1833883 Members
1624 Online
110063 Solutions
New Discussion

patching serviceguard

 
SOLVED
Go to solution
George Barbitsas
Frequent Advisor

patching serviceguard

Hi,

We patch our systems twice a year....this time around I came across the patch PHSS_36898....with crazy instructions in the readme...The documentation specifically indicates that both nodes need to be patched.
I would like to know if there is a procedure of patching mc serviceguard without actually stopping the applications, because there is no way the clients will authorize downtime for this update...I know for a fact that with VCS and older Firstwatch software you had a way to stop the clustering software without actually stopping the application contained within...
7 REPLIES 7
Ivan Krastev
Honored Contributor

Re: patching serviceguard

You can patch the SG cluster with the following procedure:
1. Move all packages from node1 to node2 - small dowtime is required and it depends from application start/stop times.
2. Patch node1, reboot, check.
3. Join node1 in cluster, move packages to node1.
4. Patch node2, check, redistrubute packages ...

regards,
ivan
Ganesan R
Honored Contributor

Re: patching serviceguard

Hi,

Not much details about your cluster setup.

Why not you patch one node after moved the packages to other node if it is capable of running. Then move back the packages to patched node and patch the other one. It is the feature of MC Service Guard right?
Best wishes,

Ganesh.
Doug O'Leary
Honored Contributor
Solution

Re: patching serviceguard

Hey;

The VCS process you're talking about is called freezing the cluster after which you can literally stop the cluster (leaving the application running) and do whatever you need to do.

That functionality is not present in MC Serviceguard.

While you can normally do rolling patches with MCSG, you will occasionally have MCSG patches that require the cluster to be down. It sounds like you finally ran into one of those.

I haven't read the patch notes, but it sounds like this is one of those occasions. In order to install that patch, you will have to drop the cluster which means an application outage.

Doug O'Leary

------
Senior UNIX Admin
O'Leary Computers Inc
linkedin: http://www.linkedin.com/dkoleary
Resume: http://www.olearycomputers.com/resume.html
George Barbitsas
Frequent Advisor

Re: patching serviceguard

I am trying to avoid downtime....we have multiple clusters and each cluster has 2 nodes(Psite/Osite)....Is there a way to stop the cluster daemons install the patch and restart serviceguard without affecting the application?

thx
George Barbitsas
Frequent Advisor

Re: patching serviceguard

ok thx Doug, thats what I was looking for...If I cant do this without downtime, then I will reschedule @ a later date...

thx

Re: patching serviceguard

Eric,

HP-UX patches are not "cumulative" in that you don't generally have to install a previous patch before installing a newer one.

So the long list of special instructions includes *all* the special instructions you might have to carry out for *all* the Serviceguard patches that have been released for SG 11.16, which is *a lot* of patches - 12 in total if you look at the patch family tree:

http://www12.itrc.hp.com/service/patch/patchTree.do?patchid=PHSS_36898&sel={hpux:11.11,}&BC=main|search|patchDetail{PHSS_36898,{hpux:11.11,}}|

So PHSS_36898 includes the fixes from all those other patches, but the special instructions have to assume that you don't have *any* of those previous patches installed.

If as you say, you patch your systems twice a year I seriously doubt that you don't already have a very recent SG patch installed, so many of these special instructions probably apply to fixes you have already applied.

Read the special instructions carefully and review against the patch family tree and I'm confident you will find that most of the special instructions can be ignored as the fix is already in place.

HTH

Duncan

I am an HPE Employee
Accept or Kudo
George Barbitsas
Frequent Advisor

Re: patching serviceguard

Question was answered