NODES AND SCHEDULER
A server went down? Your app won’t even notice.
The scheduler spreads replicas across different servers.
Replicas on different nodesAutomatic failoverAMD EPYC servers
scheduler · load balancer
replica-1us-1 · healthy
replica-2us-2 · healthy
replica-3recreating on us-3
load balancerrouting
HOW IT WORKS
Automatic distribution and recovery.
01DistributesReplicas of the same service land on different servers.
02MonitorsContinuous health checks on every instance.
03RecoversAn instance went down: it leaves the load balancer and is recreated on another node.
Availability without an on-call team.
✓No single point of failure for your app✓Modern hardware✓Isolation✓Transparent
WHAT EACH PLAN INCLUDES
- Distribution
- replicas spread across nodes
- Detection
- health check
- Recovery
- automatic recreation
- Hardware
- AMD EPYC (migration in progress)
- Network
- private between services
with only 1 replica, a node failure causes brief downtime until it’s recreated. Use 2 or more for production.
Frequently asked questions
Do I need to configure anything?
No. Just run 2 or more replicas.
Does the database fail over too?
Not automatically; app replicas don’t make the database highly available.
Can I choose the server?
There’s no need; the scheduler decides.