interactive: inspect repeated CI image-build failure #141
Labels
No labels
burndown-2026-06
headless
autonomy
async-consult
autonomy
epic
autonomy
headless
autonomy
live-collab
c#
priority
P0
priority
P1
priority
P2
priority
P3
priority
P4
role/ai
role/creator
role/design
role/director
role/engineer
role/exec
role/human
role/ops
role/qa
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
coilyco-gaming/eco-app#141
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?
The new per-mod matrix is landed in commit
9154779and locally passes all 523 tests, lint, mypy, pre-commit, and Ward validation. Forgejo run 105 failed in build-image task 14518 after 11m02s, before the four publish-mod matrix tasks started: https://forgejo.coilysiren.me/coilyco-gaming/eco-app/actions/runs/105\n\nThe preceding combined workflow failed in the same image-build phase in run 104, task 14491: https://forgejo.coilysiren.me/coilyco-gaming/eco-app/actions/runs/104\n\nThe sanctioned Actions log endpoint returns HTTP 404 for both completed jobs using either the run index or ActionRun id, so this surface cannot inspect the failing Docker step. An attended operator should open run 105 in Forgejo, capture the exact failing build-image log, and verify whether the runner hit storage, network, build timeout, or a Docker build error. Do not probe this with another push. The four package jobs are correctly dependency-gated and did not run.Root cause confirmed from run 105 task 14518. The image build reaches
mods/jobs/src, thendotnet restorefails with NU1301 after timing out on the publichttps://api.nuget.org/v3/index.jsonservice index. This is the same known runner-egress blocker from #69, not an image tag, package token, or per-mod matrix failure. The matrix never starts becausebuild-imagefails first.Application-side retries were already tried and retired because they turned outages into 20 to 40 minute sinks. The durable infrastructure prerequisite is coilyco-flight-deck/infrastructure#571. Keep this issue open until an operator verifies the internal NuGet feed from both runner and DinD build paths, then rerun the image build and four independent package publishers.
Canonical infrastructure blocker corrected to coilyco-flight-deck/infrastructure#563. The earlier Telegram curl-container test already showed destination-sensitive failure inside the Forgejo DinD path while normal Kubernetes pods could reach the same external service. NuGet is now a second affected destination. infrastructure#571 is closed as a narrower duplicate.
The image and four package jobs remain correctly blocked pending live verification of both
api.telegram.organdapi.nuget.orgfrom the general DinD and dedicated Docker-build paths.Resolved. Infrastructure main now deploys a two-replica allowlisted Forgejo runner HTTPS proxy and injects FORGEJO_EGRESS_PROXY into both runner pools. Flux applied revision b2ec6ec with both proxy replicas ready, and the operator recycled both runner StatefulSets. Eco commit
58ab8bdpasses the proxy into Docker's predefined HTTP_PROXY and HTTPS_PROXY build arguments. Workflow-dispatch run 107 completed successfully from 07:34:05Z to 07:45:08Z. Test, frontend, build-image, publish-eco-jobs-tracker, publish-eco-replay, publish-eco-store-exporter, and publish-eco-telemetry all succeeded. This verifies the in-container NuGet restore and all four independent Forgejo package publishers.