[{"data":1,"prerenderedAt":492},["ShallowReactive",2],{"blog-en-what-to-review-in-the-first-48-hours-of-an-inherited-aws-account":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"slug":11,"author":12,"image":13,"tags":14,"language":20,"format":21,"translationKey":22,"body":23,"_type":486,"_id":487,"_source":488,"_file":489,"_stem":490,"_extension":491},"/en/blog/2026-06-18-what-to-review-in-the-first-48-hours-of-an-inherited-aws-account","blog",false,"","What I would review in the first 48 hours of an inherited AWS account","A safe, senior-level approach to prioritising IAM, costs, public exposure, backups, Terraform, observability and production dependencies.","2026-06-18","what-to-review-in-the-first-48-hours-of-an-inherited-aws-account","Alejandro Rodríguez","/og-image.png",[15,16,17,18,19],"aws","audit","security","terraform","reliability","en","audit-lesson","inherited-aws-first-48-hours",{"type":24,"children":25,"toc":472},"root",[26,34,40,45,50,55,62,67,72,77,82,87,93,98,103,108,113,119,124,129,134,139,144,150,155,160,165,170,175,181,186,191,196,201,206,212,217,222,227,232,237,243,248,253,258,263,268,274,279,284,289,294,299,305,310,315,340,345,350,356,361,366,371,385,391],{"type":27,"tag":28,"props":29,"children":30},"element","h1",{"id":11},[31],{"type":32,"value":33},"text","What to review in the first 48 hours of an inherited AWS account",{"type":27,"tag":35,"props":36,"children":37},"p",{},[38],{"type":32,"value":39},"Taking over an inherited AWS account is not the time to prove how fast you can move by changing production.",{"type":27,"tag":35,"props":41,"children":42},{},[43],{"type":32,"value":44},"An account can be running, deploying every week and serving customers while still having blind spots: old access, costs nobody can fully explain, buckets or endpoints exposed without context, backups that exist only on paper, partial Terraform coverage, noisy alarms and dependencies that live in one person's head.",{"type":27,"tag":35,"props":46,"children":47},{},[48],{"type":32,"value":49},"The first 48 hours are not for rebuilding the platform. They are for starting to understand what supports production, which risks are visible and which decisions should not be made. This is an indicative window for an initial diagnosis, not a guarantee of a complete review: the scope depends on the account's size, available access and documentation.",{"type":27,"tag":35,"props":51,"children":52},{},[53],{"type":32,"value":54},"That distinction matters. A senior approach does not start by closing ports, deleting users or rewriting Terraform. It starts by separating real signals from operational noise.",{"type":27,"tag":56,"props":57,"children":59},"h2",{"id":58},"first-what-is-production-and-what-depends-on-what",[60],{"type":32,"value":61},"First: what is production, and what depends on what?",{"type":27,"tag":35,"props":63,"children":64},{},[65],{"type":32,"value":66},"Before looking at IAM, costs or Terraform, answer a basic question: which parts of this account support the business right now?",{"type":27,"tag":35,"props":68,"children":69},{},[70],{"type":32,"value":71},"An inherited account may contain multiple regions, mixed environments, old resources, unclear names and services that look critical even though they no longer are. The reverse can also happen: a resource with a temporary name may sit in the middle of the main production flow.",{"type":27,"tag":35,"props":73,"children":74},{},[75],{"type":32,"value":76},"First identify accounts, regions, applications, databases, queues, jobs, pipelines and public entry points with real impact. You do not need a perfect diagram, but you do need enough of a map to avoid mistaking risky changes for cleanup.",{"type":27,"tag":35,"props":78,"children":79},{},[80],{"type":32,"value":81},"Ownership matters: who deploys, who responds to incidents, who approves changes, who understands the main database, and who knows what should not be touched at the end of the week.",{"type":27,"tag":35,"props":83,"children":84},{},[85],{"type":32,"value":86},"Without that context, every technical finding looks simpler than it is.",{"type":27,"tag":56,"props":88,"children":90},{"id":89},"iam-and-access-review-warning-signs-do-not-launch-a-purge",[91],{"type":32,"value":92},"IAM and access: review warning signs, do not launch a