MC/Service Guard Problem in HPUX 11.00

[roobala]

Hi

HPUX 11.00
L-1000, Server.

Cluster is started and running the packages,
while starting also no errors.
But in syslog we are getting messages continously like this.(Refer:attachment)

What could be the cause for this, Can someone help us

Thanks in Advance,
Roobala

[melvyn burnard · Accepted solution]

The message you are referring to is, I assume, the one about the missing cluster lock disc.
If so, it seems you have configured a cluster with a clusterlock disc, but have not initialized the cluster lock disc.
Was the VG that contains this PV activated on the node on which you ran the cmapplyconf command?
If not then MC/SG was unable to write the cluster lock info to the PV.
You will need to halt the cluster, vgchange -c n , vgchange -a y , cmapplyconf, vgchange -a n , cmruncl

and that should fix it.

[Steffi Jones_1]

The warning message about "Cluster lock on disk /dev/dsk/... is
missing!" could be result of corrupted SG binary OR if the cluster lock
disk was replaced or moved to a different disk.
Make sure cluster lock VG and cluster lock PV is defined in the
ascii cluster configuration file, then to fix:

1. Halt the packages and cluster.

2. Activate the cluster lock VG on the configuration node ONLY, so the
lock disk can be initialized, for example:

vgchange -a y /dev/vglock

2b. If needed, do the following on the VG:

vgchange -c n /dev/vglock

3. Generate the cluster binary configuration file & distribute
to the nodes with cmapplyconf command, for example:

cmapplyconf -C /etc/cmcluster/cmclconf.ascii
-P /etc/cmcluster/pkg1/pkg1conf.ascii

4. Deactivate cluster lock VG.
vgchange -a n /dev/vglock

5. Restart the cluster.

[Emil Velez]

I agree with the above answers. You are runnig without a cluster lock disk. The cluster lock disk is necessary to prevent "split brain syndrome" If you have a network failure where the systems cannot communicate the lack of a cluster lock disk will prevent your cluster from reforming (in a 2 node cluster) and it will panic. 2 Node cluster need a cluster lock disk. The cluster lock disk needs to be a disk that a package uses for data with the data mirrored since you want to periodically do I/O to it and detect whether it is failed or not. The cluster lock disk can be in a disk array (one of the supported ones). This will provide a cluster lock disk that does not fail as far as the operating system is concerned.

Good luck and I hope these thoughts help

[James R. Ferguson]

Hi:

Another thread, just today (!) appears to have the same issue:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x9e0ea24d9abcd4118fef0090279cd0f9,00.html

See KB document #KBRC00001982 and/or #W3618850 for corrective action.

...JRF...

[Stephen Doud]

Please refer to this article in the ITRC knowledge database:

UXSGKBAN00000282

If the cluster-lock disk listed in the output report of cmscancl is a legitimate clustered disk, and you need to keep the cluster up, you can eliminate the messages by using an unsupported command called "cminitlock".
It can be downloaded along with a READMe from the response center.