Terraform Skill

Terraform Release and Provisioner Safety

Prevent a valid-looking Terraform workflow from publishing unvalidated bytes or widening a change's blast radius. Keep the user's business outcome and the actual mutation surface ahead of plan counts, green wrappers, or process completeness.

Route the task

• Release, shared gateway, saved plan, staging receipt, or production promotion: read release-safety-and-environment-parity.md and pre-deploy-validation.md before approval or mutation.
• Environment isolation, including cached initialization, environment retirement, shared data backend, DNS ownership, state, or snapshots: read multi-env-isolation.md.
• Pre-deploy checks or a validator: read pre-deploy-validation.md.
• Fresh instance or empty data disk: read zero-to-deploy-checklist.md.
• One known provisioner symptom: use the matching pattern below.

Operating contract

1. Use the repository's canonical release entry. When shared-gateway publication requires standard CI, agents may initiate that CI within existing authorization; normal release, restoration, rollback, and static-content transfer all use its controlled executor and existing wrappers. Do not substitute a local Terraform apply, root SSH/SCP, helper, or console write. Verify the implemented workflow and supported inputs before dispatch; lint or image-build CI alone is not a deployment lane. A missing lane blocks the live write, not preparation or read-only diagnosis.
2. Prefer provider resources, image baking, cloud-init, or configuration management. HashiCorp recommends exhausting purpose-built alternatives because Terraform cannot model provisioner side effects predictably. When a provisioner remains necessary, make its artifact, target, lock, validation, and readback explicit.
3. Inventory every resource or recovery tool that can write the target runtime. A validator attached to only one writer does not protect another writer of the same shared service.
4. Give staging and production one required-key schema. Let values differ; never let a key be required in one environment and optional, defaulted, absent, or allowed-empty in another.
5. Validate the exact candidate bytes, Compose-rendered environment, and immutable runtime image before the first live write or restart. Keep post-deploy checks too: they detect damage but cannot prevent the first bad mutation.
6. Treat a saved plan as an executable artifact. Bind it to reviewed source/artifact identity and apply that exact file. A successful apply is not a staging receipt; record promotion evidence only after every required live verifier succeeds.
7. Honor the user's production authorization, including standing authorization, at the last reversible point. A deadline, PLAN_DIGEST, CONFIRM_*, or agent inference does not supply it. Do not add personal signatures or per-run human approval when the authorized contract is standard CI.
8. Stop after the requested result is verified. Do not turn a single-service fix into full-stack drift reconciliation, recovery redesign, or unrelated hardening.

Provisioner traps (symptom → fix)

Use these incident-derived symptom patterns to choose the next falsifying check. Confirm the current source and runtime before promoting a historical cause into the present diagnosis.

terraform plan says "No changes" but the site is down

Symptom: Browser ERR_SSL_PROTOCOL_ERROR or similar, terraform plan shows No changes, but the service is actually broken.

Interpretation: No changes is not a live-health check. Terraform refreshes provider-managed attributes, but a provisioner's remote files and container side effects are not automatically represented by those attributes. A broken site can follow an authorized apply, another IaC writer, or an out-of-band change; this symptom alone does not identify which happened. An HTTP redirect or healthy container does not prove the affected HTTPS route works.

Diagnosis:

1. Probe the affected hostname and user path, then inspect the loaded route and live files.
2. Compare the exact deployed source/artifact and state-tracked hashes, where present, with live hashes. Read the actual provisioner commands and release logs for every writer of that directory. Shared mtimes and deployed .tftpl/.static files are clues to inspect the archive and copy path, not proof of an out-of-band write: a normal archive extraction can produce them too.
3. Reconstruct each writer's candidate file set and delete/exclude scope. In particular, a valid rsync --delete can remove another owner's route omitted from its candidate while retaining syntactically valid configuration. Check the complete existing hostname set, not just the main API.

Recovery: Use the repository's canonical release entry and its plan/apply wrapper; where standard CI is required, initiate the supported recovery or rollback through that same CI. Before a replacement, verify the candidate against the current live baseline and name exactly which files/resources may change. A frozen old commit can be reproducible and still roll back a newer route, policy, environment, or image. Restore only the authorized scope; stop if the current writer cannot preserve other owners' files. Do not use a broad replacement followed by a second sync as the default repair.

