Lost cluster lock disk

[MQ'ski]

Hi.
I have to nodes in cluster. LVM, Service Guard and v01 to vg08 group in cluster (Jamajka - JBOD). One of disk in vg01 was owner of cluster lock but failed.
It was replaced but I think it is not enaugh? In syslog I receive every hour that cluster lock is missing and it may cause problems until fixed. Should I stop the cluster and recreate vgcfgrestore vg01???

[Duncan Edmonstone]

...it depends when the last vgcfgbackup was done, and whether it was after the last cmapplyconf...

See Melvyns last post in this thread for more info:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xbecb91ccb36bd611abdb0090277a778c,00.html

HTH

Duncan

[Jeff Schussele]

Hi Marcin,

Yes you need to vgcfgrestore the VG - here's an example command

vgcfgrestore /dev/vg_lock /dev/dsk/c2t3d0

Of course your VG name & device may differ.

Then wait an hour or two & check your syslog - messages should go away.

No - you don't have to take the cluster down.

See the following if you need further info

http://docs.hp.com/hpux/onlinedocs/B3936-90053/B3936-90053.html

HTH,
Jeff

[Duncan Edmonstone]

The following is from ITRC Doc UXSGKBAN00000022:

When the cluster-lock disk is replaced, or the clusterlock ID is
missing on a disk, the cluster lock structure must be recreated to allow
MC/ServiceGuard to use it.

If a vgcfgbackup was performed after the clusterlock disk was
configured into the cluster, a simple vgcfgrestore command of this form
would suffice:

# vgcfgrestore -n vg07 /dev/rdsk/c4t5d3


If the vgcfgbackup was not performed after the cmapplyconf, another
cmapplyconf must be performed with the cluster halted to reestablish
the disk listed in the cluster ASCII file as the clusterlock disk. The
command would be in the form:

# cmapplyconf -C cluster_ASCII -P pkg1/pkg1.conf ...

The ServiceGuard cluster checks the clusterlock disk every hour. If
successful, the warnings that the clusterlock disk is missing should
end.


HTH

Duncan

[PIYUSH D. PATEL]

Hi,

In what way did u replace the disk....same capacity. Did it create the same device file ??

You can halt the cluster. Then do

vgcfgrestore /dev/vg01 /dev/rdsk/c*t*d*

vgchange -a y /dev/vg01

then apply

cmapplyconf -k -v -C /etc/cmcluster/cmclconfig.ascii


Piyush

[Sanjay_6]

Hi Marcin,

Did you recreate the cluster lock info on the disk once it was replaced,

http://support2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000058849413

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90053/B3936-90053_top.html&con=/hpux/onlinedocs/B3936-90053/00/00/56-con.html&toc=/hpux/onlinedocs/B3936-90053/00/00/56-toc.html&searchterms=replace%20lock%20disk&queryid=20020523-084232

If this doesn't solve your problem, you may have a defective disk, try this,

http://support2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000058826927

Hope this helps.

Regds

[MQ'ski]

thanks a lot... I will try...

[melvyn burnard]

If you have the valid vgcfgconfig file then this is enough to simply restore it.
If not, thene either re-apply the binary which will require the cluster to be down, or you could log a call with your local HP RC, and request the unsupported utility cminitlock.

Hope you get it fixed.

[MQ'ski]

I think I can restore it but for claryfication
# vgcfgrestore /dev/vg_lock /dev/dsk/c0t8d0
is vg_lock a vgname which should be restored?
M

[melvyn burnard]

vg_lockname should be replaced by the name of your cluster lock vg.
If it is vg01, then use vg01. If it is vglock, use vglock.
Get the idea?

[MQ'ski]

Yes I got the idea now.
Finaly... I understand that if backup is done only vgcfgrestore /dev/vg_lock /dev/dsk/phis_addrs should be done?
And I didn't find if cluster must be stopped.
I suppose not to stop?

[melvyn burnard]

Well the vgcfgrestore is only useful IF a vgcfgbackup was done AFTER the cluster binary was created.
If so, then yes the vgcfgrestore should be all you need.
If not, then you will need to either:
halt the cluster, activate the cl vg on one node, cmapplyconf, deactivate the vg, cmruncl and then monitor the system logs to ensure it worked,
or:
Place a call with your local HP RC (assuming you have support with them) and request the cminitlock utility. Note this is an UNSUPPORTED utility, but it does work. I have no knowledge of it casuing any problems.

HTH