replacing disks

[Roberto Gallis]

Hi all,
I have to replace two old disks in mirror with two newest and bigger disks. These disks are external connected to two D class server in cluster. These disks belong to a volume group shared by the 2 node cluster.
So I think I must destroy the old vg and recreate it vith the new 2 disks.
But, do I need to delete end reinstall special files with insf when disks are replaced?
Do I need to reboot the 2 nodes?

Regards

Roberto

[Printaporn_1]

HI,

if you can effort the downtime better reboot both node to get rid of weird error and device file will automatic create for you.

[Vincenzo Restuccia]

# Shutdown your system and replace the disk that is failing.

Boot into single user mode and override the quorum, as:

# ISL> hpux -is -lq /stand/vmunix

# vgcfgrestore -n /dev/vg00 /dev/rdsk/cXtYdZ

# mkboot /dev/rdsk/cXtYdZ

# mkboot -a "hpux -lq /stand/vmunix" /dev/rdsk/CXtYdZ

# lvlnboot -Rv


# vgchange -a y /dev/vg00

# vgsync /dev/vg00

# shutdown -r y 0

[Pedro Sousa]

Roberto,
is this VG in shared, or exclusive mode?
If it is in shared mode, then you'll need to de-activate it first, then activate it with "vgchange -a e" (this way you wont need to stop the cluster, just the package associated with the vg). After this, on both cases:
- connect the new disks to both systems (if slots are available)
- "ioscan -fnC disk"
- For the new HW addresses without device files, you'll need to do "insf -H " (do this on both systems)
- on the system you have the VG activated do: "pvcreate /dev/rdsk/cxtxdx" for each new disk
- "vgextend vg /dev/dsk/cxtxdx /dev/dsk/cxtxdx" (will extend the VG to the new disks)
- "pvmove /dev/dsk/ /dev/dsk/
- "pvmove /dev/dsk/ /dev/dsk/"

- "vgreduce vg /dev/dsk/cxtxdx /dev/dsk/cxtxdx"
- if it succceeds, then you'll need to "vgexport -m /tmp/vgxx.map -p -v "
- copy the file to the other system and, on that system, do "vgexport vgxx"
- "mkdir /dev/vgxx"
- "mknod /dev/vgxx/group c 64 0x000001" (this hex number has to be the equal to the original system, be carefull with this.
- vgimport -m /tmp/vgxx.map vgxx /dev/dsk/ /dev/dsk/
- try to activate it on this node.

Hope this helps. maybe it seems a little bit hard, or confusing.

[Roberto Gallis]

For Vincenzo:
I do not need to replace a failed disk on vg00 (boot disk). I need to replace 2 old disk of 2GB with new disks of 18GB.
These disks are external conneted by scsi and shared on two cluster nodes.
The vg is a part of a package of the cluster. I know I have to stop package and rebuild the vg and reshare again.
But I do not know if changing disks type can make problem with original special files.
I the reboot reinstalls special files... But I am concerned if boot is necessary or not

Regards
Roberto

[Pedro Sousa]

Hi again.
You do not need to reboot your system to install new special files! You only need to use that command called "insf".
So, if your vg is in shared mode you're maybe using MC/SG OPS?
These disks will be placed on a rack?

[Roberto Gallis]

Hi Pedro,
yes, there's SG... I know all procedures to do for recreating vg and package.
My question was:
If disk is /dev/dsk/c3t2d0 (for example) and the new disk has the same path, but the disk is bigger, do I need to reinstall special file even if the path is the same??
That's all.

Regards

Roberto

[Pedro Sousa]

OK!
My answer to that is NO. Once the device file is created, you do not need to recreate it again.
But don't forget 'pvcreate'.
good luck.

[Pedro Sousa · Accepted solution]

Roberto, final answer:
I made the following experience on a K server with 10.20 some minutes ago:
- added a disk with 4Gb
- created the device file
- "diskinfo /dev/rdsk/cxtxdx" - OK
- removed it
- ioscan showed NO_HW - OK
- added a disk with 2GB to smae slot
- ioscan showed CLAIMED - OK
- "diskinfo /dev/rdsk/cxtxdx" - HANGED!!!
so, I cannot guarantee, maybe that disk has some problem, but maybe you should re-create the files.
Just a last thought.

[Roberto Gallis]

thanks Pedro for the test!!!

Best Regards

Roberto