purge",{"type":27,"tag":35,"props":94,"children":95},{},[96],{"type":32,"value":97},"IAM is usually one of the first areas worth looking at, but also one of the least forgiving of impulsive changes.",{"type":27,"tag":35,"props":99,"children":100},{},[101],{"type":32,"value":102},"An initial review should look for risk signals: old human user accounts, access with no clear owner, very broad permissions, roles that are hard to explain, long-lived credentials, inconsistent MFA use or policies that have grown through accumulation.",{"type":27,"tag":35,"props":104,"children":105},{},[106],{"type":32,"value":107},"The goal in the first 48 hours is not perfect IAM. It is to understand how people access the account, how systems access it, which access supports deployments or integrations, and where uncertainty warrants investigation.",{"type":27,"tag":35,"props":109,"children":110},{},[111],{"type":32,"value":112},"Revoking permissions without knowing which workflow uses them can break production, CI/CD or an external dependency. That is why the reasonable output is not a mass cleanup, but a prioritised list: obvious risks needing attention, access needing an owner, and areas needing more context before changes.",{"type":27,"tag":56,"props":114,"children":116},{"id":115},"public-exposure-what-is-open-and-why",[117],{"type":32,"value":118},"Public exposure: what is open, and why?",{"type":27,"tag":35,"props":120,"children":121},{},[122],{"type":32,"value":123},"Public exposure is not bad by definition. A web application, API or load balancer exists precisely to receive traffic.",{"type":27,"tag":35,"props":125,"children":126},{},[127],{"type":32,"value":128},"The useful question is different: what is exposed, whether the exposure is intentional, who maintains it, and what the impact of getting it wrong would be.",{"type":27,"tag":35,"props":130,"children":131},{},[132],{"type":32,"value":133},"In a first review, look at surfaces such as load balancers, security groups, buckets, endpoints, DNS, certificates and services accepting traffic from the Internet. Not to apply generic recipes, but to classify them.",{"type":27,"tag":35,"props":135,"children":136},{},[137],{"type":32,"value":138},"Some exposure is expected, some is tolerated because of a historical decision, and some is accidental. Treating all of it the same leads to poor decisions: either too much is ignored or changes are made too quickly.",{"type":27,"tag":35,"props":140,"children":141},{},[142],{"type":32,"value":143},"A typical warning sign in an inherited account is that nobody can explain why a rule, bucket or endpoint is configured the way it is. That does not mean changing it on the fly. It means establishing context, impact and priority.",{"type":27,"tag":56,"props":145,"children":147},{"id":146},"costs-understand-the-bill-without-turning-it-into-finops",[148],{"type":32,"value":149},"Costs: understand the bill without turning it into FinOps",{"type":27,"tag":35,"props":151,"children":152},{},[153],{"type":32,"value":154},"Costs tell an architecture story too.",{"type":27,"tag":35,"props":156,"children":157},{},[158],{"type":32,"value":159},"In the first 48 hours, there is no need to set up a full FinOps programme or promise savings. The goal is more basic: understand which services account for the bill, what has changed recently, and whether there are anomalies worth investigating.",{"type":27,"tag":35,"props":161,"children":162},{},[163],{"type":32,"value":164},"Sometimes spending comes from legitimate growth. Sometimes it comes from oversized resources, logs retained without a clear policy, forgotten snapshots, cross-zone traffic, NAT Gateway, environments nobody uses, or architectural decisions that made sense at another time.",{"type":27,"tag":35,"props":166,"children":167},{},[168],{"type":32,"value":169},"The important thing is not to turn the cost review into a hunt for someone to blame. The senior-level question is: which part of the spending is aligned with production, and which part is still uncertain?",{"type":27,"tag":35,"props":171,"children":172},{},[173],{"type":32,"value":174},"If nobody can explain the main cost categories without depending on one particular person, that is already a useful finding. Not necessarily urgent, but relevant to the