cmgetconf warning/error after OS upgrade

[Thierry D]

Hi everyone,

While running the command cmgetconf -c .... on the cluster, I receive the following error.

# cmgetconf -c mycluster mycluster-conf.ascii
The following data must not be used when reapplying cluster configuration.
Please correct the volume group before running cmgetconf again and applying the cluster configuration.
#

From what I could see all is working correctly, I also check in the manual (well the one from Linux because I could not find the one from HPUX) and nothing interesting in it.

When I try to run the command on a package there is no error / warning .

Also the command cmcheckconf -C asciifile returns a positive feedback ...

OS: B.11.31
Model: ia64
SG Version: A.11.19.00

Also from the syslog.log I could see the following ...

Feb 1 15:15:06 nodeA cmclconfd[6414]: Client provided hostname nodeA which does not match resolved name .
Feb 1 15:15:08 nodeA cmclconfd[20437]: LVML:Querying volume group /dev/vgswap for node nodeA
Feb 1 15:15:08 nodeA cmclconfd[20437]: Volume group /dev/vgswap has no cluster attributes
Feb 1 15:15:08 nodeA cmclconfd[20416]: LVML:Querying volume group /dev/vg00 for node nodeA
Feb 1 15:15:08 nodeA cmclconfd[20416]: Volume group /dev/vg00 has no cluster attributes
Feb 1 15:15:08 nodeA cmclconfd[20437]: LVML:Querying volume group /dev/vglogiciel for node nodeA
...


Any idea, thoughts where I should focus ?

Regards,
Thierry

[Steven E. Protter]

Shalom Thierry,

How did you do the upgrade?

What was the update-ux command line?

Did you take a look at the log?

SEP

[Thierry D]

Hi Steven,

I wish I could tell you I did the upgrade but it is another team who does patching and OS upgrade ...

Tho they should have use the following command to do it (could not see anything from the history file ...)

update-ux -s sdux-server:/depot/11.31/Release0909 HPUX11i-DC-OE

Also from the log files at both locations there were no errors nor warnings ... (few files were skipped but these were locales/languages fileset ...)

/var/opt/swm/swm.log -> no errors
/var/adm/sw/swagent.log -> no errors

Also I tried to modify the hosts file, since the first message within syslog indicated a resolution problem but nothing changed (adding the fqdn in addition to the short name).

[John Bigg · Accepted solution]

I would normally expect an error when this message is reported. The message indicates that there is a problem with the LVM configuration, such as a volume group is configured differently on one node than another, so this is what I would focus on. I think 11.20 has improved this and you won't get that message without an error saying what VG is causing trouble.

Another way to diagnose the problem would be to run cmquerycl -v -n -n ... since if there is a VG config problem this command may give you more details.

At the end of the day, the cluster ascii file is probably correct, but may miss off one or more volume groups in the VOLUME_GROUP section at the end of the file. This would only cause a problem is the cluster attributes were incorrectly removed from a cluster vg in which case a package vgchange would then fail when the package started.

[Thierry D]

Hi John,

thanks for the tips, I wish I could install the 11.20 version ...

When you say there is a VG config issue, what kind of problem do you think of ?

I ran cmquerycl to get different information, but to be honest I dont see anything wrong or at least that seems wrong ..

I can sse a list of info regarding LAN from each node and so on, and a list of mount FS and disks available on both nodes ...

Regards,
Thierry

[John Bigg]

I was meaning check the vgdisplay output to ensure the same pvs are in the vgs on each node and that they can be activated on each node.

If there were problems, then I would expect cmquerycl to report problems too. If cmgetconf shows this message but cmquerycl does not report VG errors then I would be tempted to ignore the message since you indicate the cluster ascii file appears correct.

[Thierry D]

Hi John,

ok I see what you mean know and I think i did found what is wrong but I am not sure 100% about it.

In the cluster ascii config file there are a total of 20 defined VGs for applications/packages.
In the different packages config file's there are a total of 26 defined VGs .

Also I could see there are 10 or so VGs which are either included/missing from one another
So I suppose this is the problem you were referring to ....

The only thing I would like to understand now is :
Why are the Application/Package VGs declared within the cluster config and not just only within the package config file ??

I could see the following from the cluster config file but still, doesn't it mean it should be done on the package side ??

# List of cluster aware LVM Volume Groups. These volume groups will
# be used by package applications via the vgchange -a e command.
# Neither CVM or VxVM Disk Groups should be used here.


Regards,
Thierry