cluster lock disk missing

[Frank de Vries]

We have a repeatingly message in syslog.
However the disk it is referring too is in /etc/lvmtab and in ioscan.
So for me the disk is not realy missing.

Suppose it went missing in the cluster config ?
When I check with cmquery -C cluster.ascii -c cluster -n node1 -v , I get a ascii file with no specification for LOCK_DISK , is that normal ?

More importantly how to fix this.
Our cluster is hpux 11.11 + MC/SG A.11.15.00
two nodes and two packages.

Thx,

[melvyn burnard · Accepted solution]

was anything done to the origional cluster lock disk, like replacing it?
It is essentially saying that the disk referrred to in the configuration does NOT have the bit set marking it as a cluster lock disk.
to fix this, halt the cluster, vgchange -c n the cluster lock vg, vgchange -a y the cluster lock vg, then use cmapplyconf to write the bit to the disk, vgchange -a n again, and then restart the cluster, that should fix it.
Also, you are running a version of Serviceguard that is out of support.

[likid0]

With that version of SG you need to stop SG, apply the config and start SG again

[Frank de Vries]

There was no disk replacement as such.
There was a move (as is) of the San attached storage cabinet from one room to another.
Can't see how that is a change, but there you go.

I told customer it is out or support, but they do not want to do anything about it, my brief is clear: only keeping the lights on, i.e. only firefighing when there is a fire.
And sometimes that sucks.

After the vgchange, do I need to generate a new ascii file with cmquerycl or just use the existing one that is lying around with cmapplyconf.
thx

[Vivek_Pendse]

No need to re-create the ascii file again with cmquerycl. You can make changes in existing ascii files & can reapply it with cmcheckconf & cmapplyconf.
If, the configuration is correct in ascii file, then it is just the job of 15 to 20 mins only.
You need to disable the cluster aware of cluster lock disk & just activate on any node.

Thanks,
Vivek

[Bob_Vance]

Do you have the ASCII file that was used to create the cluster?
What does it have for
CLUSTER_LOCK_VG
& CLUSTER_LOCK_PV
?
What does a
cmviewcl -v
show for the lock information?

Does the PV match what you're getting in the syslog?
Does a diskinfo on the raw device name work?

Is the VG containing the lock disk(s) active?
(that's not a requirement for SG -- you can have a separate VG for the lock disks and have it inactive.
)
If not, try to activate it:
vgchange -a e vgname
.
Does vgdisplay -v on the VG show the correct PV(s).

(if the VG was inactive, then, when done, de-activate it again with
vgchange -a n
)


bv