Upgrading N-classs with HPux11.00 to HPux11.V1 (how ?)

[Frank de Vries]

We have 2 N-class , clustered with Mc/Sguard A.11.12 and OS version HPux 11.00.
running loadsa apps.

We need to plan an upgrade to 11.V1
(By the Way hpux 11.V2 is not possible
unless we change most hardware inside, so
we will opt for that)

The question I am having:

Does it make sense to do this upgrade on top
of the old HpUX11.00. Can I expect all apps
to be running afterwards ?
Is there a good cookbook anywhere I can use , or is this a tricky upgrade that may jeopardise our production system.
(We will have at best only a weekend
to do it all). So it is business critical.

It seems smarter to install fresh on a third N-class and migrate the applications like Oracle
and WebSphere ? But this will take loats more
time of course. And Time is something we haven't got. (Like everywhere)

You can give both technicall and functional/cost/time aspects if you
have had experience on that.

Your input would be cordially appreciated.
Cheers

[MarkSyder]

Do not use update-ux - you will end up in a mess. Take an ignite backup (preferably two) and do a cold install of HP-UX 11.11.

I hope you do not have application software/user data etc. on vg00. If you do, you will need to back it up to (e.g.) Data Protector and restore it afterwards. If you do not, you will just have to import the application/user VGs after you have done the rebuild.

Mark Syder (like the drink but spelt different)

[Luk Vandenbussche]

Frank,

I have done several updates with update-ux succesfully. I don't agree with the statement of reinstalling the OS.

I always test the upgrade on an standby server first. (Ignite of production to test server) So you can perfectly test the upgrade and application before you do it on your production servers.

[Frank de Vries]

Luc
Okay, I can see that your approach would work for the OS upgrade:

How much time did you have to spend on test the upgrade on the testserver.

(In the end of the day the OS may upgrade well, but it is the applications and services that need to stay available to the customers)

For instance Cobol link libraries will have
to be relinked and tested.
Stuff like that.

What do you comment on these things,
I would be interested to know

[Luk Vandenbussche]

Frank,

We take an other independent server.
On this server we restore an ignite tape and and try to start all the applications.

If everything is working fine, then we perform an upgrade on the test server and test all the applications.
If everything is working fine on the new version then we do the upgrade on the production servers.

I have prepared a server at a customer site to do this upgrade. I took me one day, now he is testing the applications, if this is OK we will perform an upgrade on the test server.

If necessary, we can provide you test server in Belgium ....

[Frank de Vries]

Impressive Luk :)

We have a couple of N-classes that just
freed up. I will use those.

thanks for your reply.

[Bill Hassell]

WShether you upgrade or do a cold-install, you will likely spend a lot of time verifying and/or updating your applications. Although the OS upgrade is forward compatible with executables (both 32 and 64 bit), not all applications are well-behaved or well-written to be independent of the OS version. You will have to research both your current versions of Oracle, Webshere and any other applications to see if the current version is supported on 11v1.

Since this is production, I would take the safest path: cold install on a spare server, install the latest versions of the applications, license and configure them and then test them, all in parallel with the current production server. Otherwise, you may be picking up the pieces for weeks afterwards (and spending a lot more time on the upgrade than just a weekend).

[Stephen Doud]

What are you concerned most about?
Your questions is focused on upgrade issues.
Does Serviceguard play into your upgrade questions?
Do you want to keep the applications running as much as possible during the upgrade, or is it permissible to shut the packages down?
If your questions a solely focused on HP-UX upgrade, I can transfer your query to a better area for this question.

[Frank de Vries]

What are you concerned most about?
> It is just that a lot is running on
this cluster.
- 9 Oracle db's and 6 oracle clients
- 2 Apache Web Server for load bal.
- 2 WebSphere app server for load bal.
- Tivoli openview agent
- dataprotector agent
- CA Unicenter autosys job management scheduler
- Mc/SG A.11.12
- Cobol development kitYour questions is focused on upgrade issues.
Does Serviceguard play into your upgrade questions?
> Yes we have SG A.11.12 and would like to take the opp. to go to A.11.16
Do you want to keep the applications running as much as possible during the upgrade, or is it permissible to shut the packages down?
> We have potentialy a day for planned downtime.
If your questions a solely focused on HP-UX upgrade, I can transfer your query to a better area for this question.
> That would interest me too actually

[Michael Steele_2 · Accepted solution]

Two pieces of advice. A) Split the nodes. B) Move all data but O/S off of vg00 and related disks.

A) This is also your roll back plan. If something goes wrong nodeB is still intact.

So Install / Upgrade on nodeA but continue running the live production environment on nodeB. Create you mount points and bring nodeA with your other vg's after satisfied that the new O/S is up. You'll have to unmount the data from nodeB before mounting on nodeA, just make sure you can see everything in ioscan first.

Once you've migrated the data over to nodeA, make an ignite tape and use it on nodeB and follow through on the regular vgexport and vgimporting in the service guard procedures.

B) vg00 has to be exclusively O/S. So clean it up first. 'pvdisplay' all the vg00 disks and verify they have no data. O/S only.