Move CI and runtime off Node 22 before its April 2027 end-of-life #47
Labels
No labels
automerge
bug
dependencies
duplicate
help wanted
invalid
lang
rust
migrations
opencode
question
security
type
documentation
type
enhancement
type
maintenance
type
refactor
vulnerability
wontfix
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
Personal/backstage#47
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Goal
Move the Backstage CI workflow, the backend Docker image, and the documented runtime off Node 22 once Node 24 or Node 26 passes the full validation suite.
Context
Node 22 reaches end of life in April 2027, about five months after this issue was opened. Personal/backstage#42 pins
.woodpecker/application-checks.yamlandpackages/backend/Dockerfileto Node 22 because newer majors fail the Playwright suite here:better-sqlite3for that ABI)The same failure appears in Woodpecker (#123, #125 on
node:26-trixie). The app and backend start, and typecheck, lint, unit tests, and both builds pass on Node 26. The failing step is the guest sign-in against/api/auth/guest/refresh, which returns 404 for the whole retry window while the welcome page stays visible. The backend starts without errors when run alone on Node 26, so the cause is still unknown.Other facts:
node:24-trixietonode:26-trixiein CI and the Dockerfile (Personal/backstage#37). Pinning to 22 in the repository will be overwritten unless the Renovate configuration keeps this workflow on Node 22.enginesallows22 || 24 || 26; README and AGENTS.md currently say Node 22 or 24.Scope
Acceptance criteria
application-checkspipeline passes on the new Node major on Woodpecker, including E2E.bun install --frozen-lockfile.packages/backend/Dockerfileuses the same major; the image starts and serves the frontend.enginesonly after the new major is validated.Deadline
Node 22 end-of-life: April 2027. Complete before then.
Related
Correction to the context above. The Woodpecker E2E failures on
node:26-trixie(#123, #125, #126) were caused by mixed Node majors within one pipeline, not by Node 26 itself. The install step ran on the image in the pinned list while two steps still usednode:24-trixie, sobetter-sqlite3built for one ABI was loaded by another runtime (ABI 127 vs 137 in the backend log).Fix in
5f6b081: every application-check step now usesnode:22-trixie.The local Node 24 and Node 26 failures in the table were measured with a correctly built binary for that ABI, so they are not explained by this and still need to be re-run on a fully consistent setup. Node 22 is still the validated runtime until that re-test is done.
Implemented the Node 24 migration on branch
fix/node-runtime-validation: all application-check steps, the backend Dockerfile,package.jsonengines, the Nix dev shell, README, and AGENTS.md now use Node 24. Added a Renovate rule limiting official Node Docker image tags to 24.x until a new major is explicitly validated.Validation evidence:
nix develop --command node --versionreports v24.21.0.node:24-trixie-slimbackend image and ran its default production command against a temporary PostgreSQL container;GET /returned HTTP 200 and served the frontend root.application-checkssteps andpr-review.Committed as
63b9984and pushed asfix/node-runtime-validation. PR #50 is open; keep this issue open until it is merged.Completed: PR #50 was merged as
7c24d93413f209e42f725429cee870885eeec798. Woodpecker PR pipeline #132 passed allapplication-checksincluding E2E, pluspr-review; the post-merge Renovate validation also passed. Local Node 24 full validation and the production-image smoke test passed, so all acceptance criteria are satisfied. The earlier overlapping E2E UI timeouts did not reproduce in isolated Node 24/26 runs.