Subject: Re: NFS hangs on 2.0 client
To: Andrey Petrov <petrov@netbsd.org>
From: Jeff Rizzo <riz@tastylime.net>
List: tech-net
Date: 01/07/2005 12:15:33
Andrey Petrov wrote:
>It would be interesting to see on what wchans processes are waiting (ps -alx).
>I see similar behaviour on ibmnws with pcn nic. In my case client usually
>sleeps in getput and nfsd in nfsmblk, that means that nfs_writerpc waits
>for packet being sent, and client waits for nfsd to complete. I also see pcn reset
>by watchdog before hung and it seems pcn loses mbuf pointer without freeing it
>during reset.
>
> Andrey
>
>
Well, the two 'ls' and one 'df' that i've tried since the hang began (on
the client) are in vnlock:
6004 138 17990 0 -2 0 72 656 vnlock D p2 0:00.00 ls
-F /pkg
6004 24466 17990 0 -2 0 68 600 vnlock D p2 0:00.00 ls
-F /pkg
6004 7288 20140 0 -2 0 32 476 vnlock D p3 0:00.00 df -h
...but the find from the pkgsrc bulk build is in getblk:
0 16347 17264 0 10 20 152 748 wait IN+ p0 0:00.05
/bin/sh -c /usr/bin/find /pkg_comp/packages -type l -name
ap-php-4.3.10.tgz -print | /usr/bin/xargs /bin/rm -f
0 24707 16347 26 -5 20 180 768 getblk IN+ p0 0:00.27
/usr/bin/find /pkg_comp/packages -type l -name ap-php-4.3.10.tgz -print
0 25113 16347 31 2 20 336 512 piperd IN+ p0 0:00.01
/usr/bin/xargs /bin/rm -f
...anything more?
+j