VGs in cluster

[Gabriel_23]

I am currently trying to shutdown the cluster and mount volume groups that initially belong to the cluster. Do I have to manually vgchange each vg in the cluster or can I edit the cluster config file to take the VGs out of the cluster automatically?
Please help if possible.

[Jeff Schussele]

Hi Gabriel,

If you mean you want the VGs up & the pkg(s) down, then after halting the packages just:

vgchange -a e /dev/vg_name
and
mount /dev/vg_name/lv_name /mnt_point

when you're done
umount /mnt_point
and
vgchange -a n /dev/vg_name

To make the filesystems available.
And no you cannot remove the VGs from the cluster while it's up.

HTH,
Jeff

[Sundar_7 · Accepted solution]

The cluster-aware vgs cannot be mounted when the cluster is down. cmlvmd daemon should be running in the system for you to be able to activate the volume groups that belong to the clustetr.

1) Shutdown the package (not the cluster !)using the volume group.

2) execute vgchange -c n /dev/vgname for all the VGs that you want to be able to mount when the cluster is down

3) Now you can halt the cluster and mount the volume groups

Again, for the package to be able to mount the volume groups, you need to execute vgchange -c y /dev/vgname when the cluster daemons are running in the system.

If you can tell us what you are trying to achive, our folks could possibly suggest you the best possible way to achive the same :-)

[Gabriel_23]

right now there are VGs (on the array)in the cluster that are in cluster mode. However, I want to take Service Guard out the equation and mount the same VGs from the array. I want to know if the only way to do this is to shutdown the cluster, manually issue vgchange -c n on each VG designated in cluster and then mount manually. Hope this makes sense.

[Sundar_7]

Yes, you need to execute vgchange -c n on all the VGs involved before you can activate them out of the cluster.

[Gabriel_23]

Ok. Thanks to everyone for taking the time out to help me out...

[Jeff Schussele]

Hi Sundar,

Read my vgchange command line again.
Note is does *not* use -a y - it uses -a e
In that case one does not need to use -c n
Why run an extra command just to use -a y?
I sure don't like to type that much.
The *only* requirement to use -a e is that the cluster daemon must be running.

Later,
Jeff

[Sundar_7]

Hi Jeff,

I fully understand it was vgchange -a e and not vgchange -a y.

But if I have understood Gabriel's question, he wanted to activate the volume groups when the **CLUSTER** is down and not just the **PACKAGE**.

***I want to take Service Guard out the equation and mount the same VGs from the array***

*** am currently trying to shutdown the cluster***

Now we are clear that clsuter daemons are not going to be running, so you cannot use vgchange -a e ...Am I missing something here ?

-- Sundar.

[Jeff Schussele]

Sundar,

Understood.
I guess what's confusing is the last sentence
"Do I have to manually vgchange each vg in the cluster or can I edit the cluster config file to take the VGs out of the cluster automatically?"
To me that can be read A) The cluster is up & running & I want to remove the VGs from it. or B) I want to start the cluster w/o mounting the VGs. Either way implies the cluster is up.

Furthermore in his later post he asks
"I want to know if the only way to do this is to shutdown the cluster, manually issue vgchange -c n on each VG designated in cluster and then mount manually."
And the simple answer is no - stopping the cluster, using -c n & -a y is not the only way. You can stop the pkg(s) & use -a e prior to mounting. Here, I'm used to having multiple pkgs on servers, so I view halting the cluster as an absolute last resort. Obviously this user has a single pkg cluster.

I think we were just looking at it from different angles. I can see your point, I would hope you can see mine.

Rgds,
Jeff

[Sundar_7]

Sure, I do see your point Jeff and I fully understand -c n and -a y is not the only option if the cluster daemons are running. Have a good one :-)

[Gabriel_23]

Sorry for the confusion guys. I should have asked another way initially. Sundar understood it the way I meant it...but I understand why I confused you Jeff. Not sure how often I can give points, but I have learned a lot from you two just talking and will give you more.
Thanks again,

Gabe