cancel
Showing results for 
Search instead for 
Did you mean: 

Cluster upgrade from version 5 to 6

mokkan
Level 6
Certified

Hello,

 

we would like to upgrade veritas cluster 5 to 6 and. Some of the service groups are running parrallely. What is the best way to upgrade it?

Also, can we run veritas cluster 5.1 and 6 on other node in 2 nodes cluster?

1 ACCEPTED SOLUTION

Accepted Solutions

mikebounds
Level 6
Partner Accredited

If you have parallel groups, but are not using CVM, then you can just force stop VCS leaving applications running, unload GAB and LLT and upgrade VCS (including GAB and LLT) on ALL nodes.

If you have SF (non-CVM) on the nodes you can then upgrade SF on inactive node and then switch SGs from one live node to upgraded node and upgrade node that SGs were moved from.

If you have CVM, then you can't unload LLT and GAB without umounting the CFS mounts, but from 5.1, LLT and GAB support communicating on different versions, but the "had" VCS engine does not, so here the procedure is:

 

Use "installer -upgrade_kernelpkgs inactive_node_name" and this will upgrade packages like VRTSllt, VRTSgab, VRTSvxfs and VRTSvxvm, but not VRTSvcs, so after this LLT and GAB are communicating on different versions and VCS is left untouched.   Then you can switch SGs from one live node to upgraded node and upgrade node that SGs were moved from in the same way.
 
After all nodes are upgraded, then use "installer -upgrade_nonkernelpkgs node1 node2 ..." which upgrades packages like VRTSvcs and VCS agent packages on all nodes while applications are running.
 
Mike
 

View solution in original post

10 REPLIES 10

mikebounds
Level 6
Partner Accredited

If you have parallel groups, but are not using CVM, then you can just force stop VCS leaving applications running, unload GAB and LLT and upgrade VCS (including GAB and LLT) on ALL nodes.

If you have SF (non-CVM) on the nodes you can then upgrade SF on inactive node and then switch SGs from one live node to upgraded node and upgrade node that SGs were moved from.

If you have CVM, then you can't unload LLT and GAB without umounting the CFS mounts, but from 5.1, LLT and GAB support communicating on different versions, but the "had" VCS engine does not, so here the procedure is:

 

Use "installer -upgrade_kernelpkgs inactive_node_name" and this will upgrade packages like VRTSllt, VRTSgab, VRTSvxfs and VRTSvxvm, but not VRTSvcs, so after this LLT and GAB are communicating on different versions and VCS is left untouched.   Then you can switch SGs from one live node to upgraded node and upgrade node that SGs were moved from in the same way.
 
After all nodes are upgraded, then use "installer -upgrade_nonkernelpkgs node1 node2 ..." which upgrades packages like VRTSvcs and VCS agent packages on all nodes while applications are running.
 
Mike
 

mokkan
Level 6
Certified

Thank you very much for your reply. We are running VxFS as our files sytem. When you say inactive node? It means we stop the current service group which are running in parallel and do the upgrade on that node?

Can we join 2 node cluster with 5.1 and 6.0 ? I mean after we upgrade it to 6.0, can re joing with 5.1 cluster node? It is possible?

 

mokkan
Level 6
Certified

One more question. Veritas cluster only suppor Vxfs? Or can we run any cluster file sytem ?

TonyGriffiths
Level 6
Employee Accredited Certified

Hi

VCS supports a number of file system types for "local" mounts. Please check the Bundled Agents guides (Mount agent) for your platform.

Be sure to check what type of "cluster file system" you are using. VxFS can be used in a cluster as a local file system, which can be switched to nodes in the cluster, though is only online on one node at any point in time.

There is also a solution named Cluster File System, in which the file system is mounted and accessible on mutiple nodes. This solution has a seperate documentation set

 

cheers

tony

mikebounds
Level 6
Partner Accredited

You can have a 2-node cluster with one node running:

5.1SP1 for kernel pacakages like LLT, GAB, vxfs and vxvm

and the other node running:

6.0 for kernel pacakages like LLT, GAB, vxfs and vxvm.

But you only have different version of LLT, GAB, vxfs and vxvm in your cluster for a short period until you have upgraded 2nd node - you should not run the 2 nodes at different versions long term.

But VRTSvcs ("had" daemon) must be the same version on all nodes.

As Tony says VCS supports other filesystems as in bundled agents guide so for 6.0 for Linux this is:

 

vxfs, bind, ext2, ext3,ext4, xfs, nfs, or reiserfs.
 
Mike
 

mokkan
Level 6
Certified

Thanks a  lot both of you.

How do I verify that file system is clustered or not?  if we hae Vxfs file sytem, how do I know is it clustered file sytem or not?

mikebounds
Level 6
Partner Accredited

A clustered filesystem means you are able to mount the filesystem on more than one node at a time.  This is only possible if you have SFCFS licence and this would mean diskgroup would be imported on both nodes at the same time and so if you ran "vxdisk list", the disks would show as "online shared" as oppose to just "online" with "shared" if you have normal vxfs where you can only mount on one node at a time.

Mike

mokkan
Level 6
Certified

Thank you very much mikebounds. How about the following command?

 

cfsmntadm display -v

 

Would that show as well?
 

mikebounds
Level 6
Partner Accredited

Yes that should show cfs mounts if you have them.

Mike

Daniel_Matheus
Level 4
Employee Accredited Certified

Hi Genn,

 

what do you mean with verify file system in clustered?