Identify internal disk.

[Ryan Phlip]

Hi,

I have a failed internal disk on my DS20 Tru64 Alpha Server. I want to physically identify this disk on the server. This is to make sure I am removing the correct disk from the server.

This is a cluster system.

Many Thanks,
Ryan Philip

[Vladimir Fabecic]

Please post output of:
# hwmgr -show scsi | grep dsk42
# hwmgr -view device | grep dsk42

[Venkatesh BL]

I think you can use:
# hwmgr -flash light -dsf /dev/disk/dsk42c

This will make the disklight blink.

[Vladimir Fabecic]

Not all disk can flash lights.
You can also try this and see on which disk activity led flashes:
# dd if=/dev/rdisk/dsk24c of=/dev/zero

[Ryan Phlip]

Hi,

I had read the administration guide and tried the flash and dd commands but there are no lights on the internal disk, so it becomes difficult. There are 4 9GB disk and 2 18GB disk. By using the Hwmgr flash command I cannot identify because the lights are not there.

Can someone explian to me the how to Identify the bus 0. Is there a way to do that5?

Regards,
Ryan Philip

[Rob Leadbeater]

Hi,

Can you post the output of the commands previously requested by Vladimir ?

dsk42 could be a single disk, or it might be part of a hardware RAID disk.... With the information provided so far, it's impossible to know...

Cheers,

Rob

[Kapil Jha]

If flash light is not there....only esacpe from hardware engineer is to use dd command and put hands on disk and u can feel which disk is working.
There would be 2 internal disks so chance are bright ;)

BR,
Kapil+

[Vladimir Fabecic]

Can someone explian to me the how to Identify the bus 0. Is there a way to do that5?

In SRM it should be bus "a".
For example DKA100 (in "console mode") should be BUS 0 ID 1 LUN 0. So, for example, if you are looking for disk in BUS 0 and SCSI ID 1, it should be DKA100.
As far as I know bus 0 should be "built in" controler bus.
In DS20 bus "b" or BUS 1 is usually a PCI SCSI controller.
If you need more details, first post what I asked for.

[Ryan Phlip]

Hi,

Firstly dsk42c is a single disk. It is not in any raid. The disk is spoiled so it does not respond. the output of the dd command is

dd if=/dev/disk/dsk42c of=/dev/zero
/dev/disk/dsk42c: Device busy

The other output is attached.

Regards,
Ryan Philip

[Vladimir Fabecic]

Command should be:
dd if=/dev/rdisk/dsk42c of=/dev/zero
not
dd if=/dev/disk/dsk42c of=/dev/zero
Try again.
So it is disk DKA100 in SRM.
And view in your disk shelf:
----dkao-----
----DKA100--- FAILED DISK
----dka200---
----dka300---
----dka400---
----dka500---
-------------

[Ryan Phlip]

Hi,

Vladimir that command works. Thanks.

How do I find out given the h/w path if the disk is in used or not.

Example : I know the disk H/W like /dev/rdsk/dsk42c how do i find out if its in use or already configured.

I have been told that a long time back spare disks have been lying in the server which are not in use. So I could use one of those.

Kind Regards,
Ryan Philip

[Rob Leadbeater]

Hi Ryan,

If you're using AdvFS, if the disk is being used, then it will be referenced somewhere under /etc/fdmns. So this will give some output if it is being used:

# ls -lR /etc/fdmns | grep dsk42

If you're using UFS then you'll see it in /etc/fstab.

You should also sanity check that the disk isn't being used as a swap disk with

# swapon -s

Cheers,

Rob

[Vladimir Fabecic]

Rob answered your question.
So post output of:
# ls -lR /etc/fdmns
# cat /etc/fstab
# swapon -s
if you want us to check if there are unused disks.

[Venkatesh BL]

>dd if=/dev/disk/dsk42c of=/dev/zero
>/dev/disk/dsk42c: Device busy

You should not see this error unless the block device is currently mounted. Are you sure that the AdvFS domain representing this device is unmounted now?

[Ryan Phlip]

Hi,

Firstly, nothing has been done on the system. I mean I identified the disk is spoiled its in a domain with one fileset mounted but you cannot do anything as the disk is totally gone. It not mirrored.

I am sending the outout as requested.

Regards,
Ryan

[Vladimir Fabecic]

dsk107 , dsk108 and dsk109 do look free for use.
Check again.
Send output of:
# disklabel -r dsk107
# disklabel -r dsk108
# disklabel -r dsk109

[Ryan Phlip]

Hi,

It does not look free to me. I am sending you the output.

Yes there is no entry in /etc/fdmns/

Thanks,
Ryan

[Ryan Phlip]

Hi,

It does not look free to me. I am sending you the output.

Yes there is no entry in /etc/fdmns/

Thanks,
Ryan

[Vladimir Fabecic]

Yes, these disks are labeled and have advfs created.
But do not see entries for them in /etc/fdmns.
Manually create entries for advfs domains, scan for file sets and try to mount them.
Do you know how to do it?

[Venkatesh BL]

1) If you can find a _really_ free disk on the system, use it.

2) It is possible that there was an AdvFS domain created earlier using these disks (dsk107, 108 etc) and that they forgot to 'rmfdmn' when the domain was no longer required. Or, they reinstalled the OS which left these disks to remain marked as 'AdvFS'

[Ryan Phlip]

Hi,

Firstly, Sorry for the delay in reply. I work accross platforms and had work on HP-UX boxes.

Hi Vladimir

I am not too strong in Tru64 so do not know how to create entries in advfs domain or to be more specific I do not understand when you say manually create. After that you said to scan which i understand is

#hwmgr scan scsi

Please guide me.

Hi Venkatesh BL,

Yes I have the same question running in my mind. Ans to ur second point, It can be a possibility.
Now I have done some study and want to be sure if these disk are been used for some purpose or are previous Failed disk.

I need to know a work around for this.

Please can you guide me and in the mean while I shall try and find more info and fwd to you all.

Many Thanks Guys,

Ryan Philip

[Vladimir Fabecic]

I also think it is possible that there was an AdvFS domain created earlier using these disks.
The procedure shoud be:
# mkdir /etc/fdmns/tempdom1
# ln -s /dev/disk/dsk107c /etc/fdmns/tempdom1/dsk107c (create simbolic link to tell which device should this file domain contain)
# showfsets tempdom1 (to see which file sets were created in this domain, it can be more than one file sets, for this example lets say that it contains file set called "fs1")
# mount -t advfs tempdom1#fs1 /mount_point

[Ryan Phlip]

Hi,

I am facing another problem. I cannot umount the mount point on that domain. It does not respond. I cannot perform any task on that disk. I am sending the errors.

Regards,
Ryan Philip

[Venkatesh BL]

>root on shenzi> df -k | grep shenzi_collect
shenzi_collect#collect 17782536 472912 10501000 5% /shenzi_collect

Here,
shenzi_collect: AdvFS domain
collect: AdvFS fileset belonging to the domain
/shenzi_collect": Mount point directory for the fileset.

I see that you've passed wrong arguments to many advfs commands.

In any case, if there is an IO error, the kernel would not be able to read and display details from the disk.

[Vladimir Fabecic · Accepted solution]

Try
# cfsmgr -U -d shenzi_collect
and post error message