replace mirrored boot disk on 11.31 intanium

[Marilyn DuChene]

Having problems replacing my mirrored boot disk.

Here's the procedure I was following:

Replacing /dev/disk/disk3 which is a mirrored boot disk on our rx7640 running 11.31
pvchange -a N /dev/disk/disk3
(replaced the disk)
scsimgr replace_wwid -D /dev/rdisk/disk3
ioscan -m lun (got new lun path - which is /dev/rdisk/disk34)
created /tmp/idf file which contained: 3, EFI 499MB, HPUX 100%, HPSP 400MB
idisk -f /tmp/idf -w /dev/disk/disk34
insf -e -H 1/0/0/3/0.0x6/0x0
io_redirect_dsf -d /dev/disk/disk3 -n /dev/disk/disk34
/dev/disk/disk3 is being used. Cannot perform redirection.

The io_redirect_dsf command is not working. I found the error in a patch (PHCO_37864) and applied this patch; however I am still getting the same error.

Question is what do I do now?

[Steven E. Protter]

Shalom,

http://docs.hp.com/en/5991-1236/When_Good_Disks_Go_Bad.pdf

SEP

[Marilyn DuChene]

I read the when good disks go bad. I was following the instructions on the HP System Administrator's Guide to LVM 11.31. Question is do I forget all about the scsimgr commands & leave all that alone and then just start pvcreating this new disk?

[Tim Nelson]

I have yet to have a disk failure on 11.31 but it would seem that the legacy procedure should work just as well.

If you can, vgreduce the bad device (or follow the procedures to vgcfgrestore with the new dev_file).
and/or
reduce the old_dev from the vg(if you can)
prep the disk,partition and boot files
extend the new_device into the vg
and extend the mirrors to it.

If I had the time I guess I would do as you are doing and find the procedure to redirect the new wwn and dev_file for future use.

[Marilyn DuChene]

OK - Thanks - I guess back to the old way. I can only say the new way with pvchange -a N and trying the scsimgr & io_redirect_dsf is not working at all. I do have a call open with HP - but so far all they come up with is reboot.

[Torsten.]

Something is still accessing the disk - question is, what is it?

I would search the "ps" output if there is any diagnostic disk monitor related process is still running or hang.

The specific monitor process may depend on the chassis where the disk is mounted.

Of course a reboot should solve this too, but it's worth to check.

[Marilyn DuChene]

Yeah - now that I think back you are probably right. It seems maybe setboot was the culprit, but I'm not sure. Here's what I was seeing in syslog: setboot: "/dev/rdisk/disk3" device failed to open. Maybe I should have change the path of the alt disk in setboot to be the same as the primary disk? It's too late. I already went back and lvreduced, vgreduce & recreated everything from scratch.
I still have messy ioscan's with 2 disks at the same hardware path - 1 the old disk3 NO_HW and 1 with the new disk, which is working.

[Torsten.]

The procedure performed by Marilyn is correct.

The procedure for replacing a bad boot disk on 11.31 is obviously different to 11.23.

Link to documentation:

http://docs.hp.com/en/5992-3385/ch04s08.html#v1446838

HP-UX System Administrator's Guide: Logical Volume Management: HP-UX 11i Version 3 > Chapter 4 Troubleshooting LVM

Replacing a Bad Disk

[Marilyn DuChene]

If anyone tries the new procedures for 11.31 - I would run the setboot and path both the primary & alternates to the working disk you have left (before starting the procedure.) That way setboot isn't impeding the process of the io_redirect_dsf command. Also HP was advising to have patches PHCO_37328, PHCO_36555 and PHCO_37864 (which I had - but didn't really help).

[IT Response · Accepted solution]

Attached are my personal notes on how to complete the disk replacement. Hope the give some points. In my experience, you either use "scsimgr replace" or io_redicted_dsf but never both. What you should now is that on sas controller disk, after the io_redirect_dsf the "new" dsf will be left as NO_HW until you reboot the system.