"Pete's" <empete2000@yahoo.com> schrieb im Newsbeitrag
news:6724a51f.0401220607.36d4155e@posting.google.c om...
> Vital Info:
> 2 Node cluster, p660-6h1's running AIX 4.3.3 ml09 with HACMP 4.4.1.
> It has shared SCSI disks using 2 2104-DU3's with 10 total Physical
> Volumes. [snip]
> Currently 'Node 1'(which is the primary node) has the cluster
> resources and is running fine.
>
> 'Node 2'(which is the secondary) is in the cluster, but, has lost all
> information regarding the shared disk, i.e. PVID's and Volume Group
> information is gone.[some snipped]

Hallo Pete,
sounds like you got your punishment already ;-). Here's the good news: the
solution might be less painful because we are talking AIX here
Disclaimer: I am writing from memory and I assume that the HACMP
configuration itself was working properly before and did not change through
your diag adventure...
>
> On Node 2
> 1. Assign PVID's to the shared disk since they no longer have any
> assigned.
Nope. Don't assign manually, just let configuration manager read what's
there (node1 still sees the old/original PIVDs doesn' it? Then they can be
read by node2 again, too.)
> 2. Shutdown the cluster software.
> On Node 1
> 3. Shutdown the cluster software
> 4. exportvg vg01
Nope!!!!!! just varyoffvg vg01
> On Node 2
# cfgmgr -vlscsi0 ... (or similar, as you like. Use -v verbose option to
watch hardware detection)
> 5. importvg vg01
# importvg vg01 <one_of_the_vg01_member_disks_rediscovered_by_cfgm gr>
Assign Major Number (same as on node1) during import if fs of vg is exported
via NFS. (I would in any case).
> 6. varyoffvg vg01
> On Node 1
> 7. importvg vg01
Just varyonvg vg01 after varyoffvg
> 8. varyoffvg vg01
Why? Did not understand why you want to do this step. Do you want to run
HACMP from node2? Should you run node1 as primary and node2 as backup (e.g.
not concurrent cluster) I would start HACMP on node1 and then do a manual
takeover to node2 and back to see whether everything's working. (As I said
above: it should if your HACMP config was fine before and did not change).
Synchronize cluster resources bevore starting takeover tests.
> 9. Start the cluster software.
> On Node 2
> 10. Start the cluster software.
>
> I tried this before except that the PV's did not have id's assigned
> and it failed(I'm assuming it failed because of no pvid's).
Yes. Actually that's the starting point. You need to see disks from node2.
Write again if cfgmgr cannot access the disks because of some lvm locks or
scsi reserve bits.

HTH,
Andreas