TOPIC:

I have a 2-node Citrix Hypervisor 8.2 (XenServer) deployment, each host-node having an on-board dual port NIC and an add-on quad port NIC for a total of 6-NICs per host.

Shared storage is available via Synology NAS over NFS. Connectivity to storage is redundant links using LACP. There is only a single switch providing ALL connectivity for infrastructure, servers & clients. Switching will remain a single point of failure, but if clients end-points loose connectivity, not much value in having storage be up.

My goal is only to support a two-node pool where I can restart a host without fencing and having the other server shut itself down as a result.

I do not need the iSCSI nor NoSAN features from HA-Lizard. I only need to resolve the self-fencing scenario.

Adding the extra NICs for connectivity to storage changes the 4-NIC design detailed in all the HA-Lizard guides.

I am considering allocating NICs as:

NIC 0+2 = Management & HA-Lizard Sync
NIC 1+3 = NFS Storage
NIC 4+5 = VM Guest

Is this the best design in sharing management & HA-sync on the same bond thru a switch?


For extra credit -

Is this a REAL problem?
HA-Lizard install prompts:
-- This host release (Citrix Hypervisor release 8.2.0 (xenenterprise)) not detected as an XCP or XenServer host
continue anyway? <CTRL + C> to exit OR <Enter> to continue

The install guides seem to all cover NEW hypervisor deployments. Is it not supported/advised to deploy in an EXISTING hypervisor?

I think your overall design is good which isolates management, storage and VM traffic.

Your install should be pretty strait forward. You only need the latest HA-Lizard RPM. Don't use the no-san installer which adds the additional iscsi storage management and dependencies.

Regarding the error.. That is harmless and a result of "XenServer" changing name to "Citrix Hypervisor".

Thanks for bringing to our attention.

So I created a new pool with a single host and installed HA-Lizard. Finally enabled HA.

What steps are required to define the HA-sync ports per design. Will this be a choice when installing HALizard on the second host?

Or will the HA just look for any and all connectivity to its peer and use anything availablr?

There is no interface specific configuration required. Health checks and HA data sync are performed across the management network.

Additionally, you need to set the value of heuristic_ips. This setting acts as an additional health check to IP(s) you set. For example, if you set this to the IP address of the external network storage device, the checks will naturally route via the interface that is part of that subnet. You can also get creative with this setting. For example, check an IP address on the management network (like a router or switch that is very near to the hosts) AND an IP on the storage network. This scenario would utilize both the management and storage networks for health checks. Keep in mind that all of the configured heuristic IP addresses must be accessible to qualify as a successful health check.