Shared LVM (FC) - HA if full multipath outage occurs

[gbonev]

Greetings,

We created POC with 2 Hypervisors and witness qdevice.
HA works as expected when pulling the plug on single HV.

Since we have decent DELL EMC / VPLEX - we decided to use shared FC with storage type LVM in order to be able to demonstrate HA.

I am trying to figure out if there is way to make Proxmox aware when one of the Hypervisors has completely lost multipath (but is still UP) - if there is way to make the VMs (re)start on the healthy Hypervisor where the multipath has no issues.

Thanks for any hints

 

[LnxBil]

That is a good question and I don't know the answer.

What is your use case for the FC problems on one host? We use multiple FC-HBAs and multiple switches so that this is very, very unlikely to happen.

 

[gbonev]

That is a good question and I don't know the answer.

What is your use case for the FC problems on one host? We use multiple FC-HBAs and multiple switches so that this is very, very unlikely to happen.

Yes indeed. We do not expect or wish to have full multipath outage at all.
In the meantime - we have 2 Data Centers with DWDN Dark Fiber between and the plan is to split the potential cluster's Hypervisors in both datacenters with disaster recovery in mind. Hence my quest for solution on potential full outage of Hypervisor's FC-HBAs

EDIT: outgae of HBAs might not be decent argument. But let's say loosing Zoning / multipath / HBAs for any reason. Endgame being not having the storage of this particular HV responding.
I suppose I could easily script the status of missing all mpaths.. But what is the proper (if any) way to tell the Prox to make HA decision based on that missing mpaths result?