vSphere 6U2, SRM 6.1 and NSX 6.2 - Datacenter Failover DR

[Thomas_]

Hi guys,

I'd like to get some high level input on what to consider when planning DR.

My environment looks like this:

Site A - DC1:

IP Subnet: 10.1.0.0/16

vCenter with emb. PSC, ext. SQL DB and SRM

3 ESX hosts registered to vCenter in DC1

Clients get IPs by local DHCP Server

Site A - DC2 (DR Site):

IP Subnet: ?

vCenter with emb. PSC, ext. SQL DB and SRM

3 ESX hosts registered to vCenter in DC2

Site B - DC3:

IP Subnet: 10.2.0.0/16

3 ESX Hosts registered to vCenter in DC1

Clients get IPs by local DHCP Server

Most applications/devices rely on DNS but a few are configured to use IPs

Site B - DC4:

IP Subnet: ?, not sure yet

no ESX Hosts, just fiber switches to connect to the various client switches in DR scenario and dedicated internet connection for ipsec to the DR site

DC1 and DC3 are connected via MPLS.

DC1 and DC2 are in the same Site but different buildings.

DC3 and DC4 are in the same Site but different buildings.

DC4 and DC2 can be connected with a dedicated internet connection via IPsec.

Assuming DC3 goes down and DC2 goes live.

I read SRM in combination with NSX can help keeping the primary IP subnet + IPs the same at the DR site.

Are VXLAN compatible switches/routers required on both sites? OSPF is a must I guess?

Lets say all my clients in the primary Site B are still up and running, just the DC went down.

The surviving clients in Site B will still have the same IP subnet 10.2.0.0/16, same as now the DR Site has - how does routing work, what subnets should I specify at DC2/DC4?

Thanks for a high level explanation!

[chuckbell]

I read SRM in combination with NSX can help keeping the primary IP subnet + IPs the same at the DR site.

That is correct. Cross VC NSX can support a universal logical switch so that the workloads can have the same IP, GW and FW policy that they have at primary site.

Are VXLAN compatible switches/routers required on both sites? OSPF is a must I guess?

VXLAN switches/routers are not required if you are asking about the physical sw/rtr. But depending on how you design is, you will need to be able to transport vxlan over the MPLS which requires a 1600 byte MTU. Routing is not a must but usually preferred to separate your failure domains between DC's. L2 can routed over L3 w/VXLAN, butVXLAN can also work over L2.

Lets say all my clients in the primary Site B are still up and running, just the DC went down.

The surviving clients in Site B will still have the same IP subnet 10.2.0.0/16, same as now the DR Site has - how does routing work, what subnets should I specify at DC2/DC4?

Not sure I follow the failure scenario but there are multiple ways to design cross-VC NSX. In general routing can be locally egress the sites perimeter or tunnel back to the primary site. It all depends on the use case.

Check out the newly updated DR guide on the community document section posted a few days ago. It may help add some color to my repsponses.

Disaster Recovery with NSX and SRM