cd.yml (Build and deploy) has failed on every commit since ~2026-06-26 (runs 288, 290, 292…). The routing/ingestion images therefore haven't been rebuilt — the routing pod is frozen on d39a18d. This silently blocked the #73 instrumentation fix from deploying via CI.
Root cause
The gitea.d-ma.be → git.d-ma.be host rename left the Go module path stale. The ingestion image build dies at go mod download:
go: gitea.d-ma.be/mathias/mcp-chassis@v0.1.0: unrecognized import path
"gitea.d-ma.be/mathias/mcp-chassis": parse https://gitea.d-ma.be/mathias/mcp-chassis?go-get=1:
no go-import meta tags (meta tag git.d-ma.be/mathias/mcp-chassis did not match import path
gitea.d-ma.be/mathias/mcp-chassis)
The server now serves the go-import meta tag as git.d-ma.be/..., which no longer matches the requested gitea.d-ma.be/... import path. Local builds pass only because the module is cached; a clean CI build fetches fresh and fails.
Scope
mcp-chassis repo still declares module gitea.d-ma.be/mathias/mcp-chassis (its go.mod line 1).
ingestion/go.mod requires gitea.d-ma.be/mathias/mcp-chassis v0.1.0; imported in ingestion/cmd/server/main.go.
The top-level module (github.com/mathiasbq/supervisor, used by the routing image) has zerogitea.d-ma.be deps — so the routing image builds clean; only the ingestion build (run first in cd.yml) is affected.
Fix (cross-repo migration)
mcp-chassis: rename module → git.d-ma.be/mathias/mcp-chassis, update internal imports, commit, tag v0.2.0 (check for other consumers first).
ingestion: update import + go.mod require to the new path/tag, go mod tidy.
Confirm cd.yml goes green; remove this workaround once CI builds both images again.
Interim
The #73 routing fix (da9bdc4) was deployed by building the routing image manually on koala (clean build, no mcp-chassis dep) and bumping the deployment tag — documented on #73. This issue tracks restoring the normal CI path and the ingestion image.
Refs: #73, #35, infra rename commit (fb58b25 in .context).
## Symptom
`cd.yml` (Build and deploy) has failed on **every** commit since ~2026-06-26 (runs 288, 290, 292…). The routing/ingestion images therefore haven't been rebuilt — the routing pod is frozen on `d39a18d`. This silently blocked the #73 instrumentation fix from deploying via CI.
## Root cause
The `gitea.d-ma.be → git.d-ma.be` host rename left the Go module path stale. The ingestion image build dies at `go mod download`:
```
go: gitea.d-ma.be/mathias/mcp-chassis@v0.1.0: unrecognized import path
"gitea.d-ma.be/mathias/mcp-chassis": parse https://gitea.d-ma.be/mathias/mcp-chassis?go-get=1:
no go-import meta tags (meta tag git.d-ma.be/mathias/mcp-chassis did not match import path
gitea.d-ma.be/mathias/mcp-chassis)
```
The server now serves the go-import meta tag as `git.d-ma.be/...`, which no longer matches the requested `gitea.d-ma.be/...` import path. Local builds pass only because the module is cached; a clean CI build fetches fresh and fails.
## Scope
- `mcp-chassis` repo still declares `module gitea.d-ma.be/mathias/mcp-chassis` (its `go.mod` line 1).
- `ingestion/go.mod` requires `gitea.d-ma.be/mathias/mcp-chassis v0.1.0`; imported in `ingestion/cmd/server/main.go`.
- The **top-level module** (`github.com/mathiasbq/supervisor`, used by the routing image) has **zero** `gitea.d-ma.be` deps — so the routing image builds clean; only the ingestion build (run first in `cd.yml`) is affected.
## Fix (cross-repo migration)
1. `mcp-chassis`: rename module → `git.d-ma.be/mathias/mcp-chassis`, update internal imports, commit, tag `v0.2.0` (check for other consumers first).
2. `ingestion`: update import + `go.mod` require to the new path/tag, `go mod tidy`.
3. Confirm `cd.yml` goes green; remove this workaround once CI builds both images again.
## Interim
The #73 routing fix (`da9bdc4`) was deployed by building the routing image manually on koala (clean build, no mcp-chassis dep) and bumping the deployment tag — documented on #73. This issue tracks restoring the normal CI path and the ingestion image.
