C-series server on DR site

[rasmusan1]

Hello

I have a customer running Cisco UCS with blade chassis and 6248 FIs in primary datacenter.

now customer want to setup a DR host in a secondary site, where they want to place DR VMs (DC, SQL AlwaysOn replica, etc.).

My question is, that if customer buys C-series server and integrate with UCS Manager on primary site, using dedicated fibre cabling from C-series VIC to FI's, to utilize high-speed links to FI's as primary, is it then possible at the same time to use the 2 built-in 1gbit/s NICs? (my thought is a vSwitch with VIC NICs as active and the 1gbit/s as standby)

And what happens if DR server looses contact with UCS Manager - will it continue to run "standalone" then?

[Walter Dey]

This could be done ! however, it is not recommended, nor price efficient. I've never seen such a setup. 

Just build a second UCS domain at the secondary site ! If your storage (or shared storage for vSphere has no DR) has no DR, what's the benefit of a DR ?

Then you could make use of global pools and global service profiles.

[rasmusan1]

well, secondary site is within a couple 100 meters from primary site... on secondary site only very limited amount of virtual servers is needed thus separate USC domain/FIs will be too expensive.

Server at DR site will not use shared storage - local disks will be utilized as storage.

They want to use the VIC links as primary for replication etc., and then in the event of primary site being unavailable, fail over to using the 1gbit/s nics embedded in the server, using standard vswitch nic ordering in vsphere.

[Walter Dey]

A couple of 100m ?? I would not call this a DR site ? 

If the primary site is down, UCS FI are down, and the remote servers are only managable by FI ? you end up in a mess !

Trust me, this setup is a absolute nightmare. The motivation is just saving cost; 

[rasmusan1]

sure the motivation is saving cost - but for a customer that is only SMB size, that will always be an issue... so instead of thinking enterprise setup and blaming customer to be cheap, please instead focus on my question and helping this customer.

So, as I understand, you think they will be better off with the C-series server just managed as standalone, and connecting to the network by using standard ethernet?

[Kirk J]

Greetings.

When a UCS rack server is directly connected to the FIs via the 10Gb CNA VICs, the onboard LOM ports are disabled.  You would need to add one of the available Broadcom or Intel PCI-E nics listed on the spec sheet for the server, and would then be able to manually configure those NIC ports from the OS. Those ports would need to be connected to another switch besides the FIs (which is what it sounds like what you are wanting to do).

Is your intention to have limited additional rack servers at the "dr site" that are remotely syncing VMs/workloads, and would continue to fully function if your primary site (that includes the UCSM, FIs, Servers) all goes down, power outage, tornado sucks them up, etc, etc?

Assuming the $$ requirements to setup two UCSM domains isn't feasible, would it be possible to setup additional Rack servers with 10Gb connectivity at the DR site in standalone mode, that could carry on with the minimally required replicated VMs?  This assumes you have some kind of 10Gb switch connected at the 'dr site' that connects to the uplinks on the FIs.  You could still probably do this via the addin 1Gb cards in the UCSM integrated rack servers, connected to 1Gb upstream switch that connected to the 'dr site' rack servers via 1Gb, but this would obviously take longer for syncs/replication to happen.

Thanks,

Kirk