SUMMARY: Cluster rolling upgrades - PK3 on half a cluster...

From: <Andrew.Raine_at_mrc-dunn.cam.ac.uk>
Date: Mon, 15 Oct 2001 15:03:53 +0100 (BST)

Dear All,

Thanks to Paul Lambert, Blake Roberts and Thomas Blinn for their
helpful suggestions.

Thomas, in particular, had some insightful comments on the clu_upgrade
scripts. However, after much digging around in there, I called Compaq
support, and their advice echoed his:

The only safe way to sort out the mess I have created is to:

Boot up the lead-member of the cluster in standalone mode (i.e. the
system from which the cluster was created). Patch it. Re-create the
cluster, upon which all the members will be running the patched kernel,
and the confused record of what I did to the original cluster will be
safely obliterated.

I'll schedule some more downtime, and get out the documentation...

Thanks again,

Andrew

--
Dr. Andrew Raine, Head of IT, MRC Dunn Human Nutrition Unit, 
Wellcome Trust/MRC Building, Hills Road, Cambridge, CB2 2XY, UK
phone: +44 (0)1223 252830   fax: +44 (0)1223 252835
web: www.mrc-dunn.cam.ac.uk email: Andrew.Raine_at_mrc-dunn.cam.ac.uk
Original Q:
> Dear Managers,
> 
> With the help of this list, I solved my NFS problem by applying PK3 to
> one of my machines.  However, now I need to install some more software,
> and I realise that I've got myself in a bit of a hole.  I'm hoping that
> someone can help me out!
> 
> Systems:
> 
> DS20 + ES40, TU 5.1 TC 5.1
> 
> History:
> 
> the two machines were clustered at the time we bought the ES40, but the
> DS20 had to be booted back into 4.X to support some people who were in
> the middle of some important work.  The ES40 was left in a "cluster" of
> one, NFS-serving to some workstations (TU, IRIX and Linux).  Problems
> with NFS on the ES40 were resolved by applying PK3.
> 
> Because the ES40 was still running TC5.1, I had to do this via the
> rolling upgrade route.  However, as I knew I couldn't complete the roll
> on the DS20 at the time, I stopped after the "install stage" on the
> ES40 (which was the "lead member" for the upgrade), and ran clu_upgrade
> postinstall on it.
> 
> Now, with the DS20 finally back in the cluster I need to finish the
> rolling upgrade and apply PK3 to the DS20.  However, the DS20 never
> knew that the cluster was being upgraded, so it thinks that I need to
> start at the clu_upgrade check setup <id> stage, while the ES40 thinks
> that it is at the "roll" stage.
> 
> Is there any (easy!) way I can convince the two machines to agree on
> the stages they think they are at?  Do I need to shutdown the DS20,
> uninstall PK3 on the ES40, boot the DS20 and start the rolling patch on
> both machines again?
> 
> Thanks for any light you can shed on my dilemma!
> 
> Regards,
> 
> Andrew

Received on Mon Oct 15 2001 - 13:53:55 NZDT