roadmap.",{"type":27,"tag":56,"props":176,"children":178},{"id":177},"backups-and-recovery-do-not-assume-confidence-just-because-production-is-running",[179],{"type":32,"value":180},"Backups and recovery: do not assume confidence just because production is running",{"type":27,"tag":35,"props":182,"children":183},{},[184],{"type":32,"value":185},"An account being in production does not mean it has a reliable recovery strategy.",{"type":27,"tag":35,"props":187,"children":188},{},[189],{"type":32,"value":190},"Backups may be configured without clarity about what they cover, their retention, who reviews them, which systems are left out, or when a restore was last tested.",{"type":27,"tag":35,"props":192,"children":193},{},[194],{"type":32,"value":195},"In an inherited account, this review should be careful. It is not about inventing RTOs or RPOs in two days, nor about declaring everything wrong because there is no formal document. It is about knowing which systems are critical and how much confidence there is if something fails.",{"type":27,"tag":35,"props":197,"children":198},{},[199],{"type":32,"value":200},"Databases, volumes, buckets, repositories, secrets, configurations, Terraform states and external dependencies may require different approaches. The worrying sign is not only the absence of a backup; it is also the absence of anyone who can explain how something important would be recovered.",{"type":27,"tag":35,"props":202,"children":203},{},[204],{"type":32,"value":205},"A practical question: if the most critical component had to be restored tomorrow, would the team know what to do without improvising under pressure?",{"type":27,"tag":56,"props":207,"children":209},{"id":208},"terraform-state-and-real-infrastructure-compare-the-map-with-the-territory",[210],{"type":32,"value":211},"Terraform, state and real infrastructure: compare the map with the territory",{"type":27,"tag":35,"props":213,"children":214},{},[215],{"type":32,"value":216},"Terraform can be a great help in understanding an inherited AWS account, but it does not always describe the whole reality.",{"type":27,"tag":35,"props":218,"children":219},{},[220],{"type":32,"value":221},"There may be infrastructure managed by Terraform, manually created resources, urgent changes made outside the normal workflow, old modules, remote state with unclear ownership, or environments that do not follow the same pattern.",{"type":27,"tag":35,"props":223,"children":224},{},[225],{"type":32,"value":226},"The initial review should answer three things: what is actually under infrastructure as code, what is outside it, and how risky it is to change.",{"type":27,"tag":35,"props":228,"children":229},{},[230],{"type":32,"value":231},"There is no need to turn this part into a Terraform restructuring exercise. In the first 48 hours, it is enough to identify whether state is reasonably protected, environments are separated, changes go through review, plans are understandable, and there is relevant drift between what is declared and what is deployed.",{"type":27,"tag":35,"props":233,"children":234},{},[235],{"type":32,"value":236},"The key is not to confuse “we have Terraform” with “the infrastructure is safe to change”. Terraform is part of the map. The actual account is the territory.",{"type":27,"tag":56,"props":238,"children":240},{"id":239},"observability-would-you-see-a-problem-coming",[241],{"type":32,"value":242},"Observability: would you see a problem coming?",{"type":27,"tag":35,"props":244,"children":245},{},[246],{"type":32,"value":247},"An inherited account can have observability tools without signals useful for operations.",{"type":27,"tag":35,"props":249,"children":250},{},[251],{"type":32,"value":252},"There may be dashboards, logs, metrics, alarms, CloudWatch, CloudTrail or AWS Config, yet the team still cannot answer basic questions when something goes wrong: which service is affected, since when, which deployment may have contributed, whether customers are affected, or who needs to act.",{"type":27,"tag":35,"props":254,"children":255},{},[256],{"type":32,"value":257},"The observability review should not start by asking how many dashboards exist. It should start by asking which decisions the available information enables.",{"type":27,"tag":35,"props":259,"children":260},{},[261],{"type":32,"value":262},"An alarm nobody responds to is