Merge pull request 'CI: add deploy-node1-runtime with hard phase6 gate' (#2) from codex/deploy-gate-node1 into main

.gitea/workflows/deploy-node1-runtime.yml

@@ -0,0 +1,212 @@
+name: deploy-node1-runtime
+
+on:
+  workflow_dispatch:
+    inputs:
+      deploy_ref:
+        description: "Git ref to deploy on NODA1 (branch/tag/sha)"
+        required: false
+        type: string
+        default: "main"
+      redeploy_runtime:
+        description: "Rebuild/restart gateway+experience-learner after git sync"
+        required: false
+        type: boolean
+        default: false
+      ssh_host:
+        description: "NODA1 SSH host override"
+        required: false
+        type: string
+      ssh_user:
+        description: "NODA1 SSH user override (default root)"
+        required: false
+        type: string
+
+concurrency:
+  group: noda1-runtime-deploy
+  cancel-in-progress: false
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    env:
+      DEFAULT_SSH_HOST: ${{ secrets.NODA1_SSH_HOST }}
+      DEFAULT_SSH_USER: ${{ secrets.NODA1_SSH_USER }}
+      DEPLOY_REF: ${{ inputs.deploy_ref }}
+      REDEPLOY_RUNTIME: ${{ inputs.redeploy_runtime }}
+    steps:
+      - name: Resolve SSH target
+        shell: bash
+        run: |
+          set -euo pipefail
+          host="${DEFAULT_SSH_HOST:-}"
+          user="${DEFAULT_SSH_USER:-root}"
+
+          if [ -n "${{ inputs.ssh_host }}" ]; then
+            host="${{ inputs.ssh_host }}"
+          fi
+          if [ -n "${{ inputs.ssh_user }}" ]; then
+            user="${{ inputs.ssh_user }}"
+          fi
+
+          if [ -z "$host" ]; then
+            echo "Missing SSH host (workflow input or secret NODA1_SSH_HOST)" >&2
+            exit 1
+          fi
+
+          echo "SSH_HOST=$host" >> "$GITHUB_ENV"
+          echo "SSH_USER=$user" >> "$GITHUB_ENV"
+
+      - name: Prepare SSH key
+        shell: bash
+        env:
+          SSH_PRIVATE_KEY: ${{ secrets.NODA1_SSH_KEY }}
+        run: |
+          set -euo pipefail
+          set +x
+          if [ -z "${SSH_PRIVATE_KEY:-}" ]; then
+            echo "Missing secret NODA1_SSH_KEY" >&2
+            exit 1
+          fi
+          mkdir -p ~/.ssh
+          chmod 700 ~/.ssh
+          key_path=~/.ssh/noda1_ci_key
+          if printf '%s' "$SSH_PRIVATE_KEY" | grep -q 'BEGIN OPENSSH PRIVATE KEY'; then
+            printf '%s\n' "$SSH_PRIVATE_KEY" | tr -d '\r' > "$key_path"
+          else
+            printf '%s' "$SSH_PRIVATE_KEY" | tr -d '\r' | base64 --decode > "$key_path"
+          fi
+          chmod 600 "$key_path"
+          if ! ssh-keygen -y -f "$key_path" >/dev/null 2>&1; then
+            echo "Invalid SSH private key in NODA1_SSH_KEY" >&2
+            exit 1
+          fi
+          echo "SSH_KEY_PATH=$key_path" >> "$GITHUB_ENV"
+
+      - name: Deploy runtime to NODA1
+        shell: bash
+        run: |
+          set -euo pipefail
+          set +x
+          mkdir -p artifacts
+          log="artifacts/deploy-node1-runtime.log"
+          ssh \
+            -i "${SSH_KEY_PATH}" \
+            -o BatchMode=yes \
+            -o IdentitiesOnly=yes \
+            -o StrictHostKeyChecking=accept-new \
+            -o ConnectTimeout=10 \
+            "${SSH_USER}@${SSH_HOST}" \
+            "set -euo pipefail; \
+             cd /opt/microdao-daarion; \
+             if [ -n \"\$(git status --porcelain)\" ]; then \
+               echo 'WARN: dirty git tree on NODA1; skip checkout/pull and continue with gate'; \
+             else \
+               git fetch origin; \
+               git checkout '${DEPLOY_REF:-main}'; \
+               git pull --ff-only origin '${DEPLOY_REF:-main}'; \
+             fi; \
+             if [ '${REDEPLOY_RUNTIME:-false}' = 'true' ]; then \
+               docker compose -f docker-compose.node1.yml up -d --no-deps --build --force-recreate gateway experience-learner; \
+             fi; \
+             git rev-parse HEAD" \
+            | tee "$log"
+
+      - name: Print deploy artifact paths
+        if: always()
+        shell: bash
+        run: |
+          set -euo pipefail
+          ls -la artifacts || true
+
+  phase6_gate:
+    needs: [deploy]
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    env:
+      DEFAULT_SSH_HOST: ${{ secrets.NODA1_SSH_HOST }}
+      DEFAULT_SSH_USER: ${{ secrets.NODA1_SSH_USER }}
+    steps:
+      - name: Resolve SSH target
+        shell: bash
+        run: |
+          set -euo pipefail
+          host="${DEFAULT_SSH_HOST:-}"
+          user="${DEFAULT_SSH_USER:-root}"
+
+          if [ -n "${{ inputs.ssh_host }}" ]; then
+            host="${{ inputs.ssh_host }}"
+          fi
+          if [ -n "${{ inputs.ssh_user }}" ]; then
+            user="${{ inputs.ssh_user }}"
+          fi
+
+          if [ -z "$host" ]; then
+            echo "Missing SSH host (workflow input or secret NODA1_SSH_HOST)" >&2
+            exit 1
+          fi
+
+          echo "SSH_HOST=$host" >> "$GITHUB_ENV"
+          echo "SSH_USER=$user" >> "$GITHUB_ENV"
+
+      - name: Prepare SSH key
+        shell: bash
+        env:
+          SSH_PRIVATE_KEY: ${{ secrets.NODA1_SSH_KEY }}
+        run: |
+          set -euo pipefail
+          set +x
+          if [ -z "${SSH_PRIVATE_KEY:-}" ]; then
+            echo "Missing secret NODA1_SSH_KEY" >&2
+            exit 1
+          fi
+          mkdir -p ~/.ssh
+          chmod 700 ~/.ssh
+          key_path=~/.ssh/noda1_ci_key
+          if printf '%s' "$SSH_PRIVATE_KEY" | grep -q 'BEGIN OPENSSH PRIVATE KEY'; then
+            printf '%s\n' "$SSH_PRIVATE_KEY" | tr -d '\r' > "$key_path"
+          else
+            printf '%s' "$SSH_PRIVATE_KEY" | tr -d '\r' | base64 --decode > "$key_path"
+          fi
+          chmod 600 "$key_path"
+          if ! ssh-keygen -y -f "$key_path" >/dev/null 2>&1; then
+            echo "Invalid SSH private key in NODA1_SSH_KEY" >&2
+            exit 1
+          fi
+          echo "SSH_KEY_PATH=$key_path" >> "$GITHUB_ENV"
+
+      - name: Run phase6 smoke (hard gate)
+        shell: bash
+        run: |
+          set -euo pipefail
+          set +x
+          mkdir -p artifacts
+          for attempt in 1 2; do
+            log="artifacts/phase6-gate-attempt${attempt}.log"
+            if ssh \
+              -i "${SSH_KEY_PATH}" \
+              -o BatchMode=yes \
+              -o IdentitiesOnly=yes \
+              -o StrictHostKeyChecking=accept-new \
+              -o ConnectTimeout=10 \
+              "${SSH_USER}@${SSH_HOST}" \
+              "set -euo pipefail; cd /opt/microdao-daarion; git rev-parse HEAD; make phase6-smoke" \
+              | tee "$log"; then
+              cp "$log" artifacts/phase6-gate.log
+              exit 0
+            fi
+
+            if [ "$attempt" -eq 2 ]; then
+              echo "phase6 gate failed after retry" >&2
+              exit 1
+            fi
+            sleep 15
+          done
+
+      - name: Print gate artifact paths
+        if: always()
+        shell: bash
+        run: |
+          set -euo pipefail
+          ls -la artifacts || true