Refs: #73, #35, infra rename commit (`fb58b25` in `.context`).
mcp-chassis renamed module git.d-ma.be/mathias/mcp-chassis, tagged v0.2.0 (69d389d). Old v0.1.0 (old path) left intact so unmigrated consumers don't break.
ingestion (this repo) migrated to git.d-ma.be/mathias/mcp-chassis v0.2.0 + go mod tidy (b9d0331). Fresh go mod download now resolves; build + tests green. This is the fetch CI was failing on, so CD should go green again.
Other consumers — latent, not yet migrated
A repo-wide scan found two more importers of the old path. They still pin v0.1.0 (immutable tag, old module path) so they build from cache today, but each will fail the same go mod download the next time its image rebuilds from clean:
gitea-mcp — cmd/gitea-mcp/main.go + go.mod
tapir — internal/web/oidc/oidc.go (+ go.mod)
Both need the same one-line treatment: bump import + go.mod require to git.d-ma.be/mathias/mcp-chassis v0.2.0, go mod tidy. Worth doing proactively before their next deploy rather than discovering it red. (Out of scope for the #35 critical path; tracked here.)
## Migration progress (2026-06-30)
- **mcp-chassis** renamed `module git.d-ma.be/mathias/mcp-chassis`, tagged **v0.2.0** (`69d389d`). Old `v0.1.0` (old path) left intact so unmigrated consumers don't break.
- **ingestion** (this repo) migrated to `git.d-ma.be/mathias/mcp-chassis v0.2.0` + `go mod tidy` (`b9d0331`). Fresh `go mod download` now resolves; build + tests green. This is the fetch CI was failing on, so CD should go green again.
### Other consumers — latent, not yet migrated
A repo-wide scan found two more importers of the old path. They still pin `v0.1.0` (immutable tag, old module path) so they build from cache today, but **each will fail the same `go mod download` the next time its image rebuilds from clean**:
- `gitea-mcp` — `cmd/gitea-mcp/main.go` + `go.mod`
- `tapir` — `internal/web/oidc/oidc.go` (+ go.mod)
Both need the same one-line treatment: bump import + `go.mod` require to `git.d-ma.be/mathias/mcp-chassis v0.2.0`, `go mod tidy`. Worth doing proactively before their next deploy rather than discovering it red. (Out of scope for the #35 critical path; tracked here.)
After the mcp-chassis v0.2.0 migration + ingestion bump (b9d0331), CD run 294 went green — first successful build since the rename. It built ingestion:b9d0331 + routing:b9d0331 and auto-bumped both infra deployment tags; flux rolled them out. The #73 routing fix reached prod through the normal CI path as intended.
Remaining (kept open under this issue): gitea-mcp and tapir still import the old gitea.d-ma.be/mathias/mcp-chassis path (pinned v0.1.0). They build from cache today but will hit the same go mod download failure on their next clean rebuild. Each needs the one-line bump to git.d-ma.be/mathias/mcp-chassis v0.2.0 + go mod tidy before its next deploy.
## hyperguild CD restored (2026-06-30)
After the mcp-chassis v0.2.0 migration + ingestion bump (`b9d0331`), CD run **294 went green** — first successful build since the rename. It built `ingestion:b9d0331` + `routing:b9d0331` and auto-bumped both infra deployment tags; flux rolled them out. The #73 routing fix reached prod through the normal CI path as intended.
Remaining (kept open under this issue): **gitea-mcp** and **tapir** still import the old `gitea.d-ma.be/mathias/mcp-chassis` path (pinned `v0.1.0`). They build from cache today but will hit the same `go mod download` failure on their next clean rebuild. Each needs the one-line bump to `git.d-ma.be/mathias/mcp-chassis v0.2.0` + `go mod tidy` before its next deploy.
Already fixed by b9d0331 (fix(ingestion): migrate mcp-chassis import to git.d-ma.be path) — ingestion/go.mod now requires git.d-ma.be/mathias/mcp-chassis v0.2.0, imported in ingestion/cmd/server/main.go as chassisauth. CD (cd.yml) has been green on every push since run #318 (2026-07-09), confirmed via workflow_run_status. Closing.
