Closed ben-manes closed 1 year ago
I'll close since this seems to have been broader and the dns fixed.
Hi @ben-manes, thanks for reporting the issue!
The current logic is that in block
mode, harden-runner tries to resolve each of the allowed domains in the pre
step. It reverts the agent (in the pre
step) if it cannot resolve even one of them (maybe because the domain has a typo and is incorrect, or it cannot be resolved at that time). Once the agent is reverted (maybe uninstalled
would be a better term), the hosted runner handles all DNS calls. So, in this case, the job did not fail because of harden runner. As you mentioned, there was probably a broader DNS issue.
Please let me know if you have feedback on the current logic, and we can improve it.
The domain is on the allow list and has worked in prior runs. It appears that Google's dns entry is not resolving, causing the workflow to fail. I can resolve it locally, but have not tried it through that dns. Can the runner be more resilient to a dns outage?
See failed job