# Redundancy and Failover

Redundancy provides alternative components or capacity, while failover transfers service to those alternatives when a failure occurs.

- Type: framework
- Domain: Engineering & architecture
- Level: Intermediate
- Last reviewed: 2026-10-19
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/redundancy-and-failover/>)

## What it means

Redundancy means maintaining additional resources so a single component’s loss need not stop service. Failover is the detection and transition process that makes the alternative serve the workload. Active-active and active-passive arrangements differ in how spare capacity is used. Merely having replicas does not establish successful failover or protect against every shared failure.

## Why it matters

Failure handling depends on routing, state, capacity, and dependencies working together. Planning the transition helps teams avoid discovering during an incident that the spare system cannot accept real traffic.

## When to use it

Use redundancy for workloads whose availability requirements justify its cost. Match separation to the failure being addressed, such as an instance, zone, or region. Keep backups and recovery plans for destructive data changes that replicated infrastructure can reproduce everywhere.

## How it works

1. Identify relevant failure domains and shared dependencies.
2. Choose alternative capacity and a state-replication arrangement.
3. Define health signals, routing changes, and safeguards against conflicting active owners.
4. Verify the surviving capacity can support the required workload.
5. Exercise failover and return-to-normal behavior, including data reconciliation.

## A practical example

A service runs across two zones, but both instances depend on one stateful database. A failover exercise exposes that shared point of failure. The team adds a tested database recovery path and checks that either zone can handle the required traffic alone.

## Where it breaks

Redundancy increases cost and maintenance and may create correlated failures through shared configuration or credentials. Failover can interrupt sessions or lose recent updates. Multiple replicas are not independent backups when corruption propagates between them.

## Common mistakes

- Counting replicas without examining their shared failure domain.
- Provisioning a standby too small for the recovery workload.
- Testing the switch away from failure but never testing safe return.

## Sources

1. [Architecture strategies for designing for redundancy](<https://learn.microsoft.com/en-us/azure/well-architected/reliability/redundancy>) — Microsoft · Official documentation
2. [REL10-BP01 Deploy the workload to multiple locations](<https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/rel_fault_isolation_multiaz_region_system.html>) — Amazon Web Services · Official documentation

## Related topics

- [Load Balancer](<https://wiki.soroush.uk/topic/load-balancer/>)
- [Disaster Recovery](<https://wiki.soroush.uk/topic/disaster-recovery/>)
- [Chaos Engineering](<https://wiki.soroush.uk/topic/chaos-engineering/>)
