SUMMARY: Upgrading firmware in a cluster

From: Edgardo Safranchik <esafran_at_aluar.com.ar>
Date: Mon, 10 Jan 2000 12:49:55 -0300

Thanks to: Knut Hellebø, Claudio Tantignone, Thomas Strandenæs, Dagmar Galama, Piet Lauwen, Alan Davis, Rodrigo Poblete, John Speakman, Miguel Ward for their answers. There are two different opinions about the problem, here there are:

* Opinion 1: Stopping the entire cluster

Some people suggest to stop the whole cluster, do the firmware upgrade over all nodes and startup. This is the safest procedure. The justification they have is that the upgrade does a bus reset over the shared storage. The question, of course, is: why my boss has approved the investiment in a cluster if I can't have it providing a 7x24 service?

* Opinion 2: Not stopping the cluster

Most of people agree that it isn't necesary to stop the cluster, some of them have done the upgrade this way without problems (even with OpenVMS over the Alphas). The procedure follows:

1- Be sure that the services doesn't have "automatically relocate if other member becomes available" configured.
2- Stop one node
3- The services are reallocated over the other node (if the ASE is well configured ...)
4- Shutdown the node
5- Upgrade firmware
6- Startup
7- Go to 2 for the other node.

In both cases it is very safe to have the CD with the old firmware, and make a full backup before the procedure.

In my case I will try the second form, not stopping the cluster, because the problem I want to solve is many "SCSI bus resets" and other errors with the shared disks, so the justification of the first opinion doesn't matter to me.

Thanks to everybody

Edgardo Safranchik

----- Original Message -----
  From: Edgardo Safranchik
  To: Managers de Tru64 Unix
  Sent: Wednesday, January 05, 2000 8:44 AM
  Subject: Upgrading firmware in a cluster


  Has anybody done a firmware upgrade in a cluster environment without putting the whole cluster offline?
  The problem is to continue offering a 7x24 service, so I want to have one node providing service while I make the upgrade into the other.
  I have a 2-node AS4100 cluster with Digital Unix 4.0D + TruCluster 1.5 + PK3 and firmware rev. 5.1 and I don't want to stop the services that the cluster is offering.
  I understand that an OS software upgrade requires to stop the cluster, but I suppose that a firmware upgrade doesn't.

  Has anybody have experience with this? What are your opinions?

  Thanks in advance

  Edgardo

  -------------------------------------------
  Lic. Edgardo S. Safranchik
  Tecnología Informática
  Gerencia de Sistemas
  Aluar Aluminio Argentino SAIC
  TE: (5411)-4725-8000 (ext 2262)
  email: esafran_at_aluar.com.ar
Received on Mon Jan 10 2000 - 15:51:40 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:40 NZDT