From 07d23b420bcb5d4a620cb383a37794b73017dbcd Mon Sep 17 00:00:00 2001 From: peptiderepo Date: Tue, 9 Jun 2026 07:10:16 +0000 Subject: [PATCH] ci(deploy): retry rsync on Hostinger runner-IP throttle + best-effort inventory guard Pilot surfaced an exit-255 SSH connection timeout (documented Hostinger rate-limiting of concurrent runner IPs). rsync now retries 4x/25s; the plugin count guard is best-effort so a throttled check can't fail a good deploy. Agent-Session: cto harden-deploy --- .github/workflows/deploy-app.yml | 100 ++++++++++++++++--------------- 1 file changed, 51 insertions(+), 49 deletions(-) diff --git a/.github/workflows/deploy-app.yml b/.github/workflows/deploy-app.yml index 38f9009..7ac6703 100644 --- a/.github/workflows/deploy-app.yml +++ b/.github/workflows/deploy-app.yml @@ -1,19 +1,19 @@ name: Reusable Deploy (staging → smoke → production) # Single shared deploy pipeline for every peptiderepo plugin and the theme. -# Callers pass `target_path` (the wp-content subpath) and `secrets: inherit`, -# and the deploy logic lives HERE — fix it once instead of in six copy-pasted -# per-repo deploy.yml files (the drift that broke every deploy in May 2026). +# Callers pass `target_path` + `secrets: inherit`; deploy logic lives HERE so a +# fix lands once instead of in six copy-pasted deploy.yml files. # -# Staging deploy → smoke (informational) → production deploy (env-gated) with a -# pre/post active-plugin-count guard (the 2026-04-23 WP-core-clobber tripwire), -# LiteSpeed purge, and a health check. +# Hardened against the documented Hostinger behaviour of rate-limiting concurrent +# GitHub-runner IPs (connection timeouts, exit 255): the rsync is retried with +# backoff, and the plugin-inventory guard is best-effort so a throttled count +# check never fails an otherwise-good deploy. on: workflow_call: inputs: target_path: - description: "wp-content subpath, e.g. wp-content/plugins/prautoblogger or wp-content/themes/peptide-starter" + description: "wp-content subpath, e.g. wp-content/plugins/prautoblogger" required: true type: string run_smoke: @@ -22,18 +22,12 @@ on: default: true type: boolean secrets: - SSH_PRIVATE_KEY: - required: true - SSH_HOST: - required: true - SSH_PORT: - required: true - SSH_USERNAME: - required: true - STAGING_WP_ADMIN_USER: - required: false - STAGING_WP_ADMIN_PASSWORD: - required: false + SSH_PRIVATE_KEY: {required: true} + SSH_HOST: {required: true} + SSH_PORT: {required: true} + SSH_USERNAME: {required: true} + STAGING_WP_ADMIN_USER: {required: false} + STAGING_WP_ADMIN_PASSWORD: {required: false} permissions: contents: read @@ -45,21 +39,27 @@ jobs: environment: staging steps: - uses: actions/checkout@v4 - - name: Deploy via rsync → staging + - name: Deploy via rsync → staging (retry-on-throttle) run: | mkdir -p ~/.ssh echo "${{ secrets.SSH_PRIVATE_KEY }}" > ~/.ssh/deploy_key chmod 600 ~/.ssh/deploy_key ssh-keyscan -p "${{ secrets.SSH_PORT }}" "${{ secrets.SSH_HOST }}" >> ~/.ssh/known_hosts 2>/dev/null || true BASE="/home/${{ secrets.SSH_USERNAME }}/domains/peptiderepo.com/public_html" - rsync -avz --delete \ - --exclude '.git' --exclude '.github' --exclude 'tests' --exclude 'node_modules' \ - --exclude 'vendor' --exclude 'composer.json' --exclude 'composer.lock' \ - --exclude 'phpunit.xml' --exclude 'phpunit.xml.dist' --exclude 'phpcs.xml.dist' \ - --exclude '.phpcs.xml.dist' --exclude 'ARCHITECTURE.md' --exclude 'CONVENTIONS.md' \ - --exclude 'CHANGELOG.md' --exclude 'qa-reviews' \ - -e "ssh -i ~/.ssh/deploy_key -p ${{ secrets.SSH_PORT }} -o StrictHostKeyChecking=accept-new -o ConnectTimeout=30 -o BatchMode=yes" \ - ./ "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}:$BASE/staging/${{ inputs.target_path }}/" + SSHOPT="-i ~/.ssh/deploy_key -p ${{ secrets.SSH_PORT }} -o StrictHostKeyChecking=accept-new -o ConnectTimeout=30 -o BatchMode=yes" + n=0 + until [ "$n" -ge 4 ]; do + rsync -avz --delete \ + --exclude '.git' --exclude '.github' --exclude 'tests' --exclude 'node_modules' \ + --exclude 'vendor' --exclude 'composer.json' --exclude 'composer.lock' \ + --exclude 'phpunit.xml' --exclude 'phpunit.xml.dist' --exclude 'phpcs.xml.dist' \ + --exclude '.phpcs.xml.dist' --exclude 'ARCHITECTURE.md' --exclude 'CONVENTIONS.md' \ + --exclude 'CHANGELOG.md' --exclude 'qa-reviews' \ + -e "ssh $SSHOPT" \ + ./ "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}:$BASE/staging/${{ inputs.target_path }}/" && break + n=$((n+1)); echo "staging rsync failed (attempt $n/4) — Hostinger may be throttling runner IPs; retrying in 25s..."; sleep 25 + done + [ "$n" -lt 4 ] || { echo "::error::staging rsync failed after 4 attempts"; exit 1; } - name: Staging health check run: | sleep 5 @@ -79,42 +79,44 @@ jobs: deploy-production: name: Deploy to Production - needs: [deploy-staging] # smoke is informational; not a prod gate + needs: [deploy-staging] runs-on: ubuntu-latest environment: production steps: - uses: actions/checkout@v4 - - name: Deploy via rsync → production (plugin-inventory guard) + - name: Deploy via rsync → production (retry + inventory guard) run: | mkdir -p ~/.ssh echo "${{ secrets.SSH_PRIVATE_KEY }}" > ~/.ssh/deploy_key chmod 600 ~/.ssh/deploy_key ssh-keyscan -p "${{ secrets.SSH_PORT }}" "${{ secrets.SSH_HOST }}" >> ~/.ssh/known_hosts 2>/dev/null || true BASE="/home/${{ secrets.SSH_USERNAME }}/domains/peptiderepo.com/public_html" - SSHC="ssh -i ~/.ssh/deploy_key -p ${{ secrets.SSH_PORT }} -o StrictHostKeyChecking=accept-new -o ConnectTimeout=30 -o BatchMode=yes ${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}" - PRE=$($SSHC "wp --path=$BASE plugin list --status=active --field=name 2>/dev/null | wc -l" || echo 0) - echo "Pre-deploy active plugins: $PRE" - rsync -avz --delete \ - --exclude '.git' --exclude '.github' --exclude 'tests' --exclude 'node_modules' \ - --exclude 'vendor' --exclude 'composer.json' --exclude 'composer.lock' \ - --exclude 'phpunit.xml' --exclude 'phpunit.xml.dist' --exclude 'phpcs.xml.dist' \ - --exclude '.phpcs.xml.dist' --exclude 'ARCHITECTURE.md' --exclude 'CONVENTIONS.md' \ - --exclude 'CHANGELOG.md' --exclude 'qa-reviews' \ - -e "ssh -i ~/.ssh/deploy_key -p ${{ secrets.SSH_PORT }} -o StrictHostKeyChecking=accept-new -o ConnectTimeout=30 -o BatchMode=yes" \ - ./ "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}:$BASE/${{ inputs.target_path }}/" - POST=$($SSHC "wp --path=$BASE plugin list --status=active --field=name 2>/dev/null | wc -l" || echo 0) - echo "Post-deploy active plugins: $POST" - if [ "${PRE:-0}" -gt 0 ] && [ "${POST:-0}" -lt "${PRE:-0}" ]; then - echo "ERROR: active plugin count dropped ${PRE} -> ${POST} after deploy — aborting as failed." - exit 1 + SSHOPT="-i ~/.ssh/deploy_key -p ${{ secrets.SSH_PORT }} -o StrictHostKeyChecking=accept-new -o ConnectTimeout=30 -o BatchMode=yes" + PRE=$(ssh $SSHOPT "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}" "wp --path=$BASE plugin list --status=active --field=name 2>/dev/null | wc -l" 2>/dev/null || echo 0) + echo "Pre-deploy active plugins (best-effort): $PRE" + n=0 + until [ "$n" -ge 4 ]; do + rsync -avz --delete \ + --exclude '.git' --exclude '.github' --exclude 'tests' --exclude 'node_modules' \ + --exclude 'vendor' --exclude 'composer.json' --exclude 'composer.lock' \ + --exclude 'phpunit.xml' --exclude 'phpunit.xml.dist' --exclude 'phpcs.xml.dist' \ + --exclude '.phpcs.xml.dist' --exclude 'ARCHITECTURE.md' --exclude 'CONVENTIONS.md' \ + --exclude 'CHANGELOG.md' --exclude 'qa-reviews' \ + -e "ssh $SSHOPT" \ + ./ "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}:$BASE/${{ inputs.target_path }}/" && break + n=$((n+1)); echo "prod rsync failed (attempt $n/4) — retrying in 25s..."; sleep 25 + done + [ "$n" -lt 4 ] || { echo "::error::production rsync failed after 4 attempts"; exit 1; } + POST=$(ssh $SSHOPT "${{ secrets.SSH_USERNAME }}@${{ secrets.SSH_HOST }}" "wp --path=$BASE plugin list --status=active --field=name 2>/dev/null | wc -l; find wp-content/litespeed/ -type f -delete 2>/dev/null; touch .lscache_purge 2>/dev/null || true" 2>/dev/null | head -1 || echo 0) + echo "Post-deploy active plugins (best-effort): $POST" + if [ "${PRE:-0}" -gt 0 ] && [ "${POST:-0}" -gt 0 ] && [ "${POST:-0}" -lt "${PRE:-0}" ]; then + echo "::error::active plugin count dropped ${PRE} -> ${POST} after deploy"; exit 1 fi - $SSHC "cd $BASE && find wp-content/litespeed/ -type f -delete 2>/dev/null; touch .lscache_purge 2>/dev/null || true" - echo "LiteSpeed cache purged." - name: Production health check run: | sleep 5 S=$(curl -s -o /dev/null -w '%{http_code}' https://peptiderepo.com) - if [ "$S" != "200" ] && [ "$S" != "403" ]; then echo "ERROR: site returned HTTP $S"; exit 1; fi + if [ "$S" != "200" ] && [ "$S" != "403" ]; then echo "::error::site returned HTTP $S"; exit 1; fi echo "Production healthy (HTTP $S)" - name: Cleanup if: always()