AdvFS domain not mounting

[Flávio Gondim de Morais]

Hi for everyone.

I've searched for all over and did not found any answers for my problem, so i think you ppl could give me a help.

Here is my problem: While the system is booting, one domain freezes to mount. OK, so i removed it from fstab and system is up, no problem. I check for voldisk list, volprint -ht and syminq and it is OK(plexes ok,no resync/sync, not disabled), the disk is correctly assigned to this machine and it the server knows it!! so, every cmd that i try on this domain it seems to be freezing, i've tryied running a verify and a fixfdmn, but nothing. the last command i tryied was nvbmtpg and here is what i got:

# ./nvbmtpg dsk7a -l
Bad AdvFS magic number on /dev/rdisk/dsk7a.

any clue, hints or something?

Thank you very much.

Regards.

Flávio.

[Mark Poeschl_2]

You are apparently using LSM (voldisk, volprint, plexes), so I'm not sure I would expect 'nvbmtpg dsk7a -l' to work. What does 'nvbmtpg -l' look like? What version/patch level of Tru64 are you using?

[Flávio Gondim de Morais]

Hi!

here:
# ./nvbmtpg domain_oradb04 -l
open_vol: open for volume "/dev/vol/oracledg/oradbvol04" failed: Device busy

# uname -a
OSF1 orasvr01 V5.1 1885 alpha

[Flávio Gondim de Morais]

just to remind you, like i said, it is not mounted and the disk is not assigned for multiple machines!

some more information:

# volprint -ht oradbvol04
Disk group: oracledg

V NAME USETYPE KSTATE STATE LENGTH READPOL PREFPLEX
PL NAME VOLUME KSTATE STATE LENGTH LAYOUT NCOL/WID MODE
SD NAME PLEX DISK DISKOFFS LENGTH [COL/]OFF DEVICE MODE

v oradbvol04 fsgen ENABLED ACTIVE 23566336 SELECT -
pl oradbvol04-01 oradbvol04 ENABLED ACTIVE 23566336 CONCAT - RW
sd dsk7-01 oradbvol04-01 dsk7 0 23566336 0 dsk7 ENA

[Hein van den Heuvel]

Good information, bad situation.
You probably lost that dsk7, where the volume lives.

Tru64 Build 1885, that is aka V5.1A
It _might_ work better in V5.1B or with max patches applied, but I doubt it will make a difference.

In desperation I would also try:

dd if=/dev/vol/oracledg/roradbvol04 of=/dev/nl bs8192 count=1000

It might give a little more insight.

fwiw,
Hein.

[Flávio Gondim de Morais]

hi thx in replying, i've already did a dd before to see if i had any contents on it.. and dd seems to returns me something!! a lot of chars but its something!! any more possibilities before restoring a backup?

[jim owens_1]

You should have the latest 5.1A patch kit 6 aka BL24 and all ERPs installed. The ERPs include updates for LSM and fixfdmn.

If the advfsd or smsd daemons are enabled on your system, kill them. They repeatedly try to access the unmounted domain and cause problems.

Then run "fixfdmn -v 2 domain_oradb04" and tell us what it says.

[Flávio Gondim de Morais]

hi!

well, i've tryied before.. but no answer... its freezing.. so i've killed only smsd the advfsd was unable to kill and i tryied again, but after 2 hrs still freezing!!

# /sbin/advfs/fixfdmn -v 2 domain_oradb04
fixfdmn: No '-d' directory specified, using current working directory to
save the fixfdmn log and undo files.
fixfdmn: Gathering volume information.

fixfdmn: Do you want to abort? (y/n) y

fixfdmn is not able to continue, no changes made to domain, exiting.

[Flávio Gondim de Morais]

# cat fixfdmn.domain_oradb04.log
11:59:55 | INFO | Gathering volume information.
14:16:59 | ERROR | fixfdmn is not able to continue, no changes made to domain, exiting.

[jim owens_1]

If advfsd can not be killed then it must be in the kernel and has that domain locked. While the domain is busy you can not do commands like fixfdmn which require the device to be available. Which means you won't be able to recreate the domain from a backup either in this state.

You must prevent those daemons from starting at boot time. One of the easiest ways is:

# cd /sbin/rc3.d
# mv S53advfsd XS53advfsd
# mv S99smsd XS99smsd

Shutdown all your applications. Then use

# reboot -dq

To force it to take a dump and reboot - note that if the kernel is hung, shutdown may also hang.

When the system comes back up, you chould be able to run fixfdmn. And the dump will be useful if you call HP service.