// System Design · 09
Redundancy
One instanceleads toDuplicate standing by
A spare that's ready the moment the primary fails.
How it fits together
- Load balancer uses / calls Node 1 (50% traffic)
- Load balancer uses / calls Node 2 (50% traffic)
- Load balancer uses / calls Primary (1 all traffic)
- Primary replicates / copies Standby (state sync)
- Load balancer uses / calls Standby (2 on failover)
Spot it
- A single node failure means an outage
- Failover has never actually been tested
- Backups exist but restores are untested
It costs
You pay for capacity you hope never to use, and a spare nobody tests is not a spare.
Go deeperThe full explanation and what to ask for · 1 min readExpandCollapse
Active-passive keeps a standby warm and switches on failure. Active-active spreads live traffic over all copies, so failure just means less capacity.
Redundant copies that share a power supply, region or config deploy fail together. Independence matters more than count.
Ask for it
- active-active
- active-passive
- N+1
- replication
- failover drill
Run this service active-active across two availability zones with at least N+1 instances, and add a failover test we can run on demand.