noise. A dashboard only one person understands is a dependency. A log that exists but does not help an investigation is operational debt too.",{"type":27,"tag":35,"props":264,"children":265},{},[266],{"type":32,"value":267},"In the first 48 hours, the goal is to identify whether the team could detect degradation, understand its impact and relate it to recent changes. If not, the platform may be operating with more apparent confidence than real confidence.",{"type":27,"tag":56,"props":269,"children":271},{"id":270},"production-dependencies-what-is-missing-from-the-diagram",[272],{"type":32,"value":273},"Production dependencies: what is missing from the diagram",{"type":27,"tag":35,"props":275,"children":276},{},[277],{"type":32,"value":278},"A review of an inherited account should not stop at the obvious.",{"type":27,"tag":35,"props":280,"children":281},{},[282],{"type":32,"value":283},"There are also certificates, domains, third-party integrations, pipelines, overnight jobs, queues, webhooks, secrets, service users, repositories, images, runners, external tools and manual processes supporting critical parts of the system.",{"type":27,"tag":35,"props":285,"children":286},{},[287],{"type":32,"value":288},"Many of these dependencies do not appear in the first diagram. Sometimes they are not even documented. They emerge by asking what happens if a provider goes down, a certificate expires, a job fails, a credential changes, or the person who knows the workflow is unavailable.",{"type":27,"tag":35,"props":290,"children":291},{},[292],{"type":32,"value":293},"Minimum documentation is highly valuable here. Not endless manuals, but ownership, impact, failure signals and basic operating steps.",{"type":27,"tag":35,"props":295,"children":296},{},[297],{"type":32,"value":298},"An inherited AWS account is not understood just by looking at AWS. It is understood by looking at the complete system that keeps production running.",{"type":27,"tag":56,"props":300,"children":302},{"id":301},"how-to-prioritise-after-the-48-hours",[303],{"type":32,"value":304},"How to prioritise after the 48 hours",{"type":27,"tag":35,"props":306,"children":307},{},[308],{"type":32,"value":309},"The result of a good initial review should not be an endless list of problems.",{"type":27,"tag":35,"props":311,"children":312},{},[313],{"type":32,"value":314},"It should be a clear classification:",{"type":27,"tag":316,"props":317,"children":318},"ul",{},[319,325,330,335],{"type":27,"tag":320,"props":321,"children":322},"li",{},[323],{"type":32,"value":324},"Immediate risks that require careful attention.",{"type":27,"tag":320,"props":326,"children":327},{},[328],{"type":32,"value":329},"Issues that need investigation before anything is changed.",{"type":27,"tag":320,"props":331,"children":332},{},[333],{"type":32,"value":334},"Technical debt that can be scheduled.",{"type":27,"tag":320,"props":336,"children":337},{},[338],{"type":32,"value":339},"Future improvements that do not block production.",{"type":27,"tag":35,"props":341,"children":342},{},[343],{"type":32,"value":344},"This classification avoids two common mistakes: ignoring important signals because “everything works”, or trying to fix everything at once and creating more risk than is reduced.",{"type":27,"tag":35,"props":346,"children":347},{},[348],{"type":32,"value":349},"An inherited account needs judgement, not heroics. Sometimes the best decision in the first 48 hours is not to change anything yet, but to come away with a much clearer map of what deserves action and in what order.",{"type":27,"tag":56,"props":351,"children":353},{"id":352},"resolve-uncertainty-before-accelerating-changes",[354],{"type":32,"value":355},"Resolve uncertainty before accelerating changes",{"type":27,"tag":35,"props":357,"children":358},{},[359],{"type":32,"value":360},"Inheriting an AWS account should not turn into audit theatre or a race to demonstrate control.",{"type":27,"tag":35,"props":362,"children":363},{},[364],{"type":32,"value":365},"The priority is to understand what supports production, where there is real exposure, what access exists, which costs deserve attention, which backups inspire confidence, how closely Terraform matches reality, which observability signals are useful, and which dependencies were not