Sparetool & HACMP problems

[mateoosh]

Hello again

We have a "small" problem with SSA Sparetool and HACMP cluster switching. I'm a beginner in HACMP, so any help is very pleased. So:

1. One of our ssa disk failed , sparetool correctly detected it, changed with spare disk, maked migratepv, sync and all was OK.
2. I replaced failed disk with a new one, then used script replacing_failed_disk (included in SSA Sparetool) and all was sucessfully completed (recreated sparedisk VG, and new disk replaced with sparedisk in VG)
3. Now, during cluster switching (takeover) we become stale partitions (VG which contains previously replaced disk is one of HACMP resources). The question is: maybe HACMP recognizes disk, which belongs to HACMP Resource (VG) by serial number ? Now, when the disk is new, the serial is another and HACMP thinks, that disk is failed ?

Anyway, thanks for any help ...

BR

Mateusz K. / Poland

 

[gower]

Hi Mateoosh,

Basically you need to verify your cluster again ( clverify ) to see if the missing disk is listed and to make sure that the resource group knows about your new disk. I don’t know about “SpareTool” but you should make sure that on both nodes ( or more depending on your cluster ) the PVID’s can be seen as this is what the resource group will be using ( depending on the level of HACMP/HAES ).

Also make sure you have sync’ed cluster resources once you know the operating system can see the correct disks on both nodes.

Hope this helps.

 

[afelton]

All servers in an HACMP resource group need up to date info about VG's and LV's that are in the resource group, and this must be kept in sync.
I assume the VG is OK on one server, but has stale PP's on the other.
With HACMP stopped try exporting/importing the VG on the server showing stale PP's. (don't forget to set auto varyon to "no" after importing the VG).
Check there are no stale PP's, and vary off VG.
Start HACMP and do some failover tests, hopefully HACMP will
work correctly.

 

[mateoosh]

Hi friends !

I maded some deeper investigation and node B doesn't know about the new disk which was replaced at node A. Additionally, there's a new disk detected at node B, which not belongs to any group. So I must remove hdisk/pdisk definitions, recreate it and then export import VG. BTW.smitty importvg should be enough to correctly import VG definions on second cluster node ? And major number should be the same on both nodes sharing the vg, right ? Sorry for dumm questions, but as i sayed before, Im a beginner at HACMP.

BR

Mateusz K. / Poland

 

[afelton]

the major numbers do need to be the same for the cluster verification to work.
to import the VG do:
importvg -V majornumber -y VGname hdisk??

this will import and varyon VG
to set VG to not auto varyon do:
chvg -a n VGname

check VG is OK on node B and vary off VG.
verify the cluster.
correct any errors reported during verification.
if no errors start HACMP and test failover.