[{"data":1,"prerenderedAt":575},["ShallowReactive",2],{"blog-en-aws-backups-recovery-test-acceptance":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"language":10,"slug":11,"translationKey":12,"date":13,"author":14,"image":15,"tags":16,"body":21,"_type":569,"_id":570,"_source":571,"_file":572,"_stem":573,"_extension":574},"/en/blog/2026-10-28-aws-backups-recovery-test-acceptance","blog",false,"","AWS backups: what to require from a recovery test","How to agree on RTO, RPO, scope and evidence for an AWS recovery test before investing in more business continuity infrastructure.","en","aws-backups-recovery-test-acceptance","aws-backup-recovery-test-acceptance","2026-10-28","Alejandro Rodríguez","/og-image.png",[17,18,19,20],"aws","backups","disaster-recovery","cloud-reliability",{"type":22,"children":23,"toc":557},"root",[24,32,37,42,49,54,70,75,90,95,109,115,120,225,239,252,258,263,268,301,314,319,324,330,335,347,352,366,371,377,389,394,424,430,435,448,453,458,464,469,503,508,514,528,533,538,544],{"type":25,"tag":26,"props":27,"children":28},"element","p",{},[29],{"type":30,"value":31},"text","Having copies in AWS does not answer the question that matters during an outage: which part of the business you can recover, with what data and in how much time. Nor does a correctly completed restore job answer it on its own.",{"type":25,"tag":26,"props":33,"children":34},{},[35],{"type":30,"value":36},"Before commissioning a recovery test, agree on its acceptance criteria: scenario, critical flow, objectives, isolation and evidence. The deliverable should let you decide whether what was tested is sufficient, what needs fixing, or whether you need another continuity strategy, rather than being limited to a screenshot with a green status.",{"type":25,"tag":26,"props":38,"children":39},{},[40],{"type":30,"value":41},"This guide proposes contracting and acceptance criteria; it does not present results from a real test or guaranteed times.",{"type":25,"tag":43,"props":44,"children":46},"h2",{"id":45},"first-pin-down-what-recovery-means",[47],{"type":30,"value":48},"First, pin down what recovery means",{"type":25,"tag":26,"props":50,"children":51},{},[52],{"type":30,"value":53},"There are three distinct levels: a copy exists, a resource has been restored, and the service is usable again. Moving from the second to the third may require rebuilding infrastructure, deploying the application, recovering configuration and connecting dependencies.",{"type":25,"tag":26,"props":55,"children":56},{},[57,59,68],{"type":30,"value":58},"AWS considers it an antipattern to ",{"type":25,"tag":60,"props":61,"children":65},"a",{"href":62,"rel":63},"https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/rel_backing_up_data_periodic_recovery_testing_data.html",[64],"nofollow",[66],{"type":30,"value":67},"restore a copy without querying or recovering data to verify its use",{"type":30,"value":69},". For the buyer, the consequence is concrete: the scope must include checks on data and on the agreed flow, not just the AWS operation.",{"type":25,"tag":26,"props":71,"children":72},{},[73],{"type":30,"value":74},"Define two objectives with the business and technology teams:",{"type":25,"tag":76,"props":77,"children":78},"ul",{},[79,85],{"type":25,"tag":80,"props":81,"children":82},"li",{},[83],{"type":30,"value":84},"RTO (recovery time objective): maximum acceptable outage time until the service is recovered.",{"type":25,"tag":80,"props":86,"children":87},{},[88],{"type":30,"value":89},"RPO (recovery point objective): maximum tolerable data loss, expressed as the interval between the last recoverable data point and the outage.",{"type":25,"tag":26,"props":91,"children":92},{},[93],{"type":30,"value":94},"They are objectives, not demonstrated capabilities or an implied SLA. Retention says how long you keep copies; it does not prove what loss you can avoid. After corruption, moreover, the most recent point may already contain the damage: you need to identify the last valid state.",{"type":25,"tag":26,"props":96,"children":97},{},[98,100,107],{"type":30,"value":99},"The ",{"type":25,"tag":60,"props":101,"children":104},{"href":102,"rel":103},"https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/rel_planning_for_recovery_objective_defined_recovery.html",[64],[105],{"type":30,"value":106},"AWS recovery objectives guide",{"type":30,"value":108}," asks you to consider impact and dependencies. An objective