Cluster nodes patching strategy

[Mihails Nikitins]

Hi,

Are there any known problem issues in patching cluster nodes one by one? I mean having different patch levels of SG, LVM etc on different nodes until the next planned switchover. Thanks and points in advance for sharing your experience!

BR,
Mihails

[John Poff]

Hi,

The newer versions of MC/SG support rolling upgrades, so that your cluster will function with mismatched versions while you upgrade. We've done it a couple of times and it works fine.

What is your current version of MC/SG?

JP

[Bernhard Mueller]

Mihails,

this is not a problem

if fact, in an MC/SG environment - besides creating two ignite tapes or recovery archives - I think it is a good idea to update one node first, check everything is running fine for a while, then update the second, third etc.

This leaves you more options in case anything goes wrong. The only drawback is the higher number of downtimes required, which is why patching is often down in one go.

Regards,
Bernhard

[Bernhard Mueller]

I was assuming that you are not going to upgrade the MC/SG version. In which case, like John says, update all nodes immediately to the newer version is a must.

Regards
Bernhard

[Mihails Nikitins]

To John:

My current level is:

B3935DA A.11.15.00 MC / ServiceGuard
BUNDLE11i B.11.11.0306.1 Required Patch Bundle for HP-UX 11i, June 2003

To Bernhard:

I meant applying all patches recommended by ITRC
patch / firmware database.

Thank you!

[John Poff · Accepted solution]

I'm sorry. I thought you were talking about upgrading your version of MC/SG. Patching the O/S, firmware, database, on one node at a time is no problem either. It's how we do it here. We patch our failover node (which serves as our test/development box in normal usage) one month, and then we patch the production nodes the next month. I don't remember having any problems doing it that way, and it gives you some time to shake out the patches and back one out if needed.

JP

[Mihails Nikitins]

Tnahks for your replies! As you told,

It's OK to patch software separately on the nodes.
It's a must to bring the whole cluster down to upgrade MC/SG.

Do I understand correctly the following.

Installing HP-UX patches does not change MC/SG version, i.e. patching is not an upgrade.

[Bernhard Mueller]

Mihails,

your assumption is correct. There are patches for certain MC/SG versions, but installing any patches does not change the MC/SG version itself.

AFAIK MC/SG 11.16 is not yet released and you are on 11.15 which is the latest version.

Regards,
Bernhard

[#9]

11.16 is released so 11.15 is no longer the newest version. There is no problem in running different patch levels on different nodes although I would not recommend this permanently. There are very few fixes which do not take effect until the patch is installed on all nodes.

[kamal_9]

Hi Mihails
You are Correct ,we are also upgrading the patches by node by node
BEST OF LUCK

[Prashant Zanwar_4]

I would say something for patching the OS for VG00 atleast in production setup where you cant tolerate errors. Remove root mirror hard disks and then do the patching. Always do it on the contigency/alternate node first and then take same task on Main node. It again depends if you are running backup node or not. But do it on less similar system first. As sometimes really cause of problem takes time to get resolved. And apply HP rated patches only. Not the latest one, might be one step back.
Hope above helps
Prashant