Upgrading OS on a Service Guard cluster

[Terrence]

We have been placed into the uneviable postion of being asked to upgrade the OS on a 2 node service guard cluster from 11.11 to 11.31.

And of course we have only a couple of hours a day of downtime to do this while keeping the application otherwise up and available.

The admins here all agree that the only practical way to do this is to either build from scratch on new hardware or to break the cluster, upgrade and rebuild on the current off node, and make that production. Then upgrade the other node. Finally we make the application into a package and recreate the cluster.

Since new hardware ain't happening, we're stuck trying to juggle this on the existing cluster.

I will award points for all useful advice.

[James R. Ferguson]

Hi:

Yes, "unenviable" indeed.

A tedious, alternative *might* be to upgrade from 11.11 to 11.23 and then (later) from 11.23 to 11.31.

However, if it were my choice, I'd push for either new hardware or to break the cluster in the fashion you and your colleagues noted.

Regards!

...JRF...

[Terrence]

What if we just upgraded to 11.23?

COuld we upgrade to 11.23 and a newer version of SG to an existing cluster?

Is this PRACTICAL to a live cluster?

[James R. Ferguson]

Hi (again) Terrence:

> What if we just upgraded to 11.23?

Since you can't step from 11.11 directly to 11.31, I suggested that you move first to 11.23.

This would fulfill one of the requirements for a Serviceguard rolling upgrade. For more information see:

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c02057890/c02057890.pdf

Regards!

...JRF...

[Steven E. Protter]

Shalom,

There is no 11.11 upgrade to 11.31.

To upgrade you must first go to 11.23 then to 11.31

General methodology is to migrate all services to a single cluster node.

Halt the cluster service on the node to be upgraded.

Upgrade it. Better to just do a cold install of 11.31, then re-install all the applications and build a new cluster.

Then arrange a down time, bring down the production node, bring up all services on the new 11.31 node and repeat the process on the 11.11 node you were running production on.

Note also, you may need to do firmware updates to get 11.31 to work. Also some 11.11 certified hardware can't run 11.31.

See http://docs.hp.com for certification information.

SEP

[njia_1]

Hi everyone,

He said he needs to upgrade OS from 11.11 to 11.31 so I think the hardware he has is PARISC.

Can you install HPUX v2 on PARISC system like rp3440? I tried few times never worked, the installation just hang.

One important thing to check is that your application runs on HPUX v3 before you start doing anything.

here is what I will do.

Take the standby node out and rebuild it with HPUX v3, also install SG and other required software. Create an Ignite backup tape

at next downtime, shutdown the package and rebuild the 2nd node with the tape and also do all the vg configures on both node and configure the cluster.

take your time to create the ready-to-go Ignite backup, then you should be able to get everything done in about 4 hours unless you have a very slow server like K380 ;)

thanks