for an application is not useful if its identity system or data source cannot meet it too. In a Spanish SME or SaaS business, also review specific commitments to customers; do not infer universal obligations from company size or sector.",{"type":25,"tag":43,"props":110,"children":112},{"id":111},"what-to-fund-depending-on-the-starting-point",[113],{"type":30,"value":114},"What to fund depending on the starting point",{"type":25,"tag":26,"props":116,"children":117},{},[118],{"type":30,"value":119},"This is a decision proposal, not an AWS classification or a guarantee of timelines:",{"type":25,"tag":121,"props":122,"children":123},"table",{},[124,148],{"type":25,"tag":125,"props":126,"children":127},"thead",{},[128],{"type":25,"tag":129,"props":130,"children":131},"tr",{},[132,138,143],{"type":25,"tag":133,"props":134,"children":135},"th",{},[136],{"type":30,"value":137},"Situation",{"type":25,"tag":133,"props":139,"children":140},{},[141],{"type":30,"value":142},"Reasonable next investment",{"type":25,"tag":133,"props":144,"children":145},{},[146],{"type":30,"value":147},"Evidence or limit to require",{"type":25,"tag":149,"props":150,"children":151},"tbody",{},[152,171,189,207],{"type":25,"tag":129,"props":153,"children":154},{},[155,161,166],{"type":25,"tag":156,"props":157,"children":158},"td",{},[159],{"type":30,"value":160},"No agreed RTO or RPO",{"type":25,"tag":156,"props":162,"children":163},{},[164],{"type":30,"value":165},"Define objectives per flow and incident",{"type":25,"tag":156,"props":167,"children":168},{},[169],{"type":30,"value":170},"Impact, dependencies and owners; still no demonstrated capability",{"type":25,"tag":129,"props":172,"children":173},{},[174,179,184],{"type":25,"tag":156,"props":175,"children":176},{},[177],{"type":30,"value":178},"There are copies, but no validated restore",{"type":25,"tag":156,"props":180,"children":181},{},[182],{"type":30,"value":183},"Scoped test with prior acceptance",{"type":25,"tag":156,"props":185,"children":186},{},[187],{"type":30,"value":188},"Data, flow and times verified under declared conditions",{"type":25,"tag":129,"props":190,"children":191},{},[192,197,202],{"type":25,"tag":156,"props":193,"children":194},{},[195],{"type":30,"value":196},"The test does not meet the objective",{"type":25,"tag":156,"props":198,"children":199},{},[200],{"type":30,"value":201},"Fix the blocker and repeat, or review the strategy",{"type":25,"tag":156,"props":203,"children":204},{},[205],{"type":30,"value":206},"Recorded cause and new measurement; closing a task is not enough",{"type":25,"tag":129,"props":208,"children":209},{},[210,215,220],{"type":25,"tag":156,"props":211,"children":212},{},[213],{"type":30,"value":214},"You must withstand the loss of a region",{"type":25,"tag":156,"props":216,"children":217},{},[218],{"type":30,"value":219},"Evaluate and rehearse recovery from the planned destination",{"type":25,"tag":156,"props":221,"children":222},{},[223],{"type":30,"value":224},"A rehearsal in the original region does not prove that coverage",{"type":25,"tag":26,"props":226,"children":227},{},[228,230,237],{"type":30,"value":229},"There is no need to decide upfront that everything must be multi-region. Nor is it wise to defer a known regional requirement with a local test that does not cover it. AWS frames the ",{"type":25,"tag":60,"props":231,"children":234},{"href":232,"rel":233},"https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/rel_planning_for_recovery_disaster_recovery.html",[64],[235],{"type":30,"value":236},"strategy choice",{"type":30,"value":238}," as a trade-off between objectives, cost and complexity.",{"type":25,"tag":26,"props":240,"children":241},{},[242,244,250],{"type":30,"value":243},"If even the dependencies are unclear, start with an ",{"type":25,"tag":60,"props":245,"children":247},{"href":246},"/blog/que-revisaria-en-las-primeras-48-horas-al-entrar-en-una-cuenta-aws-heredada",[248],{"type":30,"value":249},"inventory of the inherited AWS account",{"type":30,"value":251}," (in Spanish). Without that baseline you may test a database and leave out the component that prevents its use.",{"type":25,"tag":43,"props":253,"children":255},{"id":254},"the-scope-that-must-be-written-down",[256],{"type":30,"value":257},"The scope that must be written down",{"type":25,"tag":26,"props":259,"children":260},{},[261],{"type":30,"value":262},"Choose a flow recognizable to the business and describe the