How to: 2-node image based ESXi update

[JedB]

Hi!

I am currently switching to the image-based update and am wondering whether I should also deactivate the HA for updates in a 2-node cluster (no vSAN) as before.
In a test I have seen that if I deactivate it and then run the compliance check for the update, the HA component seems to be uninstalled. Message: Solution Components of disabled Solutions vSphere HA 8.0 U3 will be removed from this host during remediation.
So is it better to leave HA activated and will vLCM then take care of it?

[ninjabrum]

Leave HA on.  Lifecycle Manager will take care of things. The HA VIB will be removed by vLCM.  Then, when the host is brought out of maintenance mode, vCenter will push the latest HA VIB.  They could word it a little better, for sure.

[JedB]

Hi Ninjabrum,

I'm still thinking about the HA. I'm pretty sure the update will fail because of the missing ressources from one of the two hosts.
That's why I always disabled HA for updates in the past with VUM.
I'm not sure, if vLCM will handle the HA correctly because there is nothing mentioned like that in the remediation steps overview.
But in the remediation settings of the vLCM I recognized an option to let vLCM disable HA during the update. Maybe that's a good idea?
Or does it help to just tun off the host monitoring of HA during the update?

[Shen88]

@JedB,

Please note this message you saw in your test when you deactivated HA explicitly is a strong indicator of what happens when HA is disabled. If HA is disabled, vLCM (or the underlying update process) might interpret this as an intention to remove the HA functionality from the host, leading to the uninstallation of its components. This is not the desired behavior if you intend to keep HA active.
* Minimizing Downtime and Risk: Leaving HA enabled allows vLCM to leverage its capabilities for a rolling update. In a 2-node cluster, this means one host is updated at a time, ensuring that your VMs remain running on the other host. If you manually disable HA, you increase the risk of downtime during the update process, as there's no automatic failover mechanism if something goes wrong.

Hence, for image-based updates in a 2-node cluster:
* Ensure HA is enabled and configured correctly before starting the update.
* Use vLCM to perform the compliance check and remediation.
* vLCM will handle putting hosts into maintenance mode and bringing them out, coordinating with HA to ensure minimal disruption.

By trusting vLCM and HA to work together, you'll have a more robust and less error-prone update process.

------------------------------
If you find this answer right, please 'Recommend' this post.

Thank you!

Regards,
Shen
------------------------------

[Bjoern Jeide]

I have asked the Broadcom support and the correct way is to edit the remediation settings of the 2-node-cluster and to activate "Disable HA admission control on the cluster". This will deactivate HA during the remediation process and ensure that the hosts will enter the maintenance mode for the update even if there is a violation of the cluster ressources. 
Doublecheck first if both hosts have enough resources for all the running VMs or shutdown some of them.