

On Fri, 31 Jul 2009 20:47:17 +0200
Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote:

> Could all of you seeing these hangs give the following details:
> - are you all on i386, or is it also seen with amd64 ?
> - what disk subsystem are you using ? is these some raidframe volume ?
> - what mount options (quota, softdep, log, ...)
> 
> I'm seeing a hang that is currently preventing me from upgrading my
> NFS server to netbsd-5. I can't reproduce it on a test server (would be too
> easy :(. It has ffs without softdeps or log but with quotas, raidframe,
> ATA and scsi disks (external RAID device really, attached to two esiop).
> It's stable running NetBSD 3.x ...

The only "hang" problem I have observed so far were netbsd-5 nfs client
processes locking in nfsrcv wchan when using nfs over udp (this doesn't
occur over tcp, yet I often see these with tcp, and it recovers after a
short delay every time):

nfs server foo:/bar: not responding
nfs server foo:/bar: is alive again

The server never seemed affected in my case, only hung client processes.
The systems are i386, no raid, no quota, no softdep but wapbl is used.
Increasing the nfs client and server threads/processes didn't help in
both udp and tcp nfs modes.  The network appears reliable otherwise
including for ssh/http and udp multicast so I doubt network issues are
causing the hangs when using udp.

As for the nfs-udp hangs, the hanging client boxes had to be rebooted
to recover.  This at first seemed to occur rarely, but after it
happened twice the same day I decided to go back to nfs-tcp.

Thanks,
-- 
Matt