HCX General Questions

[virtualizingtheworld]

Hi,

When a Compute Profile configuration such as the Datastore for HCX Service Mesh Appliances is changed to another one and Service Mesh is synced, can you confirm it does not cause any changes and new datastore will apply for future HCX Upgrades/redeploy only?

Is storage vmotion of the Service Mesh appliances including the Network Extensions VMs fully supported and not expected to cause noticeable downtime? (guess in an NE HA setup best to migrate the Standby NE first and then manual failing over)

When you sync a Service Mesh do you get to see what changes will take place before you confirm/apply them?

Can you rename a Cluster, Distributed Switch, Network Profile Name which are all part of an existing Service Mesh...is all this supported and is any action needed after the rename or it will reflect automatically within HCX?

If you have multiple clusters with HCX deployed in them (same vCenter but different DVSwitches) and one of the networks is common between them which needs to be extended, am I correct in saying that it only needs to be extended once only?

Thanks

[Tom Zukowski]

If you change the Compute Profile and resync the Service Mesh, it will rebuild the existing appliances on the new datastore.

"expected to cause noticeable downtime" is relative.  vMotion of IX appliances is not recommended while migrations are in process.  vMotion of NE appliances works with minimal downtime depending on how much traffic is running over the L2 connection.  The process that you describe is reasonable.  If you migrate the HCX appliances to a different datastore than is defined in the Compute Profile, a Service Mesh resync will cause the HCX appliances to be reinstalled on the datastore defined in the Compute Profile.  

Sync or Resynch a Service Mesh will show changes before it applies them so you can cancel it.

Are you talking about renaming those resources in vCenter?  If you rename those objects, HCX will not know that they have been renamed and fail.  

Correct, you should only extend the same network/VLAN one time, otherwise you will create a network loop which would be detrimental.  HCX does not check or mitigate network loops.

Honestly, with the changes that you are describing it would be easier to create a new Service Mesh with new appliances in the correct datastore and then start new migrations over the new Service Mesh and unextend the port groups on the existing NEs and re-extend them over the new NEs.

[virtualizingtheworld]

Thanks Tom, can I storage migrate the hcx service mesh appliances and following the migration change the Compute Profile to reflect the new datastore (and resync the Servise Mesh).
Would doing it this way avoid appliances redeployment (downtime) as HCX will see that the appliances are there already?

[Tom Zukowski]

One of my peers tested this scenario in a lab and it provided successful results.

Changed the Compute profile to add a Datastore,
SVMotioned the NE appliance to another datastore... and host for that matter
Constant ping checked a stretched VM1-VM2 VM1-Onprem VM2 in VMC
Ping showed a max of 2000ms latency on the switch over, No ping drops,
Resynced the mesh and the mesh showed no configuration changes needed,

So in my "limited lab" the answer is, Change the compute profile to add the extra datastore, Perform SVMotion , then hit resync,
In my instance practically 0 downtime,

[virtualizingtheworld]

Thanks Tom