Deploy Workflow Hardening (Implemented 2026-02-07)
Status: Implemented | Scope: Make routine deploys boring, repeatable, and crawl-safe on the single VPS.
Outcomes
- Added a safe deploy wrapper with an explicit backend-only mode for external frontend outages:
scripts/vps-hetzdeploy.sh(--mode full|backend-only)- Added an installer so operators use a real command instead of fragile shell aliases:
scripts/vps-install-hetzdeploy.sh(installshetzdeployunder/usr/local/bin/)- Preserved deploy safety properties:
- refuses dirty trees by default
- keeps worker restart crawl-safe (skips restart while jobs are running)
- keeps baseline drift checks enabled by default
Canonical Docs Updated
docs/operations/playbooks/core/deploy-and-verify.mddocs/deployment/production-single-vps.mddocs/deployment/production-rollout-checklist.md
Validation
make ciremains green.- On the VPS,
type hetzdeployreports/usr/local/bin/hetzdeploy(not an alias).
Historical Context
Detailed iteration is preserved in git history and associated incident notes.