GitLab Runner: Find the Proxy Settings for the Failed Job Stage
Runner control, image retrieval, repository fetch and in-job downloads can originate from different processes.
| First failure | Input to inspect |
|---|---|
| Runner contacts GitLab | Runner service environment |
| Pull job image | Docker daemon/pull service |
| Retrieve repository | Actual Git client/helper environment |
| Job command download | Container and that tool's own client settings |
A job variable is too late for an image pull already required to create the container. A successful Runner control connection likewise cannot establish isolated-job reachability.
For 127.0.0.1:3128, locate the listener: job-container loopback is not the Runner host. Give only the intended job networks access to a host helper and manage its upstream account deliberately.
Apply the deployed Runner/executor's supported interface, then the actual downloader's versioned options. Keep stage, requesting process and attempted endpoint in redacted diagnostics rather than putting account-specific guesses into shared wrappers.