CertKata

AWS Certified DevOps Engineer - Professional — Free Essentials

Key facts distilled from CertKata's DOP-C02 question bank. Free to read — no account needed.

The DR spectrum, and what each point costs

Disaster recovery strategies on one spectrum: cost and standby readiness rise from backup-and-restore through pilot light and warm standby to multi-site active-active, while recovery time fallsLOWER COST · LESS READYHIGHER COST · MORE READYBackup& restorenothing runningPilotlightcore onlyWarmstandbyscaled-down copyMulti-siteactive-activefull capacityRTOhourstens of minminutesnear zeroRECOVERY TIME FALLS AS SPEND RISESthe choice is RTO/RPO against standby cost

Backup and restore, pilot light, warm standby and active-active sit on one spectrum: cost and standing capacity rise together as recovery time falls. Pilot light keeps the data current with minimal running compute; warm standby runs a scaled-down copy ready to take load. RTO and cost together decide which one you buy, and RPO is decided by how current the replicated data is. None of it counts until it is exercised - only tested restores and tested failovers prove anything at all.

Build once, promote many

The artifact that was tested is the artifact that ships. Build it once, tag it immutably with the commit it came from, and promote that exact tag through every environment - rebuilding per environment breaks the guarantee entirely, because the thing in production was never the thing that passed. Reproducibility needs pinned dependencies and a versioned, containerised build environment; unpinned inputs are also the usual explanation for a build that works locally and fails in CI.

Alarms people still respond to

An alarm that is not actionable gets ignored, and then the actionable one is ignored alongside it. Each alarm should carry impact and guidance. Reduce noise without losing signal: evaluate over multiple periods, combine conditions in composite alarms, use anomaly detection where normal varies, and suppress during known maintenance rather than deleting. Missing data is an explicit decision, not a default - and heartbeat and agent-health alarms are what catch monitoring that has quietly stopped.

Guardrails: SCPs, boundaries and explicit deny

IAM policy evaluation order: an explicit deny always wins, then an explicit allow is considered, and anything not explicitly allowed is implicitly deniedAPI requestAny explicit Deny?DENYYESNOAny explicit Allow?ALLOWYESNOImplicit DENY(the default)an explicit Deny can never be overridden by an Allowa permissions boundary caps what any Allow can grant

A service control policy sets a ceiling on what an account may do and can never grant a permission - it is the wrong answer to any question asking how to give access. Member-account IAM cannot override it, which is what makes it a guardrail rather than a suggestion. An explicit deny wins over any allow, and a permissions boundary caps what a delegated administrator's grants can reach, which is how privilege escalation gets prevented rather than merely detected.

Deployment strategies and automatic rollback

Blue/green keeps the previous version running so reversal is immediate; canary and linear traffic shifting expose a fraction first, and CodeDeploy alias shifting with alarms is how that becomes automatic rather than watched. ECS blue/green validates the replacement task set before any traffic moves. In-place rolling updates hold capacity with a minimum-healthy-hosts setting. Whatever the shape, post-deployment validation and a deployment timeout are what stop a bad release sitting there.

This is a preview. Get the full DOP-C02 cheat sheet — every domain, bundled with the practice question bank.

Sign up for the full version