vgcfgrestore after replacing failed hard disk

[tempsample]

Hi,

I have noted few difference in steps after repplacing HDD.

step 1:

vgcfgrestore

lvcreate (mirror)

step 2:

partition the disk (idisk,efi_fs,vgextend,lvcreate)

is that difference based on hardware model ?

[Torsten.]

Old PA-RISC servers are using the complete disk as part of vg00, newer Integrity servers need partitioned disks and use the second partition only for vg00.

So on a Integrity server you use idisk for partitioning, populate the EFI area to make it bootable and (on both systems) you use vgcfgrestore to restore the VG configuration (this includes mirroring), so the only step to do is a vgsync.

see

When_Good_Disks_Go_Bad_WP
http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c01911837/c01911837.pdf

Systems with SAS or Smart array controller may need different/additional steps, take care!

[tempsample]

Hi Torsten,

thanks for the reply.

From your reply I understand that for

Old PA RISC servers,below process should be followed.

1.change the failed disk

2.pvcreate -B

3.vgcfgbackup -n vg00 [new disk path]

4.vgsync vg00

for Itanium

1.change the failed disk.

2.idisk

3.efi_fs_init

4.pvcreate -B

5.mkboot -e -l

6.efi_cp-d

7.vgcfgrestore

8.vgsync vg00

am i understanding correct ?

[Torsten.]

more or less, but please follow the steps from the guide i linked above!

Do you have a failed disk right now? What server model? What OS version?

[tempsample]

Hi Torsten,

i am Just asking for information.

[Torsten.]

As said, the LVM related steps are in "When_Good_Disks_Go_Bad_WP".

SAS and smart array related steps (sasmgr, scsimgr related commands, e.g. io_redirect_dsf, etc) are in the appropriate hardware manuals for these controller cards.

[Torsten.]

Example:

https://h20565.www2.hp.com/portal/site/hpsc/template.PAGE/public/psi/mostViewedDisplay/?sp4ts.oid=4201434&spf_p.tpst=psiContentDisplay&spf_p.prp_psiContentDisplay=wsrp-navigationalState%3DdocId%253Dmmr_kc-0108998%257CdocLocale%253Den_US&javax.portlet.begCacheTok=com.vignette.cachetoken&javax.portlet.endCacheTok=com.vignette.cachetoken

(google found this, please login to view)

Document title is

"HP-UX LVM - SAS Disk drive replacement "

[Matti_Kurkela]

On Itanium hardware (rxNNNN models and Integrity Superdomes), the system disk needs to be partitioned with the idisk command, and the PV of the system disk will be located in the 2nd partition (the 1st partition will be reserved for the EFI bootloader in Itanium hardware).

On PA-RISC hardware (rpNNNN and older models, and 9000 Superdomes), the partitioning is not used, as the PA-RISC bootloader fits within the boot LIF at the very beginning of the disk.

There is another difference based on HP-UX version/patchlevel: if your HP-UX version supports the "pvchange -a N" command, you won't need to use lvreduce/lvextend to remove the mirrors and add them back: you just disable LVM access to the failed disk with "pvchange -a N <device>", replace the physical disk (and if using SAS-based disks, enter the necessary commands to make the new disk take the place of the old one), rebuild the partitioning on Itanium, then you can restore the entire LVM configuration to the replaced PV with vgcfgrestore and reinstall the bootloader with mkboot. Then you can re-enable LVM access with "pvchange -a y <device>", and the LVM should automatically start syncing all the LV mirrors.

On older HP-UX versions where "pvchange -a" is not supported, it can be necessary to lvreduce the mirrors before the disk is replaced (at least if the disk is not already completely dead and showing as NO_HW in ioscan). If the mirrors are lvreduced, then they must always be lvextended back after disk replacement.