Both vSAN uplinks down impact

[andvm]

Hi,

Am testing a scenario to observe the outcome when a host loses both vSAN Dist Switch uplinks.

I am seeing that when this occurs the test VM gets restarted by HA on another Host and it’s accessible again.

However vCenter (6.7) shows the VM as residing on the new host at one time, then on next refresh it shows it residing on the old Host.

When browsing to the old host VM list (still accessible as only vSAN uplinks brought down) it shows the test VM in the VM list and the when it refreshes it disappears and VM is listed on the other  Host it failed over to.

Any suggestion why the behaviour is occurring and what further actions are needed?

Thanks

[andvm]

FYI: After I re-enabled one of the vSAN uplink the VM was showing correctly and permanently on the failed over host.

[depping]

Do you have Host Isolation configured? And is your HA correctly configured as well to ping the correct IsolationAddress? As what you are describing sounds like the host isolation address is incorrectly configured.

[andvm]

Hi depping​

No, das.isolationAddressX is not set and none of the vSAN clusters are setup with this. (vSAN network is not routable so there is no Gateway)

Each Host in the cluster is a Standalone fault domain. (Think this is the default behaviour if no fault domain is setup)

Although I also see "Fault domain failures to tolerate" set to 2 (Not sure how cluster can tolerate 2 failures at one time without data loss assuming vSAN default storage policy is in use) - Unless I am misinterpreting the meaning of this.

Let me know if you have any further thoughts or recommendations.

Thanks

[depping]

Please read this section in the documentation and follow the best practices around configuring HA:

vSphere HA considerations | VMware® vSAN™ Design and Sizing Guide | VMware

[andvm]

In practical terms do you mean...

Setup an SVI and add das.isolationAddressX to point to it

Set das.useDefaultIsolationAddress=false

There is only a vSAN datastore for now so Heartbeat datastores is not possible.

If I understand correctly setting das.isolationAddressX and/or Heartbeat datastores will allow for additional checks to confirm if a Host vSAN connectivity is down.

What I’d like to understand is if having heartbeat datastores is enough or if I would still need to set

das.isolationAddressX ?

Thanks

[depping]

What I mean is follow the best practices listed:

1. Configure the isolation response to be: power off and restart VMs
2. Configure "das.usedefaultisolationaddress=false"
3. Configure a "das.isolationaddress0" on the vSAN network
4. Yes, the isolation address can be an SVI or any other reliable IP address

You do not need heartbeat datastores.