Platform1 min read
Updates without downtime and with automatic rollback
The system runs on two replicas and updates one at a time, rolling back by itself if the new version fails.
Create my accountHow it works now
The main service now runs on two replicas. In each update, one replica is swapped at a time, and the old one only stops when the new one is healthy. If the new version fails its health check, the system goes back to the previous one by itself.
What you gain
- Updates without taking the service down
- One replica at a time, with the old one kept until the new one is healthy
- Automatic rollback if something goes wrong
What changes for you
Less risk of instability when we ship improvements.
How it works
- Two replicas of the main service
- One replica is updated at a time, with a wait between them
- The old replica only leaves when the new one passes the health check
- If the new one fails, the previous one returns by itself