incident: logical deletion, lost resource, or account or region unavailability are not equivalent scenarios. Also document what you will not simulate.",{"type":25,"tag":26,"props":264,"children":265},{},[266],{"type":30,"value":267},"Before running, agree on:",{"type":25,"tag":76,"props":269,"children":270},{},[271,276,281,286,291,296],{"type":25,"tag":80,"props":272,"children":273},{},[274],{"type":30,"value":275},"Which data and recovery points will be used, where they are and how a valid state will be chosen.",{"type":25,"tag":80,"props":277,"children":278},{},[279],{"type":30,"value":280},"Destination account and region, permissions, KMS keys, network and required capacity.",{"type":25,"tag":80,"props":282,"children":283},{},[284],{"type":30,"value":285},"Application version, artifacts, configuration and secrets the flow needs.",{"type":25,"tag":80,"props":287,"children":288},{},[289],{"type":30,"value":290},"Identity, DNS, queue and external provider dependencies, and which will be simulated.",{"type":25,"tag":80,"props":292,"children":293},{},[294],{"type":30,"value":295},"Owner of execution, person accepting functional validation, and authority to stop the rehearsal.",{"type":25,"tag":80,"props":297,"children":298},{},[299],{"type":30,"value":300},"Budget, permitted resources, stop conditions and owner of cleanup.",{"type":25,"tag":26,"props":302,"children":303},{},[304,306,312],{"type":30,"value":305},"Owning an encrypted copy is not enough: you need operational access to its keys and permissions in the planned scenario. A ",{"type":25,"tag":60,"props":307,"children":309},{"href":308},"/en/blog/separate-production-development-aws-accounts",[310],{"type":30,"value":311},"separate account",{"type":30,"value":313}," can help isolate the test, but it does not move copies or automatically guarantee recoverability.",{"type":25,"tag":26,"props":315,"children":316},{},[317],{"type":30,"value":318},"Before starting the recovered application, control its effects: emails, payments, webhooks, scheduled tasks and writes toward production. Block or simulate the integrations that could cause real effects. Stop the test if isolation fails, unplanned writes appear, or agreed limits are exceeded; record the result as incomplete.",{"type":25,"tag":26,"props":320,"children":321},{},[322],{"type":30,"value":323},"A restore must not by default become a copy of personal data for development. If you need real data, agree on access, minimization, retention and the applicable review of its processing. This does not certify GDPR compliance or replace legal advice.",{"type":25,"tag":43,"props":325,"children":327},{"id":326},"conceptual-example-restoring-an-rds-instance",[328],{"type":30,"value":329},"Conceptual example: restoring an RDS instance",{"type":25,"tag":26,"props":331,"children":332},{},[333],{"type":30,"value":334},"Suppose the flow depends on an RDS database instance. It is not a customer case or a measurement; it serves to make concrete what you would accept.",{"type":25,"tag":26,"props":336,"children":337},{},[338,345],{"type":25,"tag":60,"props":339,"children":342},{"href":340,"rel":341},"https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_PIT.html",[64],[343],{"type":30,"value":344},"Point-in-time recovery of an RDS DB instance",{"type":30,"value":346}," creates another instance without modifying the original. It does not rewind the production database in place. The eligible point depends on retention and the latest restorable time available.",{"type":25,"tag":26,"props":348,"children":349},{},[350],{"type":30,"value":351},"In the test, review the security groups, parameter and option groups applicable to the engine, plus access to keys. Then connect an isolated application to the new instance with compatible configuration. Validate queries and a representative flow; for example, creating and querying a test transaction without sending real payments or notifications.",{"type":25,"tag":26,"props":353,"children":354},{},[355,357,364],{"type":30,"value":356},"Do not accept only the ",{"type":25,"tag":358,"props":359,"children":361},"code",{"className":360},[],[362],{"type":30,"value":363},"available",{"type":30,"value":365}," state. AWS explains that the instance may be operational while its volumes keep loading blocks in the background, with performance not yet at full. Validation must use agreed conditions and response criteria. Without representative load, you have