Prism IP unavailability during cluster upgrade

Badge +3

we are using a VDI solution that connects to Prism REST interface to deploy Virtual Machines as they are requested.

When we are upgrading the cluster, the VDI solution can't connect to Prism REST interface and so, new virtual machines can't be deployed for a brief period time.

I don't know if we should consider this normal and we should assume this stop of the service during cluster upgrade or this could be improved some way. Perhaps reducing MAC cache time-out or something else. In our setup Nutanix cluster is located in one VLAN and the VDI solution is located in another VLAN, so the communication of the VDI solution with the Prism REST interface goes through routers and firewalls.

Any idea? Thanks!


2 replies

Userlevel 2
Badge +5
cluster AOS upgrade is rolling upgrade , it upgrade one CVM and restart, do the post restart check and move to another CVM.
Badge +3
Yes I know that. In fact, our problem happens not with AOS upgrade but with hypervisor upgrade (which implies a reboot of both the CVM and the hypervisor).

I think that when the CVM which provides Prism interface is going to reboot, the Prism interface is delegated to other CVM which assumes the control of the cluster. For this procedure our VDI server losses connection to Prism REST interface and new VDI VMs fails to be created.

My question would be if there are any tip or best practice we could follow to reduce this time of unavailibity of the service during the hypervisor upgrade.