At least two replicas run
The service runs as two or more instances, so losing one does not take it down.
HighChecked automatically
- Applies to:
- Services on platforms where you set replicas and resource limits
- Rule id:
- deployment.min-two-replicas
Why it matters
One instance is a single point of failure, and it fails more often than people expect: every deploy, every node upgrade and every crash is downtime. Two instances turn each of those into a non-event, and are also what makes a rolling deploy possible.
How to apply it
- Run at least two replicas in production, on different nodes and preferably different zones.
- Size them so the service still meets its objective with one replica gone.
- Make the service safe to run twice: no in-memory state another request depends on, and scheduled jobs that take a lock.
- Set a disruption budget so maintenance never removes all replicas at once.
Example
deployment.replicas: 3 across three zones, with a disruption budget of at most one unavailable.
References
Checklists with this rule
Related agent skill: kubernetes-deployment