Maintenance on VSA how to ?

[RHaarlemmer]

Hi,

We have a 2 node VSA with 3 managers configured. (Nodes + FOM)

Now we need to do some maintenance on the hardware where the FOM is running.

What would be the correct way to make sure the VSA keeps running and the FOM can be shutdown ?

I thought stop the manager on one of the nodes and then shutdown the FOM, is this correct ?

Just wanted to make sure and I couldn't find a clear answer on the web.

Regards,
RHaarlemmer

[Assen · Accepted solution]

Hi RHaarlemmer,

for the quorum to work and the MG to stay online, in your case you will need 2 managers to keep quorum up.

Just power off the FOM and do the Maintenance on the Host. After you have finished the task, power up the FOM again and check in CMC if it is back.

If you stop the manager on one of the nodes and power off FOM, than the VSA that remains will loose quorum, because there will be only one manager left in the management group.

BR

[Serverparken]

Hi,

I have a question in the same topic.
I have a 2-node VSA solution, with a FOM running on a separate site (over a low latency WAN link).
If I want to take one of the VSA's offline, for planned maintenance on the host where it's running, what is the best way doing this? I know that I can take one VSA node down without losing quorum (and the availability of the storage), as long as the FOM is available - but during that time, the availability of the storage all depends on the availability of the FOM (and the WAN link). In normal operation that's fine - but when doing planned/controlled maintenance (during office hours) I don't want the only running VSA node to be dependent on the FOM (and the WAN link).
Is there any way to manually control this behavior and force the only running VSA node to not go offline if it for some reason do lose contact with the FOM during the maintenance? Is there a "best-practice" way of doing this kind of maintenance? Can a Virtual Manager be used for this (or do I then have to delete the FOM)?

[oikjn]

you cannot add a virtual manager while a failover manager exists, so you would need to remove that first.

Do you have a two-node multi-site cluster, or are the two nodes simply in the same management group?  If you are that concerned about your WAN link, then it really isn't suitable for a multi-site cluster to begin with.

If the maintenance window is just a few minutes during a host update/reboot cycle or a minor hardware change, then you should just shut it off as you would normally do a reboot.  If it is going to be longer where the management group might take days of offline usage, I would unofficailly say that should just spool up and additional VSA node at the primary site and join them to the management group and then turn off the manager on the remote site and turn it on for the temporary VSA and do your thing.  That is way more work than should be needed.  if your cluster is only doing async-replication, then you might want to consider just setting up a 2nd management group for the remote site instead of having them linked so you don't have to worry about a FOM at a 3rd site.

[Serverparken]

Hi,

Thank you for your answer.

It's a two-node multi-site cluster.
I'm not very concerned about the WAN link, but as the maintenance could take a while (upgrading the ESXi hosts firmware, BIOS, etc. + upgrading ESXi) the WAN link (including routers, switches and infrastructure in the path) is the weakest link in the chain.
I suppose the best alternative is to temporarly remove the FOM and then add a Virtual Manager instead, if I want do stay safe and do this at daytime?
The other alternative, to just cross my fingers, requires this to be done outside of business hours I think...
To install a temporary VSA is not an option.

I guess the "best practice" is to rely on the FOM's availability and just take one node down for maintenance at a time, and let them re-sync in between?

[oikjn]

I wouldn't mess with the FOM unless you think there is a real chance that you will lose the 2nd site for more than a day.  If not, just shut it down and do your work and turn it back on without messing with the management group.

FYI, adding an extra VSA node is no big deal and as long as you don't add it to the cluster, you won't need to add actual storage to the node other than maybe a 5GB drive and it won't effect the cluster performance.  what will happen is that it will complain that the node isn't licensed and that it will be in a 60-day grace period, but that won't matter for this purpose as you will remove it long before that is an issue.

frankly, I"m not sure there is a really good "best practice for a 2-node multi-site syncronous cluster management group since a WAN issue on the primary site will alway cause the cluster to go offline, I'd sooner use a virtual manager loaded on the primary host and know that I'll have to manually intervene when failing over to the DR site and I would only use sync and not async replication IF the RPO needed to be so small that snapshot replication couldn't cover my data protection needs.  To me it sounds like you should either be switing to async repliction OR you really should be purchasing at least one if not two more licenses and hardware to allow NR10 at least on the primary site.