Commitec16b67unknown_key

fix(deploy): repair Hetzner git remote on every deploy — was 404ing 16h

fix(deploy): repair Hetzner git remote on every deploy — was 404ing 16h

ROOT CAUSE of the persistent admin-dashboard reload loop:

The Hetzner box at /opt/gluecron had its git `origin` remote set to
`https://gluecron.com/ccantynz/Gluecron.com.git`. The gluecron server
itself returns 404 for that URL (likely owner-case mismatch +
push-protection on the self-hosted repo). So on every deploy since
2026-05-15 10:27 UTC, the `git fetch` step in the SSH script failed:

  remote: Repository not found
  fatal: repository 'https://gluecron.com/ccantynz/Gluecron.com.git/' not found
  Process exited with status 128

That meant every commit to main for the past 16+ hours — including
d7ba05d, 67f64a3, 2e8a4d5, 170ddb2, 904927d, and 44fe49b (the four
attempts to kill the AA reload loop) — never reached the live host.
The systemd unit kept running the stale code from yesterday, which
still served the buggy PWA layout, which still produced the SW
collision loop, which the user has been staring at for hours.

Fix: repair the remote on every deploy. The GitHub Actions runner
has GITHUB_TOKEN with `contents: read` (confirmed in the previous
run's permissions block) which is enough for fetch. We re-point
origin to GitHub HTTPS with the token interpolated as basic-auth
on every deploy. `git remote set-url` is idempotent — safe to run
forever, fixes the box automatically on the next deploy.

After this lands, the next push to main will:
  1. Set the remote URL to GitHub (with auth)
  2. fetch + reset --hard onto 44fe49b (+ this commit)
  3. Run bun install / build / migrate / restart
  4. Serve the NEW layout with no PWA service workers
  5. The kill-switch in the new layout unregisters any SW left
     behind in browsers, ending the loop within one page load

No code path other than the deploy script touches the remote URL,
so no other changes needed.

Claude committed on May 16, 2026Parent: 44fe49b

1 file changed+16−1ec16b67db3d827f43e1b96b25e37c05daaf10f11

1 changed file+16−1

Modified.github/workflows/hetzner-deploy.yml+16−1View fileUnified Split

@@ -138,6 +138,16 @@ jobs:

138138 APP_BASE_URL: ${{ secrets.APP_BASE_URL || 'https://gluecron.com' }}

139139 GH_RUN_ID: ${{ github.run_id }}

140140 GH_SHA: ${{ github.sha }}

141 # FIX (2026-05-16): the box's git remote was pointing at the

142 # self-hosted URL `https://gluecron.com/ccantynz/Gluecron.com.git`

143 # which the gluecron server itself returns 404 for (owner-case

144 # mismatch + push protection). That meant 16+ hours of failed

145 # deploys — the live site has been running stale since

146 # 2026-05-15 10:27 UTC. We now re-point the remote at GitHub on

147 # every deploy and authenticate with the workflow's GITHUB_TOKEN

148 # (which has contents:read — enough for fetch). Idempotent.

149 GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

150 GH_REPO: ${{ github.repository }}

141151 with:

142152 host: ${{ secrets.HETZNER_HOST }}

143153 username: ${{ secrets.HETZNER_USER }}

@@ -148,7 +158,7 @@ jobs:

148158 # share one SSH session across all phases (git pull → install →

149159 # build → migrate → restart) so the in-script curl posts give us

150160 # the live timeline a black-box step boundary can't.

151 envs: DEPLOY_EVENT_TOKEN,APP_BASE_URL,GH_RUN_ID,GH_SHA

161 envs: DEPLOY_EVENT_TOKEN,APP_BASE_URL,GH_RUN_ID,GH_SHA,GH_TOKEN,GH_REPO

152162 script: |

153163 set -euo pipefail

154164 cd /opt/gluecron

@@ -174,6 +184,11 @@ jobs:

174184

175185 notify_step "git-pull" "in_progress"

176186 GP_START=$(date +%s)

187 # Repair the remote on every deploy. The box was historically

188 # pointed at the self-hosted gluecron-side URL which 404s; we

189 # now force it back to GitHub HTTPS with a workflow-scoped PAT

190 # for auth. `set-url` is idempotent.

191 git remote set-url origin "https://x-access-token:${GH_TOKEN}@github.com/${GH_REPO}.git"

177192 git fetch --prune origin main

178193 git reset --hard origin/main

179194 new_sha=$(git rev-parse HEAD)

180195