CI runner mac-studio fails every job before the first step (docker network pools exhausted, disk full) #133

Open
opened 2026-09-21 14:58:26 +02:00 by clawbot · 0 comments
Collaborator

Every workflow job scheduled onto the mac-studio runner (act runner v0.3.0) fails during job setup, before actions/checkout runs. Nothing in this repo causes it.

Observed today on the milestone PR #114: runs https://git.eeqj.de/sneak/vaultik/actions/runs/6489 and https://git.eeqj.de/sneak/vaultik/actions/runs/6492 both end within a second with Error response from daemon: all predefined address pools have been fully subnetted while creating the job network. Earlier, https://git.eeqj.de/sneak/vaultik/actions/runs/6427 (on #127) failed at the same point with no space left on device. Jobs that land on fsn1app1 run normally (https://git.eeqj.de/sneak/vaultik/actions/runs/6491, the push of next).

Reading: the docker daemon on mac-studio has leaked job networks from earlier runs (each run creates a GITEA-ACTIONS-TASK-<id>_... network) until no address pool is left, and its disk filled. Until it is cleaned, a CI result on any PR here depends on which runner picks the job up; the local make check gate stays the authority for review.

Acceptable, on mac-studio (operator only): remove the stale GITEA-ACTIONS-TASK-* networks (docker network prune), reclaim disk (docker system prune), and re-run the failed jobs; or take that runner out of the pool for this repo. Recommendation: prune both, then re-run https://git.eeqj.de/sneak/vaultik/actions/runs/6492.

model: claude-fable-5

Every workflow job scheduled onto the `mac-studio` runner (act runner v0.3.0) fails during job setup, before `actions/checkout` runs. Nothing in this repo causes it. Observed today on the milestone PR https://git.eeqj.de/sneak/vaultik/pulls/114: runs https://git.eeqj.de/sneak/vaultik/actions/runs/6489 and https://git.eeqj.de/sneak/vaultik/actions/runs/6492 both end within a second with `Error response from daemon: all predefined address pools have been fully subnetted` while creating the job network. Earlier, https://git.eeqj.de/sneak/vaultik/actions/runs/6427 (on https://git.eeqj.de/sneak/vaultik/pulls/127) failed at the same point with `no space left on device`. Jobs that land on `fsn1app1` run normally (https://git.eeqj.de/sneak/vaultik/actions/runs/6491, the push of `next`). Reading: the docker daemon on `mac-studio` has leaked job networks from earlier runs (each run creates a `GITEA-ACTIONS-TASK-<id>_...` network) until no address pool is left, and its disk filled. Until it is cleaned, a CI result on any PR here depends on which runner picks the job up; the local `make check` gate stays the authority for review. Acceptable, on `mac-studio` (operator only): remove the stale `GITEA-ACTIONS-TASK-*` networks (`docker network prune`), reclaim disk (`docker system prune`), and re-run the failed jobs; or take that runner out of the pool for this repo. Recommendation: prune both, then re-run https://git.eeqj.de/sneak/vaultik/actions/runs/6492. model: claude-fable-5
sneak was assigned by clawbot 2026-09-21 14:58:26 +02:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: sneak/vaultik#133