VCS - 3 node - IP change

UNIX Solaris
1

Hey Admins,

I need steps to do an IP changes on 3-node VCS cluster. All IP's (public, cluster IP's, backup IP, etc.)want to change. We have a downtime scheduled. Please let me know the steps to do so.

Thanks in advance.

2

Perhaps a rather over-familiar form of greeting there, but "hello" to you too snchaudhari2.

Can you elaborate on a few things for us:-

  • What OS and versions are you using
  • What do the manuals say about it?
  • What does your software support say?

I don't want to be flippant with these questions. Clusters are great when they work well but can be a problem when they don't. Some errors can trigger an immediate shutdown/crash as that can be seen as the best course of action in some circumstances, so it is very important to get it right.

Assuming that VCS is licenced, then you should have some support and I would advise getting them to walk you through it. It may not be directly from Veritas, but whomever you get OS support from. Of course, here I'm hoping you are not going to be using a free Linux variant. IBM AIX, HP-UX, Sun/Oracle Solaris, RedHat Linux etc. are all commercially supplied and should have support options to help you in your specific circumstances. If you are going with a free OS, you get what you are paying for.

Regards,
Robin

3

Solaris 10
VCS - 5.1.10.20

Yes. It's licensed. But just want to know steps before I can reach out to them. In general I mean. We have downtime for all 3 nodes and it's for these IP-changes.

4

The address that you publish for applications/people to use should be part of your cluster definition (and a DNS entry, of course) so that should be straightforward. The problem will come with how the nodes know each other. These will be the IP addresses you put on the cards as part of the OS and it depends how the cluster was created. These may be both the IP addresses that could be used if people knew about it and any that are non-routeable more usually used for backups or iSCSI access (if you use that) because your cluster nodes may all be set up to see each other across all the various paths.

In the extreme, you may need to do something like this for all nodes in the cluster in parallel:-

  • Stop cluster services (and prevent start on boot)
  • Edit cluster configurations for all heartbeats and node definitions
  • Set new static IP addresses on cards
  • Boot
  • Start cluster services2 (and enable start on boot if required)
  • Test

It may be better to edit the cluster configurations after the boot, and that is where their advice will be the best.

Can you list off the cluster configurations? They will be quite large so don't post them here, but your support will almost certainly want them. If you were AIX with IBM HA I'd have more advice. Sorry about that :frowning:

I hope that this is useful though,
Robin

5

Thanks Robin,

I will try these steps in couple of days. Thanks for your response. I'll keep posted.

6

Don't take this as a definitive process. I really recommend that you talk to your supplier. Suggest it to them if you wish, but make sure that they confirm the official sequence.

I may have left something out and you should not rely on my thoughts for something so critical that may be difficult to recover from.

Regards,
Robin