阅读中文版本

Automate Blue-Green DNS Cutover

Switch traffic between blue and green stacks by updating DNS with a low TTL and verifying health before and after the cutover.

# Set record TTL to 60s at least one TTL window before switching
example-dns update --name app.example.com --ttl 60

curl -fsS https://green.example.com/health || exit 1
example-dns update --name app.example.com --value green.example.com --ttl 60
for i in $(seq 1 10); do curl -fsS https://app.example.com/health && break; sleep 3; done

What this recipe does

Script a DNS-based blue-green cutover that flips traffic only after the new stack passes health checks.

Manual traffic switching is slow and error-prone, and reverting depends on remembering the previous DNS record.

Prerequisites

• DNS provider API access
• Two identical stacks (blue and green)
• Health endpoints on each stack

Procedure

1. Lower the TTL ahead of the cutover.

# Set record TTL to 60s at least one TTL window before switching
example-dns update --name app.example.com --ttl 60

Expected result: The short TTL ensures the switch propagates within a minute.

2. Verify the green stack then flip the record.

curl -fsS https://green.example.com/health || exit 1
example-dns update --name app.example.com --value green.example.com --ttl 60
for i in $(seq 1 10); do curl -fsS https://app.example.com/health && break; sleep 3; done

Expected result: Traffic reaches green and the health check passes end to end.

How to confirm it worked

Run the commands above and confirm every expected result in order. If a step does not produce the stated output, use the troubleshooting notes before continuing. A passing run shows the public hostname resolves to green, health checks pass, and rolling back restores blue within the TTL.

Troubleshooting

Old IPs still receive traffic — Wait for the TTL to drain and confirm existing connections complete before destroying blue.

Green healthy on its own but fails via DNS — Check TLS certificates and host headers for the public name.

Limitations

• DNS propagation is not instantaneous.
• Requires a rollback record prepared in advance.
• Long-lived connections may not drain quickly.