rock/homelab-frontend
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(ci): drop ineffective --network host option, runner sets it globally now
Build / Build and push image (push) Failing after 14s
Details
CI / Test, vet, build (push) Failing after 12m47s
Details

Browse Source
This commit is contained in:
Story Crater Bot
2026-08-21 16:14:16 -07:00
parent aef122b854
commit 3d63df9ba8
1 changed files with 17 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.gitea/workflows/build.yaml
+17 -10
View File
@@ -18,16 +18,23 @@ jobs:
runs-on: docker runs-on: docker
container: container:
image: docker:27-cli image: docker:27-cli
# The runner's dind sidecar shares the pod network, so the daemon is # No `options: --network host` here -- act_runner ignores that per-job
# reachable on localhost -- that part needs no extra config. The mTLS # override and always decides the job container's network from its own
# certs dind generates at startup do NOT come along for free, though: # config.yaml (container.network), which defaults to an isolated
# they live in an emptyDir mounted into the runner/dind containers, not # per-job bridge. Confirmed live: with that default, DOCKER_HOST=
# into containers a workflow spins up. Job containers get no bind mounts # tcp://localhost:2376 resolved to the job container itself, not dind,
# at all unless the path is in the runner's container.valid_volumes # so every command past `docker login` (which never touches DOCKER_HOST
# allowlist (empty by default -- confirmed live, this exact mount was # -- it only talks to the registry) failed with "Cannot connect to the
# rejected until the runner's Helm chart added a config.yaml scoping # Docker daemon". host networking is set once, for every job, in the
# valid_volumes to exactly this path). # runner's own Helm chart.
options: --network host #
# The mTLS certs dind generates at startup are a separate gap: they
# live in an emptyDir mounted into the runner/dind containers, not into
# containers a workflow spins up. Job containers get no bind mounts at
# all unless the path is in the runner's container.valid_volumes
# allowlist (empty by default -- this exact mount was rejected until
# the runner's Helm chart added a config.yaml scoping valid_volumes to
# exactly this path).
volumes: volumes:
- /docker-certs/client:/docker-certs/client:ro - /docker-certs/client:/docker-certs/client:ro
env: env: