- Integrated Systems
- About Us
- Integrated Systems
- About Us
Dec 15, 4:00 am to 10:00 am UTC
Dec 14, 10:00 pm CST to Dec 15, 4:00 am CST
Dec 14, 8:00 pm PST to Dec 15, 2:00 am PST
02-29-2012 07:59 AM
Virtual Connect Stacking High-Availability questions
Hoa had a VC stacking question:
Two C7000 stacked in vNet configuration. Uplinks came from Bay 1 and Bay 2 of the primary enclosure, where VCM running from. This configuration provides no place for VCM to migrate and providing network supports should you shut off power for maintenance purposes on the 1st enclosure. To correct that, can you take uplinks from 1st enclosure Bay 2 and move that to 2nd enclosure Bay 2, make necessary changes on VCM (remove/add uplinks) and call it done? I know that I can add/remove uplinks on a single enclosure domain at will but add/remove uplinks from two enclosures stack may be not that simple. My fear is that one must have uplinks 1st enclosure Bay 1 with 2nd enclosure Bay 2 in place before creating VC domain to call it truly in HA mode. Can a quick dirty uplinks move and editing gain you HA in a pinch? Or delete and create new domain a must?
Input from Dan:
Pretty sure you can adjust on the fly in stacked domain just fine.
Uplink changes have never given me any trouble changing on the fly before.
Changing a vNet from Direct to SUS or vice versa on the other hand is a Delete and Re-Add process.
You can always try it in the simulator.
PS: Make sure you ADD the new ports first and then remove the old ports. If you remove all ports from an SUS, any vNets associated with that SUS using SmartLink will have SmartLink automatically turned off. This is currently expected behavior but I have already requested a feature change to make sure this doesn’t happen until you actually hit the Apply button (what most customers would expect).
Response from Hoa:
No disagreement in changing uplink for single enclosure VC. The challenge is can I modify VC created with two enclosure stack without HA into HA support. Specifically, how can we change VCM to perform force fail-over within A enclosure into between enclosures.
A vendor VC setup gig went terribly wrong. Blades BfS and clustered within two enclosures. From an outage we “discover” issues now we’re responsible for the fix. Again, in addition to VCM in stack fail over can happen only on primary enclosure, domain uplinks are in the primary one as well. By powering off primary enclosure effectively kill everything in between, can’t happen again. The challenge: 1) VCM must run in HA mode. 2) Spread uplinks to both enclosure. 3) Keep production running, downtime at a minimal with no open heart surgery. Domain is not in A/A, A/S, just active with network on odd side, even side a test network going nowhere -The person setup the gig said that was for testing LACP on vNET! Now you see my point?
My bottom line question is can I modify current domain or must create new one? From VC Cookbook stack scenarios, two enclosure uplinks spread at VC deployment time set the stage for HA. As you can see it’s not as simple as single enclosure uplinks reshuffling. By rearranging uplinks in a stack configuration does that mean I now can have fail-over VCM between enclosures? Normally uplinks are independent from the server facing side but in stack/HA uplinks seem to play an additional role with VC configuration network, VCM. I am not sure for HA to work it’s a one shot at the design stage or else. Hope to see my logic is wrong again.
And from Vincent:
What Dan said applies equally to multi-enclosure stacks. You can change uplinks on the fly, no problem.
Which uplinks are active is entirely independent of which module is running VCM for management. Only the 2 upper-most modules in the first enclosure (where the domain was created) can run VCM. You cannot failover VCM to modules in another enclosure. If that enclosure is powered off for whatever reason, you will lose the ability to manage the configuration. The servers however will retain connectivity assuming your VC networks have uplinks in the other enclosures. You will need to restore the first enclosure before you can access VCM again.
Other comments or input to this discussion?