Projects › Data-Centre Disaster Recovery: Active/Standby Failover
Data-Centre Disaster Recovery: Active/Standby Failover
The order-management application runs in DC1, with a warm standby in DC2 on the same address (10.100.0.10), so branches never need to change anything when it fails over. Its databases replicate between DB-DC1 and DB-DC2 over a dedicated DC interconnect. Two branches and both data centres share a carrier metro-Ethernet service. All interfaces are addressed; there is no routing yet. Build it so branches use DC1, replication stays on the interconnect, and a silent loss of DC1 moves the branches to DC2 within the 5-second recovery objective.
Difficulty: expert · Time: about 75 minutes
Why it matters
Disaster recovery is judged by numbers agreed with the business: the recovery time objective. Advertising the same application prefix from both sites with the standby at a worse cost is a common way to fail over without changing DNS or clients. The catch is that over a carrier Ethernet service a dead router does not drop the link, so failure is only noticed when the routing protocol's dead timer expires: 40 seconds by default for OSPF. Meeting a 5-second objective means tuning for it and measuring it.
Technologies
- OSPFv2
- OSPF Timers
- Disaster Recovery
- Failover Testing
- Data Center Design
Related projects
- Data Center Spine-Leaf Fabric with VXLAN and BGP EVPN
- Two-Site Data Center Interconnect with EVPN Multi-Site
- Underlay Multicast for a VXLAN Fabric: PIM, Anycast RP and VNI Groups
- Arista eBGP Underlay: Two Leaves Through a Spine
- Arista EOS Top-of-Rack VLANs for a vSphere Cluster
- Arista EOS Routed Interfaces and Single-Area OSPF
- Arista Top-of-Rack Static Routing to the Distribution Router
- Arista Three-Router OSPF Underlay for a Colocation Cage