Model failure
Choose the scenario, affected systems, decision owner and recovery acceptance checks.
Disaster recovery
A recovery plan must survive an exercise.
Disaster-recovery architecture, runbooks and exercises for systems with defined recovery targets and owners.
Exercise types
Pick the exercise that proves the plan.
Recovery targets become honest only after an exercise. Three formats, increasing in realism and risk, each selected and approved before work begins.
ExerciseWhat it provesProduction riskEvidence produced
Tabletop walkthroughOwners, decision points and gaps in the runbookNoneReviewed runbook and action list
Isolated restoreData comes back complete on real destination capacityNone, runs beside productionDated restore result with measured timings
Controlled failoverThe full path: traffic, data, dependencies and health checksManaged, with approved rollbackMeasured recovery result against the agreed targets
Recovery targets
Recovery point and recovery time are measured in the exercise, then contracted. A target that has never survived an exercise is an intention, not a capability.
See backup evidence firstRecovery process.
Prepare the path
Confirm data, destination, network, access, dependencies, capacity and runbook.
Exercise and learn
Run the controlled scenario, measure results and close the discovered gaps.
Common questions.
Does a replicated server provide disaster recovery?
Not by itself. Recovery still needs consistent data, dependencies, traffic switching, access, capacity, health checks and an owner.
Will the exercise cause downtime?
The method and risk depend on the system. A tabletop, isolated restore or controlled failover is selected and approved before work begins.