Advanced vSAN configuration in sync

[larstr]

After patching ESXi hosts to the latest 6.7 16316930 we're now seeing a Critical alarm for this stretched vSAN cluster:

In the Hosts with this value field it lists all 12 hosts of the cluster.

This is an old Dell R730xd cluster with Perc H330 controllers. And I've found this article that recommends setting these values when using this controller. It states that these are the default values of a certain vSAN 6.2 version and newer and our version is clearly a lot newer.

So you clearly don't need to set this manually anymore. Except I found a reference that VCF 3.9 sets these values manually  behind the scenes.

Any idea what's the best way to proceed here?

Lars

[TheBobkin · Accepted solution]

Hello Lars,

"This is an old Dell R730xd cluster with Perc H330 controllers."

Are you positive these are H330 controllers and not H730p/HBA330 controllers?

If they are the former you should replace them as H330 are not supported for use with vSAN:

VMware Knowledge Base

"In the Hosts with this value field it lists all 12 hosts of the cluster."

So here's the thing: this check only triggers if there is a disparity between the settings on the hosts e.g. you can have all hosts with a custom non-default value (e.g. clom repair delay set to 60000 minutes) but it will not trigger the Health alert unless the configuration is inconsistently applied.

Perhaps the Witness is not being shown as the outlier due to not being a member of the vSphere-level cluster - did you update the Witness to the same build as the nodes?

Is the vCenter here also running on 6.7 U3?

"So you clearly don't need to set this manually anymore."

Correct, and we can also see they are all showing the default value as it is on any build later than 6.0 U2.

"Any idea what's the best way to proceed here?"

Validate that you updated the Witness and if not then update/redeploy later one - if this is not the case then I would start with seeing does the vSAN Health service logs give any more information as to the trigger (logs for the service itself and the UI output found in /var/log/vmware/vsan-health/).

Bob

[larstr]

Bob,

The controllers are listed as Perc H330 Mini and vSAN health is okay about them:

Yes, vCenter is 6.7U3 and witness is same build as the hosts.

The solution turned out to be one of the things you suspected: After applying these same non-default settings to the witness, everything is green again.

Thanks alot!

Lars

[TheBobkin]

Hello Lars,

Happy to help, but do please ensure Witness has been updated to the same build version as the hosts - this is quite important and sadly often goes missed.

Also, those are H730 controllers, not H330, difference being a couple of years and the former being supported.

Bob

[larstr]

Bob,

Ah, yes, that's of course the H730 :smileygrin:

Yes, as I confirmed above; the witness was already on the same build number as the ESXi hosts.

Thanks again for helping us get to the bottom of this.

Lars