not demonstrated capacity to serve production.",{"type":25,"tag":26,"props":367,"children":368},{},[369],{"type":30,"value":370},"This example is limited to restoring a DB instance: Aurora has its own documentation and details vary by engine. Nor does it demonstrate consistency between the database, objects and other systems recovered from independent points.",{"type":25,"tag":43,"props":372,"children":374},{"id":373},"what-aws-backup-automates-and-what-you-must-contribute",[375],{"type":30,"value":376},"What AWS Backup automates and what you must contribute",{"type":25,"tag":26,"props":378,"children":379},{},[380,387],{"type":25,"tag":60,"props":381,"children":384},{"href":382,"rel":383},"https://docs.aws.amazon.com/aws-backup/latest/devguide/restore-testing.html",[64],[385],{"type":30,"value":386},"Restore testing in AWS Backup",{"type":30,"value":388}," lets you schedule restores for supported resources, select recovery points and record job duration. It also manages later deletion of test resources.",{"type":25,"tag":26,"props":390,"children":391},{},[392],{"type":30,"value":393},"Before choosing it, check compatibility and restrictions for the specific resource, region and encryption. Being able to back up does not imply every test mode is available. Review inferred parameters: that a restore starts does not mean its network or configuration is safe for this test.",{"type":25,"tag":26,"props":395,"children":396},{},[397,399,414,416,422],{"type":30,"value":398},"Business verification remains the team's work. AWS allows adding ",{"type":25,"tag":60,"props":400,"children":403},{"href":401,"rel":402},"https://docs.aws.amazon.com/aws-backup/latest/devguide/restore-testing-validation.html",[64],[404,406,412],{"type":30,"value":405},"custom validation after the ",{"type":25,"tag":358,"props":407,"children":409},{"className":408},[],[410],{"type":30,"value":411},"COMPLETED",{"type":30,"value":413}," event",{"type":30,"value":415},", via EventBridge and a target such as Lambda, and recording the result with ",{"type":25,"tag":358,"props":417,"children":419},{"className":418},[],[420],{"type":30,"value":421},"PutRestoreValidationResult",{"type":30,"value":423},". The logic must check what was agreed and reflect failures; not assign success in advance. The documentation template contains to-be-implemented tasks, not functional validation ready to accept.",{"type":25,"tag":43,"props":425,"children":427},{"id":426},"measure-until-the-flow-is-validated",[428],{"type":30,"value":429},"Measure until the flow is validated",{"type":25,"tag":26,"props":431,"children":432},{},[433],{"type":30,"value":434},"Distinguish three measures: restore job duration, test duration, and service recovery time. They are not interchangeable.",{"type":25,"tag":26,"props":436,"children":437},{},[438,439,446],{"type":30,"value":99},{"type":25,"tag":60,"props":440,"children":443},{"href":441,"rel":442},"https://aws.amazon.com/blogs/architecture/disaster-recovery-dr-architecture-on-aws-part-ii-backup-and-restore-with-rapid-recovery/",[64],[444],{"type":30,"value":445},"AWS recovery analysis",{"type":30,"value":447}," includes detection and decision, plus rebuilding, integrating and redirecting traffic. If you start the clock when launching the restore, you leave out part of the time since the outage. If you prepare infrastructure beforehand, you must also declare it.",{"type":25,"tag":26,"props":449,"children":450},{},[451],{"type":30,"value":452},"Ask for a timeline with start, technical restore and end of validation, identifying the steps not tested. A partial result may be useful, but it does not let you claim you meet the full RTO.",{"type":25,"tag":26,"props":454,"children":455},{},[456],{"type":30,"value":457},"For RPO, identify the recovered point and verify records or verifiable items against the simulated scenario. Add integrity and consistency checks appropriate to the data: a ping does not replace them. The date of the last backup, without verifying its usable content, is not enough either.",{"type":25,"tag":43,"props":459,"children":461},{"id":460},"what-the-deliverable-should-contain",[462],{"type":30,"value":463},"What the deliverable should contain",{"type":25,"tag":26,"props":465,"children":466},{},[467],{"type":30,"value":468},"Propose these criteria before commissioning the work, both to a vendor and to the internal