visible.",{"type":27,"tag":35,"props":367,"children":368},{},[369],{"type":32,"value":370},"With that, the team can make better decisions: what to touch now, what to investigate, what to put on the roadmap and what to leave alone.",{"type":27,"tag":35,"props":372,"children":373},{},[374,376,383],{"type":32,"value":375},"If you have inherited an AWS account and need to separate real risks from operational noise, AstralDeploy can help with an ",{"type":27,"tag":377,"props":378,"children":380},"a",{"href":379},"/en/services/cloud-reliability",[381],{"type":32,"value":382},"AWS cloud reliability review",{"type":32,"value":384}," focused on understanding production, prioritising findings and planning changes with less risk to what already supports the business, without guaranteeing the absence of failures.",{"type":27,"tag":56,"props":386,"children":388},{"id":387},"sources-consulted",[389],{"type":32,"value":390},"Sources consulted",{"type":27,"tag":316,"props":392,"children":393},{},[394,406,417,428,439,450,461],{"type":27,"tag":320,"props":395,"children":396},{},[397,399],{"type":32,"value":398},"AWS IAM — Security best practices, temporary credentials, MFA and least privilege: ",{"type":27,"tag":377,"props":400,"children":404},{"href":401,"rel":402},"https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html",[403],"nofollow",[405],{"type":32,"value":401},{"type":27,"tag":320,"props":407,"children":408},{},[409,411],{"type":32,"value":410},"Amazon S3 — Block Public Access and evaluating public access: ",{"type":27,"tag":377,"props":412,"children":415},{"href":413,"rel":414},"https://docs.aws.amazon.com/AmazonS3/latest/userguide/access-control-block-public-access.html",[403],[416],{"type":32,"value":413},{"type":27,"tag":320,"props":418,"children":419},{},[420,422],{"type":32,"value":421},"AWS Cost Explorer — Cost and usage analysis to identify areas requiring investigation: ",{"type":27,"tag":377,"props":423,"children":426},{"href":424,"rel":425},"https://docs.aws.amazon.com/cost-management/latest/userguide/ce-what-is.html",[403],[427],{"type":32,"value":424},{"type":27,"tag":320,"props":429,"children":430},{},[431,433],{"type":32,"value":432},"AWS Well-Architected — REL 9, backing up data, applications and configuration, and periodically checking recovery: ",{"type":27,"tag":377,"props":434,"children":437},{"href":435,"rel":436},"https://docs.aws.amazon.com/wellarchitected/latest/framework/rel-09.html",[403],[438],{"type":32,"value":435},{"type":27,"tag":320,"props":440,"children":441},{},[442,444],{"type":32,"value":443},"AWS Well-Architected — REL 13, recovery objectives and strategies based on business needs: ",{"type":27,"tag":377,"props":445,"children":448},{"href":446,"rel":447},"https://docs.aws.amazon.com/wellarchitected/latest/framework/rel-13.html",[403],[449],{"type":32,"value":446},{"type":27,"tag":320,"props":451,"children":452},{},[453,455],{"type":32,"value":454},"HashiCorp Terraform — Planning and comparing configuration, state and remote objects: ",{"type":27,"tag":377,"props":456,"children":459},{"href":457,"rel":458},"https://developer.hashicorp.com/terraform/cli/commands/plan",[403],[460],{"type":32,"value":457},{"type":27,"tag":320,"props":462,"children":463},{},[464,466],{"type":32,"value":465},"AWS Well-Architected — Monitoring and observability focused on decisions and responsibilities: ",{"type":27,"tag":377,"props":467,"children":470},{"href":468,"rel":469},"https://docs.aws.amazon.com/wellarchitected/latest/management-and-governance-guide/observability.html",[403],[471],{"type":32,"value":468},{"title":7,"searchDepth":473,"depth":473,"links":474},2,[475,476,477,478,479,480,481,482,483,484,485],{"id":58,"depth":473,"text":61},{"id":89,"depth":473,"text":92},{"id":115,"depth":473,"text":118},{"id":146,"depth":473,"text":149},{"id":177,"depth":473,"text":180},{"id":208,"depth":473,"text":211},{"id":239,"depth":473,"text":242},{"id":270,"depth":473,"text":273},{"id":301,"depth":473,"text":304},{"id":352,"depth":473,"text":355},{"id":387,"depth":473,"text":390},"markdown","content:en:blog:2026-06-18-what-to-review-in-the-first-48-hours-of-an-inherited-aws-account.md","content","en/blog/2026-06-18-what-to-review-in-the-first-48-hours-of-an-inherited-aws-account.md","en/blog/2026-06-18-what-to-review-in-the-first-48-hours-of-an-inherited-aws-account","md",1795108324621]