Already fixed by b9d0331 (`fix(ingestion): migrate mcp-chassis import to git.d-ma.be path`) — `ingestion/go.mod` now requires `git.d-ma.be/mathias/mcp-chassis v0.2.0`, imported in `ingestion/cmd/server/main.go` as `chassisauth`. CD (`cd.yml`) has been green on every push since run #318 (2026-07-09), confirmed via `workflow_run_status`. Closing.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Symptom
cd.yml(Build and deploy) has failed on every commit since ~2026-06-26 (runs 288, 290, 292…). The routing/ingestion images therefore haven't been rebuilt — the routing pod is frozen ond39a18d. This silently blocked the #73 instrumentation fix from deploying via CI.Root cause
The
gitea.d-ma.be → git.d-ma.behost rename left the Go module path stale. The ingestion image build dies atgo mod download:The server now serves the go-import meta tag as
git.d-ma.be/..., which no longer matches the requestedgitea.d-ma.be/...import path. Local builds pass only because the module is cached; a clean CI build fetches fresh and fails.Scope
mcp-chassisrepo still declaresmodule gitea.d-ma.be/mathias/mcp-chassis(itsgo.modline 1).ingestion/go.modrequiresgitea.d-ma.be/mathias/mcp-chassis v0.1.0; imported iningestion/cmd/server/main.go.github.com/mathiasbq/supervisor, used by the routing image) has zerogitea.d-ma.bedeps — so the routing image builds clean; only the ingestion build (run first incd.yml) is affected.Fix (cross-repo migration)
mcp-chassis: rename module →git.d-ma.be/mathias/mcp-chassis, update internal imports, commit, tagv0.2.0(check for other consumers first).ingestion: update import +go.modrequire to the new path/tag,go mod tidy.cd.ymlgoes green; remove this workaround once CI builds both images again.Interim
The #73 routing fix (
da9bdc4) was deployed by building the routing image manually on koala (clean build, no mcp-chassis dep) and bumping the deployment tag — documented on #73. This issue tracks restoring the normal CI path and the ingestion image.Refs: #73, #35, infra rename commit (
fb58b25in.context).Migration progress (2026-06-30)
module git.d-ma.be/mathias/mcp-chassis, tagged v0.2.0 (69d389d). Oldv0.1.0(old path) left intact so unmigrated consumers don't break.git.d-ma.be/mathias/mcp-chassis v0.2.0+go mod tidy(b9d0331). Freshgo mod downloadnow resolves; build + tests green. This is the fetch CI was failing on, so CD should go green again.Other consumers — latent, not yet migrated
A repo-wide scan found two more importers of the old path. They still pin
v0.1.0(immutable tag, old module path) so they build from cache today, but each will fail the samego mod downloadthe next time its image rebuilds from clean:gitea-mcp—cmd/gitea-mcp/main.go+go.modtapir—internal/web/oidc/oidc.go(+ go.mod)Both need the same one-line treatment: bump import +
go.modrequire togit.d-ma.be/mathias/mcp-chassis v0.2.0,go mod tidy. Worth doing proactively before their next deploy rather than discovering it red. (Out of scope for the #35 critical path; tracked here.)hyperguild CD restored (2026-06-30)
After the mcp-chassis v0.2.0 migration + ingestion bump (
b9d0331), CD run 294 went green — first successful build since the rename. It builtingestion:b9d0331+routing:b9d0331and auto-bumped both infra deployment tags; flux rolled them out. The #73 routing fix reached prod through the normal CI path as intended.Remaining (kept open under this issue): gitea-mcp and tapir still import the old
gitea.d-ma.be/mathias/mcp-chassispath (pinnedv0.1.0). They build from cache today but will hit the samego mod downloadfailure on their next clean rebuild. Each needs the one-line bump togit.d-ma.be/mathias/mcp-chassis v0.2.0+go mod tidybefore its next deploy.Already fixed by
b9d0331(fix(ingestion): migrate mcp-chassis import to git.d-ma.be path) —ingestion/go.modnow requiresgit.d-ma.be/mathias/mcp-chassis v0.2.0, imported iningestion/cmd/server/main.goaschassisauth. CD (cd.yml) has been green on every push since run #318 (2026-07-09), confirmed viaworkflow_run_status. Closing.