A saved plan freezes Terraform's planned values, not arbitrary files or external downloads read by its provisioners at apply time. Check that the actual implementation consumes the validated immutable artifact before claiming recovery will publish those bytes. After the authorized apply, independently verify the affected user path and the previously working routes. Live hash checks can detect drift; they do not by themselves establish its writer or prevent deletion.

Staging applies cleanly but production fails port is already allocated

Symptom: Compose deploy passes staging verification, then the production apply dies at docker compose up with Bind for 127.0.0.1:<port> failed: port is already allocated.

Root cause: Staging/production parity covers configuration, not host port allocation. A port that is free on staging can already be occupied by another service on the production host. A tainted null_resource is left behind, so the retry must go through the replace flow (CONFIRM_REPLACE), not a plain re-apply.

Diagnosis:

ssh root@<target-host> 'ss -ltnp | grep <port>'   # find the real occupant
docker ps --format "{{.Names}} {{.Ports}}" | grep <port>

Prevention: Before binding a host port in a module, check the target host for the exact bind at design time — never infer freeness from staging. Prefer uncommon high ports, record the allocation in the module comment and the deploy doc, and when a container needs a host-local probe target remember 127.0.0.1 inside a container is the container itself: host-local targets belong to host-level probes (systemd scripts), not containerized monitors.

docker: not found in remote-exec

cloud-init still installing Docker when provisioner SSHs in.

provisioner "remote-exec" {
  inline = [
    "cloud-init status --wait",
    "command -v docker >/dev/null || { echo 'FATAL: Docker not ready'; exit 1; }",
  ]
}

rsync: connection unexpectedly closed in local-exec

Do not infer a universal Terraform limitation from this symptom. A second SSH client can lose to the target's connection budget, SSH policy, or a competing deploy. Keep local-exec local: package an immutable artifact there, then use a Terraform-managed upload or a purpose-built deploy system. Give every apply a unique remote staging path; never share /tmp/src.tar.gz across concurrent applies.

provisioner "local-exec" {
  command = "tar czf /tmp/src-${self.id}.tar.gz --exclude=node_modules --exclude=.git -C ${path.module}/../../.. myproject"
}
provisioner "file" {
  source      = "/tmp/src-${self.id}.tar.gz"
  destination = "/tmp/src-${self.id}.tar.gz"
}
provisioner "remote-exec" {
  inline = ["tar xzf /tmp/src-${self.id}.tar.gz -C /data/ && rm -f /tmp/src-${self.id}.tar.gz"]
}

macOS BSD tar: --exclude must come BEFORE the source argument.

cloud-init status shows "running" forever

apt-get -y does not suppress debconf dialogs. Packages like iptables-persistent block on TTY prompts.

- |
    echo iptables-persistent iptables-persistent/autosave_v4 boolean true | debconf-set-selections
    echo iptables-persistent iptables-persistent/autosave_v6 boolean true | debconf-set-selections
    DEBIAN_FRONTEND=noninteractive apt-get install -y iptables-persistent

Known offenders: iptables-persistent, postfix, mysql-server, wireshark-common.

EACCES: permission denied in container logs, container Restarting

Host volume dirs are root-owned; container runs as non-root (uid 1001). Fix before docker compose up:

mkdir -p /data/myapp/data /data/myapp/logs
chown -R 1001:1001 /data/myapp/data /data/myapp/logs

Find UID: grep adduser.*-u or USER in Dockerfile.

Provisioner fails but no diagnostic output

Keep fail-fast behavior; attach diagnostics to failure instead of disabling set -e. Otherwise an early failed command can be overwritten by a later green health check.

provisioner "remote-exec" {
  inline = [
    "set -eu",
    "trap 'rc=$?; if [ $rc -ne 0 ]; then docker logs myapp --tail 20 2>&1 || true; docker ps --format \\\"table {{.Names}}\\\\t{{.Status}}\\\" || true; fi; exit $rc' EXIT",
    "docker compose up -d",
    "sleep 15",
    "docker ps --filter name=myapp --format '{{.Status}}' | grep -q healthy || exit 1",
  ]
}