fix(ci): robust prod deploy — drop dev-health gate, no tier-clobbering sed, rm -f before recreate
Universal: Pre-Release / Build Pre-Release (${{ inputs.stability || github.ref_name }}) (push) Successful in 42s
Universal: PR Check / Branch Policy (pull_request) Successful in 1s
Universal: PR Check / Wiki Update Reminder (pull_request) Successful in 2s
Universal: PR Check / Validate PR (pull_request) Successful in 8s
Generic: Repo Health / Access control (pull_request) Successful in 2s
Generic: Repo Health / Site Health (pull_request) Has been skipped
Generic: Project CI / Lint & Validate (pull_request) Successful in 36s
Generic: Project CI / Tests (pull_request) Successful in 36s
Universal: PR Check / Secret Scan (pull_request) Successful in 56s
Universal: PR Check / Require Docs Update (pull_request) Failing after 1m11s
PR RC Release / Build RC Release (pull_request) Successful in 1m13s
Universal: PR Check / Build RC Package (pull_request) Has been cancelled
Universal: PR Check / Report Issues (pull_request) Has been cancelled
Generic: Repo Health / Scripts governance (pull_request) Has been cancelled
Generic: Repo Health / Repository health (pull_request) Has been cancelled
Generic: Repo Health / Report: Scripts Governance (pull_request) Has been cancelled
Generic: Repo Health / Report: Repository Health (pull_request) Has been cancelled

The prod deploy (deploy-mokogitea.yml, on push to main) failed the #733 release and
silently corrupted the shared compose. Three fixes:
- Removed the 'Verify dev environment is healthy' gate: it curled git.dev and exit 1
  on failure, so a dev blip false-negatives the PROD deploy (wrong-tier gate; the
  dev->rc->main pipeline + RC env is the real gate).
- Replaced 'sed s|mokogitea:...|:$TAG|' (matched ALL mokogitea service lines and
  clobbered the dev+rc ${MOKOGITEA_*_TAG} env-vars with the prod tag) with the
  env-var pattern: env $TAG_ENV=$TAG docker compose up -- drives only the target
  service, no sed.
- Added 'docker rm -f $CONTAINER' + '-p gitea-dev' + '--force-recreate' before the
  compose up, fixing the 'Container name /mokogitea already in use' conflict.

Aligns deploy-mokogitea.yml with the fixed deploy-dev/deploy-rc pattern. Host compose
was separately restored to env-var form (the bad run had clobbered dev+rc tags).
Long-term: cut over Template-Go#5 deploy-prod.yml. Refs #733, #752.

Claude-Session: https://claude.ai/code/session_01Wsno14cxE49MstXFs9G5KT
This commit is contained in:
2026-07-06 10:55:58 -05:00
parent bc3bb0f778
commit a7220ddb2b
@@ -39,17 +39,10 @@ jobs:
deploy:
runs-on: ubuntu-latest
steps:
- name: Verify dev environment is healthy
run: |
echo "Checking git.dev.mokoconsulting.tech health..."
if curl -sf --max-time 10 https://git.dev.mokoconsulting.tech/api/healthz; then
echo " Dev environment is healthy — proceeding with production deploy"
else
echo "::error::Dev environment is NOT healthy — blocking production deploy"
echo "Deploy to dev first (push to dev branch) and verify it passes before merging to main."
exit 1
fi
# NOTE: removed a "Verify dev environment is healthy" gate that curled
# git.dev and `exit 1` on failure — gating a PROD deploy on a live DEV
# health check is a cross-tier false-negative (a transient dev blip blocked
# prod). Pre-prod validation is the dev -> rc -> main pipeline + RC env.
- name: Checkout source (for version detection)
uses: actions/checkout@v4
with:
@@ -66,17 +59,19 @@ jobs:
ENV="${{ github.event.inputs.environment }}"
fi
if [ "$ENV" = "production" ]; then
echo "compose_dir=/opt/gitea" >> $GITHUB_OUTPUT
echo "compose_dir=/opt/gitea-dev" >> $GITHUB_OUTPUT
echo "container=mokogitea" >> $GITHUB_OUTPUT
echo "source_dir=/opt/gitea/source" >> $GITHUB_OUTPUT
echo "branch=main" >> $GITHUB_OUTPUT
echo "tag=$VERSION" >> $GITHUB_OUTPUT
echo "tag_env=MOKOGITEA_TAG" >> $GITHUB_OUTPUT
else
echo "compose_dir=/opt/gitea-dev" >> $GITHUB_OUTPUT
echo "container=mokogitea-dev" >> $GITHUB_OUTPUT
echo "source_dir=/opt/gitea-dev/source" >> $GITHUB_OUTPUT
echo "branch=dev" >> $GITHUB_OUTPUT
echo "tag=$VERSION-dev" >> $GITHUB_OUTPUT
echo "tag_env=MOKOGITEA_DEV_TAG" >> $GITHUB_OUTPUT
fi
- name: Write deploy key
@@ -95,6 +90,7 @@ jobs:
SOURCE_DIR: ${{ steps.config.outputs.source_dir }}
COMPOSE_DIR: ${{ steps.config.outputs.compose_dir }}
CONTAINER: ${{ steps.config.outputs.container }}
TAG_ENV: ${{ steps.config.outputs.tag_env }}
run: |
HEALTH_FMT='${{ '{{' }}.State.Health.Status${{ '}}' }}'
IMAGE_FMT='Image: ${{ '{{' }}.Config.Image${{ '}}' }}'
@@ -134,8 +130,12 @@ jobs:
echo 'Restarting container...'
cd $COMPOSE_DIR
sed -i 's|${{ env.IMAGE }}:[^ ]*|${{ env.IMAGE }}:$TAG|' docker-compose.yml
docker compose up -d $CONTAINER
# Drive ONLY the target service's tag via its compose env-var (no sed —
# a blanket 'sed s|mokogitea:...|' clobbered the dev + rc service tags too).
# Remove any lingering fixed-name container first (avoids the "name already
# in use" conflict), then force-recreate under the shared compose project.
docker rm -f $CONTAINER 2>/dev/null || true
env $TAG_ENV="$TAG" docker compose -p gitea-dev up -d --force-recreate $CONTAINER
echo 'Health check...'
for i in 1 2 3 4 5 6 7 8; do