We have a usecase of running app leader election for our service that’s present in 3 AWS regions. I’m following the application leader election here: Application Leader Election with Sessions | Consul - HashiCorp Learn.
I setup a consul cluster with 3 servers in each Datacenter (AWS region here). I noticed when I write a key, its available with the region, but when I query it in a different region, it is not present. I imagine, replication is not done across datacenters, but shouldn’t the request be forwarded?
[Editt] Seems like it doesn’t replicate on its own from my reading. Does doing some manual replication here make sense or would I lose the app leader election mathematical correctness?
If the above is not a good strategy, Can I run 3 nodes in 3 different regions and use 1 data center for all of them?
[Edit] Seems like this works for me.
- Is there a better setup to do app leader election across 3 AWS regions?
- Does latency between regions affect HA in any way? Perhaps more false positive failures, but does it satisfy Raft’s correctness?
- What would my availability be with 3 nodes across regions? (Ideally want HA)