Step replays fail: production capture, and local Free replays #581
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?
What happens
GET /api/steps/<id>/capsuleanswers 409, "could not bundle the checkpoint: Unable to create '/tmp/tmpXXXX/repo.bundle.lock'", for the summary steps of run 204 and run 203. Since braid#493, osfd runs git for a run's folder in a sandbox with a/tmpof its own, so the bundle path osfd made is not there.OSF_SANDBOX=step, Docker Desktop's kernel has no Landlock, and the run's server is refused. The step is then held asengine_down, and the replay waits until its wall time without saying why. A stopped replay also leaves its router's container running.Fix
/dev/net/tun, withOSF_SANDBOX=step-unfenced.Done when a production step can be captured, and replayed on Free on the laptop.
Shipped in #583 (merged as
c9c5e63, live on production), and archived in #584.OSF_SANDBOX=step-unfenced. Replays of run 204's summary step and of a simulated run's ran to their end on Free models./api/v1, which Hetzner no longer serves (/v1does), so its model was never found. The catalog has the new address, ready for a router that reads it.