topic Re: SG considerations for 11iv1 to v2 upgrade in Operating System - HP-UX

SG considerations for 11iv1 to v2 upgrade

arkie — Sun, 19 Sep 2010 08:45:05 GMT

Hi All,

Need your guidance regarding the following:-

An existing 2 node R3 production cluster has the following:-
1. ServiceGuard A.11.16.00
2. SG-NFS-Tool A.11.11.06
3. SG-Ext-SAP-R3 B.04.01
Current OS is HP-UX11iv1 and OE/ReleaseID is HPUX11i-OE, B.11.11.0509.

The requirement is to upgrade the OS to B.11.23.0806, FOE (or later, based on availability of Media).

Need you experts opinion on upgrade considerations for SG, SG-NFS-Tool and SG-Ext-SAP-R3.

Thanks in advance

Re: SG considerations for 11iv1 to v2 upgrade

SoorajCleris — Sun, 19 Sep 2010 14:21:41 GMT

Hi ,

Go to docs.HP.

download the doc "install and update da da da da"

for each version of product you want to upgrade to. It will be there , how to do that.

Suggestion is to do a cold install!!!

Regards,
Sooraj

Re: SG considerations for 11iv1 to v2 upgrade

melvyn burnard — Sun, 19 Sep 2010 21:37:09 GMT

first thing to do would be to check whether the relevant toolkit/software is supported on HP-UX 11.23
So, from here:
http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c02466761/c02466761.pdf
page 12, Serviceguard A.11.16 went out of support on HP-UX 11.23 last year.
However, to do the upgrade this is not so important as you can upgrade Serviceguard once the OS upgrade has been completed succesfully, and you should upgrade to at LEAST A.11.18 of Serviceguard at that time.

SGeSAP B.04.01 is also supported on HP-UX 1.23, but once you uopgrade Serviecguard you should consider upgrading the SGeSAP AND the NFS Toolkits to a later version.
Bear in mind that installing these toolkits merely loads them into the /opt/cmcluster directory, and you would need to use the newer files supplied in those locations to modify your existing package configuration and control files.
I would suggest you first do the OS upgrade, and make sure all is well at that stage and THEN consider doing teh upgrade of the Serviecguard and toolkit software

Re: SG considerations for 11iv1 to v2 upgrade

arkie — Mon, 20 Sep 2010 07:26:51 GMT

Hi Melvyn, thanks for your inputs and SG 11iv2 compatibility check link.

Can you pls send me similar compatibility check links for SG-Ext-SAP and SG-NFS-Tool

Re: SG considerations for 11iv1 to v2 upgrade

melvyn burnard — Mon, 20 Sep 2010 08:43:48 GMT

not quite the same, but take a read of this customer letter for SGeSAP:
http://www.hp.com/softwarereleases/releases-media2/notices/SGeSAPSupportLetterAug2010.pdf

Also for Serviceguard:
http://www.hp.com/softwarereleases/releases-media2/notices/SGSupportLetterAug2010.pdf

I am currently unable to locate one for the NFS Toolkit, but you should always look to use the lates version for the relevant OS and Serviceguard software installed.

Re: SG considerations for 11iv1 to v2 upgrade

nightwich — Mon, 20 Sep 2010 09:07:09 GMT

Hi arkie.

My five cents.

If you have Service Guard you should have a cluster of two nodes minimum.

So I will use this aproach:

- Make_tape_recovery for both nodes.
- Pass all packages to one node ( node 1 )
- Halt de node 2 in the cluster
- In node 2 cold install of OS SG and toolkits.
- Regoin the node 2 to the cluster.
- Pass the packages and see if everything is function correctly
- If yes do the same for the other node.

If you have any doubt reply.

Regards.

Re: SG considerations for 11iv1 to v2 upgrade

arkie — Mon, 20 Sep 2010 11:22:23 GMT

Hi Nightwich, thanks for the insights.

Major doubt here - Say, Node 1 (11iv1) is having SG A.11.16, NFS-Tool A.11.11.06 & SGeSAP B.04.01. After doing the upgrade in Node 2, let's say it has 11iv2, SG A.11.19, NFS-Tool A.11.11.08 & SGeSAP B.05.00. Will they rejoin into a cluster

Re: SG considerations for 11iv1 to v2 upgrade

nightwich — Mon, 20 Sep 2010 11:42:32 GMT

Hi arkie

I think the answer is yes.

I already do a upgrade similar to yours ( without those toolkits ) and this was the aprouch that I use.

Althought you must check if version's of SG are compatible.

Take in consideration 2 things:

1 - If you use SG_RAC you can't use this aprouch:

"SGeRAC is not supported in a heterogeneous
cluster because Oracle RAC does not support mixing HP-UX versions within a
single RAC cluster."

2 - Have a good backup of the server ( if you need to go back you can easily do it with recovery tape )

Also don't forget to assign some point to usefull answers.

Regards

Re: SG considerations for 11iv1 to v2 upgrade

arkie — Mon, 20 Sep 2010 11:53:32 GMT

I always assign points for the good answers. But the thread will be open until .. can't say now-probably after a successful upgrade. So Melvyn,Nightwich - pls keep visiting.

Re: SG considerations for 11iv1 to v2 upgrade

melvyn burnard — Mon, 20 Sep 2010 14:31:52 GMT

Major doubt here - Say, Node 1 (11iv1) is having SG A.11.16, NFS-Tool A.11.11.06 & SGeSAP B.04.01. After doing the upgrade in Node 2, let's say it has 11iv2, SG A.11.19, NFS-Tool A.11.11.08 & SGeSAP B.05.00. Will they rejoin into a cluster

NO they won't!
You CANNOT do a rolling upgrade to A.11.19 from anything below A.11.18.
If you do an upgrade make sure you have S GA.11.16, or at most A.11.17, that should work.
However, if you are looking at doing a cold install, that is NOT a rolling upgrade, and requires different steps

Re: SG considerations for 11iv1 to v2 upgrade

nightwich — Mon, 20 Sep 2010 14:52:22 GMT

Melvyn

I do not suggest a rolling upgrade, but a cold install.

- Halt the node2
- Cold install
- Upgrade SG version and toolkits and then re-joing the node to the cluster.

As you said, I also advertise that the versions of SG must be compatible (see in the manual of SG version):

"Althought you must check if version's of SG are compatible"

If they are compatible I do not see any problem.

Regards.

Re: SG considerations for 11iv1 to v2 upgrade

melvyn burnard — Mon, 20 Sep 2010 16:02:25 GMT

As I said previously, a cold install of a node is not an upgrade, it requires other steps be taken.
First of all, doing a cold install may change disk device files.
Secondly, it may also change lan id's/names
and that is just teh start.

To do a cold install requires:
1. remove any definition of the node to be installed from ALL packages, and apply these changes with cmapplyconf
2. remove any definition of the node to be installed from the cluster configuration file, and use cmapplyconf to make these changes (in other words reduce it to a single node cluster).
3. create a backup of the node.
4. install the node, but make sure you include the same version of Serviceguard.
5. copy/update the Serviceguard security files to allow the new node access.
6. using cmquerycl, create/update your cluster ascii file with the installed node, then check and apply the configuration
7. make sure the node can join the cluster
8. copy the relevant package files into teh correct directories, make sure you have the mount points created as required, import your volume management set up and verify these all work.
9. update each package to add in the new node, and apply
10. test the whole set-up.
11. repeat steps 1-10 for the second node.
looka at upgrading Serviceguard and toolkits.

And under no circumstances try to go from a.11.16 directly to a.11.19 of Serviceguard, it will NOT work.