Updating cluster LVM info

[Anthony O'Connor]

We have a 4-node cluster comprising 2 N4000s and an EMC array in each of two sites. The EMCs are SRDF'd.

If a filesystem on one node gets extended and then the package that it is part of gets failed-over to a node in the other site, is there any way to automatically update the LVM config info on the node in the 2nd site to reflect the new filsystem sizes?

This is not serviceguard, it's veritas. Any pointers?

[Sanjay_6 · Accepted solution]

Hi Anthony,

Are you talking about veritas volume manager or vxfs filesystem under LVM. As far as LVM is concerned, if it is only filesystem size increase, there is no need to update the LVM info on other nodes. If you add a new LV, Add a disk or something like that, then you need to update the LVM info on the other nodes.

Hope this helps.

Regds

[James R. Ferguson]

Hi Anthony:

Since the size of a logical volume is deduced at the time it is activated, you should not have to do anything on your failover node.

Regards!

...JRF...

[Rita C Workman]

To add to what the others have already mentioned:

If it is that you added some new disk(s) to a vg in a pkg...remember you did (or should have done) a vgexport with a mapfile; and then did a vgimport on that other box, so your cluster is covered and aware of all the disks in the vg.

Hope this helps,
Rita

[Roger Baptiste]

Hi, you dont need to do anything. The lv info is
picked up automatically on
it''s activation.

-raj

[Tim D Fulford]

Sorry I have to disagree with some of what is said above.

The official line from MCSG course is that after adding/creating new LV's in a SG package you need to do a vgexport on one node & vgimport it onto ALL nodes that the package is available on. However, all you need to do is make the character & block device.

What I have done before now is write a script to periodically compare the filenames in /dev/vg??/. & if any are missing add them with the appropriate majour & minor numbers. Put this script in cron & bing all done.

As far as the package script file (/etc/cmcluster//.sh) you only need to alter this IF it involves mounted filesystems. Then just copy the script across all nodes. You don't need to restart the pkg (I pritty sure t6his is the case)

Tim

[Tim D Fulford]

Scrubb what was said above.

1 read the question
2 Dont contradict ..JRF..

Shut up....

Tim

[Anthony O'Connor]

Thanks for replies. We're dealing with vxfs using hp-ux LVM commands here. Was concerned that if a filesystem got extended while online on one node it would encounter problems when failed-over to another node but I have just tested and verified that it's taken care of automatically when the VG is activated. Obviously bigger changes such as adding disks or new LVs require a vgexport/vgimport.