docs/ops/deploy_gate.md

@@ -0,0 +1,46 @@
+# Deploy Gate (Gitea)
+
+## Purpose
+
+`deploy-node1-runtime` is a hard release gate for NODA1 runtime changes:
+
+1. `deploy` job syncs target git ref on NODA1 (and can optionally rebuild `gateway` + `experience-learner`).
+2. `phase6_gate` job runs `make phase6-smoke` on NODA1.
+3. If `phase6_gate` fails, the workflow fails.
+
+This prevents a deploy from being considered successful without a Phase-6 closed-loop smoke pass.
+
+## Workflow
+
+File: `.gitea/workflows/deploy-node1-runtime.yml`
+
+Manual trigger inputs:
+
+- `deploy_ref` (default: `main`)
+- `redeploy_runtime` (default: `false`)
+- `ssh_host` (optional override)
+- `ssh_user` (optional override, default `root`)
+
+Required repo secrets:
+
+- `NODA1_SSH_HOST`
+- `NODA1_SSH_USER`
+- `NODA1_SSH_KEY`
+
+## Safety notes
+
+- `redeploy_runtime=false` only syncs git on NODA1 and runs gate checks.
+- `redeploy_runtime=true` recreates `gateway` and `experience-learner` containers.
+- If NODA1 git tree is dirty, workflow skips checkout/pull and still enforces `phase6_gate` (safe mode for live nodes).
+- Workflow uses SSH key validation and `IdentitiesOnly=yes` to avoid host key collisions.
+
+## Expected PASS
+
+- `deploy` job: successful SSH sync of selected `deploy_ref`.
+- `phase6_gate` job: `make phase6-smoke` returns PASS.
+- Workflow conclusion: `success`.
+
+## Failure handling
+
+- SSH/network issues: one retry is attempted in gate step.
+- Gate FAIL: treat as release blocker, inspect `artifacts/phase6-gate*.log`.