cancel
Showing results for 
Search instead for 
Did you mean: 

Upgrade A3 cluster

Upgrade A3 cluster

MLD
New Contributor
I'm trying to upgrade my 3 node A3 cluster from v3.5 to v4.0, but can't find any procedure regarding the cluster upgrade.
1 ACCEPTED SOLUTION

Robert_Haynes
Extreme Employee
Hello.

The upgrade procedure is fairly straightforward.

It would be recommended to power down the cluster first, the member nodes before the master. Once offline, take snapshots of the appliances for any disaster recovery type scenario.
Bring the cluster back online.
You would then log in, click on the tools icon (wrench/screwdriver icon) and select "Update software".
It should say 4.0 is available. Proceed with the upgrade.

View solution in original post

3 REPLIES 3

Robert_Haynes
Extreme Employee
Hello.

The upgrade procedure is fairly straightforward.

It would be recommended to power down the cluster first, the member nodes before the master. Once offline, take snapshots of the appliances for any disaster recovery type scenario.
Bring the cluster back online.
You would then log in, click on the tools icon (wrench/screwdriver icon) and select "Update software".
It should say 4.0 is available. Proceed with the upgrade.

How long should I wait between each node powering up (master first).
Can I run the update using the cluster VIP or should I update each node. Can the update run on all 3 nodes at the same time and how long does it typically take.

--
Thanks,
Matt

There is no set time; it's a matter of what you are comfortable with and what your experience has been. In my own environment I simply turn on all three nodes in a cluster and allow it to converge - this is how it is designed. However granted there are issues from time to time so waiting until say the GUI of the first node is operational (i.e. you can log in) to wait to start the next node and so on is as arbitrary a wait period as any.

I run / start the upgrade from the VIP, which is to say you run / start it from the Master member of the cluster. The upgrade runs in consecutively; and if memory serves it is a member node first, restarted following by the next member node, followed by the master... but don't quote me.

As for time - YMMV. On my 'fast' system ~ 20m for a cluster.. on my slow hypervisor... hour+.
GTM-P2G8KFN