coulomb/railiance-cluster
5
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
railiance-cluster/install/traefik/README.md
codex bad08fd712
All checks were successful
CI Smoke / host-smoke (push) Successful in 0s
Details
CI Smoke / container-smoke (push) Successful in 1s
Details
Configure bounded HTTPS upload timeout for Forgejo OCI publication 
Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a13590-a16f-7f60-be60-fd71627bc99d
2026-10-13 19:59:10 +02:00

1.5 KiB
Raw Blame History

Railiance01 Traefik upload timeout

Forgejo OCI blob uploads were cut off at 60 seconds with unexpected EOF and HTTP 499/500. The shared HTTPS entry point now declares a bounded 600-second request-read budget via the k3s-managed HelmChartConfig. It applies to all websecure request bodies; it does not disable timeouts or change authentication.

Validated against the installed chart traefik-38.0.201+up38.0.2 and its live values: the rendered difference is exactly one additional argument. Traefik remains 3.6.7. The separate HelmChartConfig survives k3s reconciliation; do not edit the generated Deployment or packaged traefik.yaml.

After node/capacity review, run from this repository:

ssh railiance01 'kubectl apply --server-side --dry-run=server -f -' < install/traefik/railiance01-helmchartconfig.yaml
ssh railiance01 'kubectl diff -f -' < install/traefik/railiance01-helmchartconfig.yaml
ssh railiance01 'kubectl apply --server-side -f -' < install/traefik/railiance01-helmchartconfig.yaml

Wait for the Helm controller to finish and the replacement Traefik pod to become Ready. Verify the actual argument, HTTPS endpoints and a previously failing large OCI publication. Rollback: set the argument to 60s in this source and reapply with the same checks; this restores the former upload limitation.

References: Traefik read timeout and k3s customization.