patterns / 3 · after dual-AZ
Multi-region
One caster + NLB per region. Failover is DNS — keep the regional NLBs.
examples/relay_iport_ha_mr— own state. Destroy dual-AZ first.
Deploy
| Need | Detail |
|---|---|
| Regions | AKL (ap-southeast-6) + SYD (ap-southeast-2) |
| Zone | On both VPCs (often SYD-owned) |
| Profiles | primary_aws_profile + secondary_aws_profile |
| IPort | Reachable from both regions |
aws sso login --profile <akl-profile>
aws sso login --profile <syd-profile>
cd examples/relay_iport_ha_mr
cp terraform.tfvars.example terraform.tfvars
terraform init && terraform apply
terraform output -raw ntrip_vip_fqdn # dial :2101
Explain
Route53 failover aliases (evaluate_target_health). Delay is mostly DNS TTL —
not “missing” an NLB.
cd examples/relay_iport_ha_mr
./scripts/failover.sh status
./scripts/failover.sh drain-primary
./scripts/failover.sh restore-primary
1–3+ minutes typical. Longer than dual-AZ because of DNS.
Teardown
cd examples/relay_iport_ha_mr
aws sso login --profile <akl-profile>
aws sso login --profile <syd-profile>
terraform destroy