version: 1
id: reliability-gameday-capstone
title: Capstone — Reliability Game Day
summary: 'Run a payments-api reliability drill: record SLO burn and freeze, execute a chaos experiment with abort, write a blameless postmortem, and commit the evidence pack.'
difficulty: capstone
estimatedMinutes: 70
prerequisites: [gameday-budget-freeze, oncall-handoff-basics, leadership-postmortem]
image: alpine:3.21
shell: /bin/sh
setup:
  - "apk add --no-cache git >/dev/null"
  - "mkdir -p /workspace/slo /workspace/chaos /workspace/gameday /workspace/docs /workspace/oncall"
  - "cd /workspace && git init -q && git config user.email learner@platformforge.local && git config user.name Learner"
  - "printf 'slo_target_percent=99.9\\nsli_percent=99.6\\nbudget_remaining_percent=-0.2\\n' > /workspace/slo/inputs.env"
  - "printf 'name: draft\\ntarget: unknown\\n' > /workspace/chaos/experiment.yaml"
  - "printf 'Ticket: payments-api game day under freeze.\\nRecord burn/freeze, run pod-kill experiment with abort, postmortem, handoff, commit.\\n' > /workspace/TICKET.md"
  - "cd /workspace && git add . && git commit -qm 'reliability drill draft'"
tasks:
  - id: burn-and-freeze
    title: Record burn and engage freeze
    description: 'Write /workspace/slo/burn.txt containing -0.2 (from inputs.env budget_remaining_percent). Create /workspace/gameday/FREEZE.md with headings Status and Rule. Status must mention freeze or frozen. Rule must mention error budget or over budget. Write /workspace/oncall/open-pages.txt containing PaymentsBurnFast.'
    hints:
      - 'Tip code FREEZE_GATE: negative remaining budget means freeze risky changes before inject.'
      - 'Tip code OPEN_PAGES: list PaymentsBurnFast explicitly in open-pages.txt.'
      - 'burn.txt should be exactly the remaining budget value -0.2.'
    checks:
      - type: file
        name: Burn remaining recorded
        path: /workspace/slo/burn.txt
        value: "-0.2"
      - type: file
        name: Freeze status section
        path: /workspace/gameday/FREEZE.md
        value: Status
      - type: file
        name: Freeze rule section
        path: /workspace/gameday/FREEZE.md
        value: Rule
      - type: command
        name: Status mentions freeze
        command: "grep -A3 '^#* *Status' /workspace/gameday/FREEZE.md | grep -Eqi 'freeze|frozen'"
      - type: command
        name: Rule mentions budget
        command: "grep -A3 '^#* *Rule' /workspace/gameday/FREEZE.md | grep -Eqi 'error budget|over budget'"
      - type: file
        name: Open page recorded
        path: /workspace/oncall/open-pages.txt
        value: PaymentsBurnFast
  - id: chaos-with-abort
    title: Run the chaos experiment plan
    description: 'Replace /workspace/chaos/experiment.yaml with name payments-api-pod-kill, target payments-api, action kill-pod or pod-kill, and steadyStateHypothesis mentioning availability or ready. Write /workspace/docs/ABORT.md mentioning error budget or freeze and rollback. Write /workspace/gameday/roles.txt with exactly three non-empty lines including facilitator, scribe, and on-call.'
    hints:
      - 'Tip code ONE_FAILURE: keep the inject to a single-pod kill.'
      - 'Tip code ABORT_BUDGET: abort criteria must reference budget/freeze and rollback.'
      - 'roles.txt needs facilitator, scribe, and on-call.'
    checks:
      - type: file
        name: Experiment named
        path: /workspace/chaos/experiment.yaml
        value: payments-api-pod-kill
      - type: file
        name: Targets payments-api
        path: /workspace/chaos/experiment.yaml
        value: payments-api
      - type: command
        name: Action is pod kill
        command: "grep -Eqi 'kill-pod|pod-kill' /workspace/chaos/experiment.yaml"
      - type: command
        name: Steady state declared
        command: "grep -Eqi 'availability|ready' /workspace/chaos/experiment.yaml"
      - type: command
        name: Abort ties to reliability controls
        command: "grep -Eqi 'error budget|freeze' /workspace/docs/ABORT.md && grep -Eqi 'rollback' /workspace/docs/ABORT.md"
      - type: command
        name: Roles file complete
        command: "[ $(grep -cve '^[[:space:]]*$' /workspace/gameday/roles.txt) -eq 3 ] && grep -Eqi 'facilitator' /workspace/gameday/roles.txt && grep -Eqi 'scribe' /workspace/gameday/roles.txt && grep -Eqi 'on-call|oncall' /workspace/gameday/roles.txt"
  - id: postmortem-and-commit
    title: Write postmortem and commit the pack
    description: 'Create /workspace/docs/postmortem.md with headings Summary, Timeline, Root cause, Impact, and Action items. Root cause must mention chaos or inject or replica. Action items must mention alert or freeze or validation. Write /workspace/docs/LEARNING.md mentioning follow-up. Commit all changes so git status is clean and HEAD has at least two commits.'
    hints:
      - 'Tip code LEARN_CLOSE: action items must change alerts, freezes, or validation — not vibes.'
      - 'Keep the postmortem blameless — systems and process.'
      - 'Confirm git status --porcelain is empty after committing.'
    checks:
      - type: file
        name: Summary present
        path: /workspace/docs/postmortem.md
        value: Summary
      - type: file
        name: Timeline present
        path: /workspace/docs/postmortem.md
        value: Timeline
      - type: file
        name: Root cause present
        path: /workspace/docs/postmortem.md
        value: Root cause
      - type: file
        name: Impact present
        path: /workspace/docs/postmortem.md
        value: Impact
      - type: file
        name: Action items present
        path: /workspace/docs/postmortem.md
        value: Action items
      - type: command
        name: Root cause mentions experiment
        command: "grep -A4 '^#* *Root cause' /workspace/docs/postmortem.md | grep -Eqi 'chaos|inject|replica'"
      - type: command
        name: Action items are concrete
        command: "grep -A4 '^#* *Action items' /workspace/docs/postmortem.md | grep -Eqi 'alert|freeze|validation'"
      - type: file
        name: Learning note present
        path: /workspace/docs/LEARNING.md
        value: follow-up
      - type: command
        name: Evidence pack committed
        command: "cd /workspace && test \"$(git rev-list --count HEAD)\" -ge 2 && test -z \"$(git status --porcelain)\""
limits: {cpus: "0.75", memory: 192m, pids: 96, timeout: 3600, network: false}