[sam-qfs-discuss] Random QFS mount hang

[McCreery, Lee CTR DISA]

Hi Bry,

We at one point had a race condition between SAMFS and NFS that caused
all kind of issues.  It might be the case that NFS shares out the base
directory before the SAM mount process can complete.  I am sure this is
not a documented procedure but it can work.

You might try to comment out the mount in vfstab, reboot, make sure to
"unshared /?????" after reboot, put volume back into vfstab, "kill -HUP"
the sam-fsd process to start the sam-sharefsd process for that volume,
attempt the mounting of /????? and re- "share /?????" the mount point
for NFS if needed.

But we have never seen not being able to "kill -9" the mount process. 

Just my 2c,

Lee

-----Original Message-----
From: sam-qfs-discuss-bounces at opensolaris.org
[mailto:sam-qfs-discuss-bounces at opensolaris.org] On Behalf Of Bryan
Collins
Sent: Tuesday, July 01, 2008 11:09 PM
To: 'sam-qfs-discuss at opensolaris.org'
Subject: [sam-qfs-discuss] Random QFS mount hang

Hi all,

We've been running a SAM/QFS shared client environment here for a couple
of years now, without any major dramas.

However, I have noticed an issue where a QFS client will not mount a
filesystem and leave the mount process hanging in the background.

MDS is V490 Solaris9, 4.6.25
QFS Client is 480R Solaris9 4.6.25

The client has 5 QFS filesystems, all mounted fine after reboot except
for 1.

The mount process is hung.
    root   428     1  0   Jul 01 ?        0:00 mount -o
shared,bg,sync_meta=1,meta_timeo=5,stripe=0,mm_stripe=1 pandoraaccess

Cant kill it (even -9 does nothing), truss shows absolutely 0 activity.
pstack shows nothing too
428:  mount -o shared,bg,sync_meta=1,meta_timeo=5,stripe=0,mm_stripe=1
pando
 00000000 ???????? (0, 0, 0, 0, 0, 0)

I cant temporarily mount the filesystem via NFS as the mountpoint is
being used on the system (device busy error from mount).

On the MDS side of things, tracing is enabled and I can see the
sam-sharefsd start up a thread for that filesystem/client combo.

I am going to give the client a reboot later tonight outside business
hours, but I'd like to know if theres another way of knocking
off this mount process so I can gain access to the filesystem on the
client.
I'd also like to hear if anyone else has come across this situation?
I am confident that everything is configured and running properly, as
this setup doesn't change and its been in production for a couple of
years (with the odd patch update and a regular monthly reboot).

Thanks
Bry

_______________________________________________
sam-qfs-discuss mailing list
sam-qfs-discuss at opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/sam-qfs-discuss