tru64 goes down to halt

[Bulent Kolay]

I have a server alpha es40. its os is tru64.
the server has gone down to p00>>> everyday for last 3 days.
the server has been working without any problem for 4 years. We didn't install any addition software recently.
when it becomes crash, I boot it . it works for a while. after that it goes down.
I investigated the messages on it.

An AdvFS domain panic has occured due to either a metadata write error or an internal inconsistency. This domain is being rendered in accessible
Vmunix: syncing disks AdvFS I/O error
Vmunix: Volume: /dev/rz8g
Vmunix: Tag: 0xfffffff7.0000
Vmunix: Page: 435
Vmunix: Block: 8064
Vmunix: Block Count: 16
Vmunix: Type of operation: write
Vmunix: Error: 5
bs_osf_complete: metadata write failed

[Steven Schweda]

sizer -v

> [...] AdvFS I/O error

Hmmm. Do you think it may be a bad disk?

> [...] Volume: /dev/rz8g

Possibly that one?

[Bulent Kolay]

in another time in /var/adm/messages, AdvFS I/O error /dev/rz8a

What am I to do in order to fix it ?
Thanks

[Rob Leadbeater]

Hi,

It sounds very much like you have a failed or failing disk.

Repair/replace the disk.
Restore from your backups as appropriate.

Cheers,

Rob

[Bulent Kolay]

if I save the disk from repair,
How can I repair it ?
Thanks

[Rob Leadbeater]

You could try reassigning bad blocks.

Check out the man page and the help for scu.

Cheers,
Rob

[Bulent Kolay]

is there a tool to try reassigning bad blocks ?

how can I get it ?

thanks

[Rob Leadbeater]

Please refer to my previous post. For example:

# man scu

# scu

scu> help

Cheers,
Rob

[Steven Schweda]

> How can I repair it ?

How old is this disk drive? What would a
replacement cost? How much time would you
like to spend playing with a cheap piece of
obsolete (and failing) hardware? What is
your time worth?

[Bulent Kolay]

I have 3 advfs partitions

disklabel -r rz8

rrz8a
rrz8g
rrz8h
I booted the server on single mode.
when I run "/sbin/scu -f /dev/rrz8a"
I doesn't get any error. scu prompted
scu>
when I run "/sbin/scu -f /dev/rrz8g"
scu>
when I run "/sbin/scu -f /dev/rrz8h"
scu>
is it normal ?
I see /var/adm/messages
Vmunix: syncing disks AdvFS I/O error
Vmunix: Volume: /dev/rz8g

Why didn't the utility able to find any bad blocks?

Thanks

[Rob Leadbeater]

> Why didn't the utility able to find any bad blocks?

Because you didn't tell it to check anything...

scu>
scu> sbtl 1 0 0
scu> verify media

If this flags a block as being faulty, then

scu> reassign lba
scu> verify media start

Repeat as necessary. If you get more than a few bad blocks then replace the drive and restore from your backups.

Cheers,

Rob

[Bulent Kolay]

Hello

I think 1 0 0 of values in sbtl is variable.
How can I learn these values for the server?

Thanks

[Kapil Jha]

scu> show edt

and from there you can get the device number.

After scu you may also consider to run fsck on rootdg as from the error it clearly reflect that you have you have AdvFS curruption.

Do you have alternate mirror disk?

BR,
Kapil+

[Martin Moore]

> After scu you may also consider to run fsck on rootdg as from the error it clearly reflect that you have you have AdvFS curruption.

Um, no. fsck works on UFS, not AdvFS. To check for and fix AdvFS corruption, you need to use fixfdmn or verify (both found in /sbin/advfs).

Martin