name: QingLong 3.0 Security Administration Kubernetes live evidence

on:
  push:
    branches:
      - next
    paths:
      - .github/workflows/ql3-security-administration-live.yml
      - deploy/containers/ql3-cluster-admin/**
      - deploy/containers/ql3-cluster-control/**
      - deploy/kubernetes/ql3-cluster/base/**
      - deploy/kubernetes/ql3-cluster/operations/cloudnative-pg/**
      - deploy/kubernetes/ql3-cluster/operations/security-administration/**
      - package.json
      - packages/ql3-cluster-admin/**
      - packages/ql3-cluster-control/**
      - packages/ql3-cluster-postgres/**
      - packages/ql3-runtime-core/**
      - pnpm-lock.yaml
      - scripts/lib/ql3-k3s-docker-live.cjs
      - scripts/ql3-security-administration-kubernetes-*.cjs
      - test/back/ql3SecurityAdministrationKubernetes*.test.cjs
  workflow_dispatch:

permissions:
  contents: read

concurrency:
  group: ql3-security-administration-live-${{ github.ref }}
  cancel-in-progress: false

jobs:
  security-administration-live:
    name: Three-node Security Administration, PostgreSQL and PVC custody
    runs-on: ubuntu-24.04
    timeout-minutes: 90
    steps:
      - uses: actions/checkout@v6
      - uses: pnpm/action-setup@v6
        with:
          version: '8.3.1'
      - uses: actions/setup-node@v6
        with:
          node-version: '24.18.0'
          cache: pnpm
          cache-dependency-path: pnpm-lock.yaml

      - name: Require enough ephemeral disk for three isolated K3s nodes
        shell: bash
        run: |
          set -euo pipefail
          available_kib="$(df -Pk / | awk 'NR == 2 { print $4 }')"
          minimum_kib="$((25 * 1024 * 1024))"
          if (( available_kib < minimum_kib )); then
            echo "Security Administration live gate requires at least 25 GiB free; found ${available_kib} KiB" >&2
            exit 1
          fi
          docker volume ls --quiet --filter dangling=true | sort > \
            "${RUNNER_TEMP}/ql3-dangling-volumes.before"
          docker system df

      - name: Install workspace dependencies without lifecycle scripts
        run: pnpm install --frozen-lockfile --ignore-scripts

      - name: Install verified kubectl v1.34.3
        shell: bash
        run: |
          set -euo pipefail
          curl --fail --location --silent --show-error \
            --output "${RUNNER_TEMP}/kubectl" \
            https://dl.k8s.io/release/v1.34.3/bin/linux/amd64/kubectl
          curl --fail --location --silent --show-error \
            --output "${RUNNER_TEMP}/kubectl.sha256" \
            https://dl.k8s.io/release/v1.34.3/bin/linux/amd64/kubectl.sha256
          test "$(cat "${RUNNER_TEMP}/kubectl.sha256")" = \
            "$(sha256sum "${RUNNER_TEMP}/kubectl" | cut -d ' ' -f 1)"
          chmod 0755 "${RUNNER_TEMP}/kubectl"

      - name: Fetch checksum-locked CloudNativePG operator manifest
        shell: bash
        run: |
          set -euo pipefail
          curl --fail --location --silent --show-error \
            --output "${RUNNER_TEMP}/cloudnative-pg.yaml" \
            https://github.com/cloudnative-pg/cloudnative-pg/releases/download/v1.30.0/cnpg-1.30.0.yaml
          echo "f8bede43fe4ee0d478c2355b204a36876b2ae4faac60f2a9452280b293da3b88  ${RUNNER_TEMP}/cloudnative-pg.yaml" | sha256sum --check

      - name: Recheck static Security Administration and evidence contracts
        run: |
          pnpm --filter @qinglong/runtime-core build
          pnpm --filter @qinglong/cluster-postgres build
          pnpm audit:security-administration-kubernetes:ql3
          node --test \
            test/back/ql3SecurityAdministrationKubernetesLiveContract.test.cjs \
            test/back/ql3SecurityAdministrationKubernetesLiveAudit.test.cjs

      - name: Preload the digest-bound K3s distribution
        run: docker pull rancher/k3s:v1.34.3-k3s1

      - name: Prepare Linux bridge netfilter for nested K3s
        shell: bash
        run: |
          set -euo pipefail
          sudo modprobe br_netfilter
          test -e /proc/sys/net/bridge/bridge-nf-call-iptables
          sudo sysctl -w net.ipv4.ip_forward=1
          sudo sysctl -w net.bridge.bridge-nf-call-iptables=1

      - name: Prove Security Administration through K3s, PostgreSQL and PVC
        env:
          QL3_SECURITY_ADMINISTRATION_KUBERNETES_LIVE: '1'
          QL3_KUBECTL_BIN: ${{ runner.temp }}/kubectl
          QL3_CNPG_OPERATOR_MANIFEST_FILE: ${{ runner.temp }}/cloudnative-pg.yaml
          QL3_SECURITY_ADMINISTRATION_REPORT: ${{ runner.temp }}/ql3-security-administration/report.json
        shell: bash
        run: |
          set -euo pipefail
          umask 077
          mkdir -m 0700 "$(dirname "${QL3_SECURITY_ADMINISTRATION_REPORT}")"
          pnpm test:security-administration-kubernetes-live:ql3 \
            "--report=${QL3_SECURITY_ADMINISTRATION_REPORT}"

      - name: Re-audit content-free report and isolated cleanup
        env:
          QL3_SECURITY_ADMINISTRATION_REPORT: ${{ runner.temp }}/ql3-security-administration/report.json
        shell: bash
        run: |
          set -euo pipefail
          test "$(stat -c '%a' "${QL3_SECURITY_ADMINISTRATION_REPORT}")" = '600'
          pnpm audit:security-administration-kubernetes-live:ql3 \
            "--report=${QL3_SECURITY_ADMINISTRATION_REPORT}"
          sha256sum "${QL3_SECURITY_ADMINISTRATION_REPORT}"
          test -z "$(docker ps -aq --filter name=ql3-security-live-)"
          test -z "$(docker network ls -q --filter name=ql3-security-live-)"
          docker volume ls --quiet --filter dangling=true | sort > \
            "${RUNNER_TEMP}/ql3-dangling-volumes.after"
          diff --unified \
            "${RUNNER_TEMP}/ql3-dangling-volumes.before" \
            "${RUNNER_TEMP}/ql3-dangling-volumes.after"
          docker system df

      - name: Upload audited content-free Security Administration evidence
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: ql3-security-administration-${{ github.run_id }}-${{ github.run_attempt }}
          path: ${{ runner.temp }}/ql3-security-administration/report.json
          if-no-files-found: error
          retention-days: 14
          compression-level: 9
          overwrite: false
          include-hidden-files: false