Failover Scenarios
Scenario 1: All Regions have healthy Edge Gateways
In this scenario, all regions are healthy, and each region contains one or more healthy Edge Gateways.
When a client seeks to resolve the API FQDN httpbin.tetrate.com:
httpbin.example.comis an alias for the global CNAMEedge-internal.global.example.comedge-internal.global.example.comis an alias for the regional CNAMEedge-internal.local.example.comedge-internal.local.example.comcontains the A records for the healthy Edge Gateways in that region
This scenario functions correctly in a 'split brain' scenario, because clients are only given the A records for the healthy Edge Gateways in their local region.
Scenario 2: A Partial Failure in one Region
In this scenario, an Edge Gateway fails in one region, but other Gateways continue to function in that region.
The Tetrate GSLB solution maintains the list of A records in each regional edge-internal.local.example.com record:
- If an Edge Gateway is unhealthy, its IP address is removed from the regional
edge-internal.local.example.comrecord - When an Edge Gateways recovers, or a new Edge Gateway is deployed, it IP addresses is added to the regional
edge-internal.local.example.comrecord
Scenario 3: A Regional Failure means an entire region is unavailable
In this scenario, an entire region fails. This may be provoked by a number of situations:
- All Edge Gateways in that region have failed
- All Tetrate GSLB controllers in that region have failed
- The Edge clusters or entire regional cloud have failed
In this case, all API FQDN queries should return the IP addresses of the healthy Edge Gateways in other regions:
- The Tetrate GSLB solution continually maintains a separate Failover Gateway name,
global-edge-internal.global.example.com, which contains the A records for all healthy Edge Gateways across all regions - In the event of a region failure, the Tetrate GSLB solution updates the Global Gateway name to be an alias for the Failover Gateway CNAME
On recovery of the failed region, the Global Gateway name is restored to be an alias for the Regional Gateway CNAME.
This approach ensures continuity in the event of a regional failure; all clients are directed to healthy Edge Gateways.