Terraform Drift Review for an Offshore Infrastructure Team

Start with the buying decision

This guide is for platform owners asking a distributed engineer to investigate divergence between declared and deployed infrastructure. The immediate decision is whether and how to separate expected provider noise, emergency changes, stale state, and unauthorized configuration before proposing reconciliation. Write that decision at the top of the working document, name the deadline, and identify who can approve it. A provider can supply facts and options, but the buyer should retain the final judgment about budget, architecture, security exceptions, and production risk.

Build the minimum evidence packet

The useful deliverable is a drift packet with workspace, state revision, plan output, resource owner, change history, risk classification, and approved next action. Give every provider or candidate the same factual starting point: stack, application boundaries, expected work, working hours, required experience, review path, environments, data sensitivity, and proposed start window. Mark estimates as estimates, and distinguish confirmed requirements from preferences that can change during discovery.

Begin with a refresh-only plan against the named workspace and pinned provider versions. Export the state lineage and serial, record backend identity without exposing credentials, and classify every difference as configuration drift, state drift, provider normalization, external control, or unexplained evidence. A console change made during an incident is not automatically wrong; locate its ticket, owner, expiry, and recovery intent. Review replace and destroy actions separately from in-place changes, including dependencies that make a small-looking correction destructive. The engineer may reproduce and annotate the plan, but the resource owner decides import, configuration update, state operation, or infrastructure change. Re-run after that decision and require a zero-surprise plan rather than treating an empty diff as the only acceptable outcome. Never edit remote state by hand or reconcile from an unreviewed local configuration.

Buyer decision record

Scroll sideways to read every column on a small screen.

Normalize the options before ranking them

Put alternatives into one comparison table. Use the same time period, currency date, capacity assumption, service boundary, and definition of completion. State who supplies management, product decisions, code review, quality checks, equipment, licenses, security administration, and release approval. An apparently expensive option may include a control or role that another proposal leaves with the buyer.

Assign ownership at each boundary

Use a simple responsible-and-approving map for scope, architecture, access, implementation, verification, acceptance, deployment, incidents, invoicing, and offboarding. The engineering manager should know which decisions cannot be delegated. Provider ownership must be paired with the authority, inputs, and response path needed to perform the work; otherwise the contract assigns responsibility without a workable operating model.

Apply the review to infrastructure governance by starting with a drift packet with workspace, state revision, plan output, resource owner, change history, risk classification, and approved next action. Preserve identifiers, versions, commands, timestamps, expected results, observed results, and named owners. The working team should explicitly separate expected provider noise, emergency changes, stale state, and unauthorized configuration before proposing reconciliation. Test the normal path, a denied or invalid path, an interrupted path, and recovery from a known state. Measure unexpected resource count, destructive plan actions, unresolved owners, state age, and approved reconciliation time; define each measure's start event, end event, data source, and decision threshold before interpreting it. The service lane is DevOps release support, but buyer-side owners retain production access, accepted risk, security exceptions, architecture choices, and release approval.

A useful comparison has four columns: declared address, observed remote object, evidence explaining the difference, and approved disposition. Include moved blocks, imports, ignored attributes, provider-computed values, and resources missing from either side. Run the saved plan through peer review and retain its hash so approval cannot drift onto a later plan. If a backend lock, refresh error, or unavailable API makes evidence incomplete, stop; do not convert partial visibility into a reconciliation proposal. Close temporary incident changes through their own owners, and schedule state backups before any approved state operation.

Test the failure case before commitment

Model this uncomfortable case: an automatic apply removes a production-side emergency control because nobody established why the declared and observed configurations differ. Ask what signal reveals the problem, who notices it, which work stops, how evidence is preserved, and who chooses recovery. A useful answer names a control and produces an artifact. “We communicate closely” is not a recovery plan unless the channel, response expectation, backup, and authority are defined.

Measure the delivery system, not online activity

For this decision, track unexpected resource count, destructive plan actions, unresolved owners, state age, and approved reconciliation time. Define the start and end event for every measure, where the record comes from, and what action a threshold triggers. Compare trends with the baseline and with changes in scope or team composition. Small samples need context; one difficult release should prompt investigation rather than a claim about long-term performance.

For the cross-time-zone handoff, record the exact revision, environment, fixture, changed paths, checks run, skipped checks, open uncertainty, stop condition, reviewer, and next authorized action. Use synthetic data and least privilege. The specific failure to rehearse is this: an automatic apply removes a production-side emergency control because nobody established why the declared and observed configurations differ. Ask which signal distinguishes that failure from a harmless variation, which action is reversible, and who can approve the consequence. Re-test after a relevant dependency, configuration, traffic shape, ownership boundary, or platform version changes. A passing sample supports only the declared scope; it is not a guarantee about systems, users, data, or conditions that were not observed.

Acceptance requires a reviewed saved plan whose address-by-address disposition has an owner, whose destructive actions are explicitly approved, and whose state lineage matches the intended workspace. Unexplained remote objects, refresh failures, stale configuration, or expired incident exceptions keep the review open. The next operator receives the plan hash and may not substitute a new plan silently.

Set a review and exit path on day one

Schedule an early operating review and a later commercial review. The operating review checks access, ticket readiness, review capacity, evidence quality, and handoffs. The commercial review compares actual use and outcomes with the assumptions in the decision record. Give improvements an owner and a date, then state what evidence will show that the correction worked.

Turn the comparison into a bounded first step

Choose the smallest paid step that can retire the largest uncertainty. It may be a discovery session, a work sample, a backlog and architecture review, or one production-shaped change in a controlled environment. Define the output, time box, reviewers, access, acceptance evidence, and stopping point. Do not label an open-ended engagement a pilot merely because it starts small.

Questions about assessing Philippine developers