team:",{"type":25,"tag":470,"props":471,"children":472},"ol",{},[473,478,483,488,493,498],{"type":25,"tag":80,"props":474,"children":475},{},[476],{"type":30,"value":477},"Agreed scenario, flow, RTO and RPO, with execution and acceptance owners.",{"type":25,"tag":80,"props":479,"children":480},{},[481],{"type":30,"value":482},"Identification of the recovered point, destination, configuration and test conditions, without exposing secrets or personal data in the report.",{"type":25,"tag":80,"props":484,"children":485},{},[486],{"type":30,"value":487},"Timeline and results of data and flow checks, with traceable evidence.",{"type":25,"tag":80,"props":489,"children":490},{},[491],{"type":30,"value":492},"Comparison against objectives, separating what was demonstrated from what was excluded: detection, traffic, third parties or real load, as applicable.",{"type":25,"tag":80,"props":494,"children":495},{},[496],{"type":30,"value":497},"Reproducible recovery procedure (runbook), necessary access and handover to whoever will operate the recovery.",{"type":25,"tag":80,"props":499,"children":500},{},[501],{"type":30,"value":502},"Incidents, observed cost, verified cleanup and next decision: accept for that scenario, fix and repeat, or widen scope.",{"type":25,"tag":26,"props":504,"children":505},{},[506],{"type":30,"value":507},"If functional validation fails, do not accept overall success because the resource was restored. If a permission fails, the finding is useful, but recovery remains undemonstrated. And if everything passes, the conclusion is still limited to the load, scenario and conditions tested.",{"type":25,"tag":43,"props":509,"children":511},{"id":510},"cost-and-closure-do-not-leave-the-test-open",[512],{"type":30,"value":513},"Cost and closure: do not leave the test open",{"type":25,"tag":26,"props":515,"children":516},{},[517,519,526],{"type":30,"value":518},"Budget preparation, validation, duplicated resources and usage time, not just the copy. ",{"type":25,"tag":60,"props":520,"children":523},{"href":521,"rel":522},"https://aws.amazon.com/backup/pricing/",[64],[524],{"type":30,"value":525},"AWS Backup pricing",{"type":30,"value":527}," covers restore testing evaluations and restored data by resource, with possible additional charges for retained resources or requests. Without region, volume and duration there is no defensible estimate for your company.",{"type":25,"tag":26,"props":529,"children":530},{},[531],{"type":30,"value":532},"Restore testing cleanup is not instant. Verify its result and pending resources; for those you created outside that mechanism, also assign an owner. Delete only authorized test resources and keep original copies and evidence per the agreed retention.",{"type":25,"tag":26,"props":534,"children":535},{},[536],{"type":30,"value":537},"Repeat tests with a frequency tied to criticality and relevant changes. There is no universal frequency that replaces that analysis. Nor does high availability replace backup: it can replicate a deletion or corruption. Multi-region adds coverage only if the necessary access, data and dependencies are designed and tested.",{"type":25,"tag":43,"props":539,"children":541},{"id":540},"agreeing-on-the-next-step",[542],{"type":30,"value":543},"Agreeing on the next step",{"type":25,"tag":26,"props":545,"children":546},{},[547,549,555],{"type":30,"value":548},"If you have copies but cannot describe a validated recovery, you can request a free call from AstralDeploy's ",{"type":25,"tag":60,"props":550,"children":552},{"href":551},"/en/services/cloud-reliability",[553],{"type":30,"value":554},"cloud reliability service",{"type":30,"value":556},". The conversation serves to review what is covered and agree whether a scoped test with acceptance criteria makes sense. Scope and access are defined afterwards; it implies no guaranteed recovery, 24/7 support, or free full audit.",{"title":7,"searchDepth":558,"depth":558,"links":559},2,[560,561,562,563,564,565,566,567,568],{"id":45,"depth":558,"text":48},{"id":111,"depth":558,"text":114},{"id":254,"depth":558,"text":257},{"id":326,"depth":558,"text":329},{"id":373,"depth":558,"text":376},{"id":426,"depth":558,"text":429},{"id":460,"depth":558,"text":463},{"id":510,"depth":558,"text":513},{"id":540,"depth":558,"text":543},"markdown","content:en:blog:2026-10-28-aws-backups-recovery-test-acceptance.md","content","en/blog/2026-10-28-aws-backups-recovery-test-acceptance.md","en/blog/2026-10-28-aws-backups-recovery-test-acceptance","md",1795108324752]