In a cluster, can I do a rolling reboot if I want to apply ECO 7 of TCPIP V5.4?

[Joanne Korol_1]

I want to do a rolling reboot of my Production VMScluster. I've done rolling reboots with VMS patches and everything went fine. I'm not 100% sure if it's okay to do with TCPIP so I don't affect any clients on our servers.

[Richard W Hunt]

My two-node cluster didn't crash when I did exactly what you suggested, though I did not allow too much time to elapse between the two reboots.

I wouldn't stop and restart services on the second node after patching the first one unless that stop/start operation is for the entire TCPIP services suite. Even then, it is probably better to just let it run until you can do the second node's reboot. And the longer you wait, the more likely you will be to run into issues with mixed versions being active on the second node.

A factor that makes a difference is whether you have shared or separate system disks. With separate disks, you actually would be MORE stable for a delay - but then, you have to do the patches twice in that case anyway.

[Joanne Korol_1]

We have a shared system disk in a 4 node cluster. Clients use the servers 24 hours a day. Once a day the clients are logged off for 5 minutes. Because of this, it allows me to take 1 server of out the DNS alias table each night when I want to perform maintenance.

At most, it would take 4 days and that is what worries me.

I'm thinking if it was okay to perform rolling reboots, I would have found the answer already. Don't you agree??

Nothing is mentioned in the release notes.

[Bill Hall]

Joanne,

I don't ever recall intentionally doing a complete cluster shutdown just for a TCPIP Services ECO install. I know I've done a lot of rolling reboots of cluster members after installing multiple VMS ECOs combined with a DECnetV ECO and/or a TCPIP Services ECO.

If in fact, a shutdown of the entire cluster was required, it should be well documented in the ECO's release notes or the installation guide.

Bill

[Jan van den Ende]

Joanne,

on our (mostly 4, sometimes temporarily 5-node) cluster, we have done ONLY rolling upgrades for the past (nearly) 11 years.

That includes EVERY upgrade, hard- as well as software.

And there HAVE been times, when it took over two weeks between first-node and last-node reboot.

You can be 99.99 % sure that everything goes well.

hth

Proost.

Have one on me.

jpe