Verifying and troubleshooting runner connectivity

After you install a CloudRadial AutomationAI runner, it should appear online within a couple of polling intervals. This article explains how to confirm that, how to read the runner's poll logs, and how to interpret the most common failures. It is for the technician operating a runner.

• Confirming the Runner Is Online
• Reading the Poll Logs
• Interpreting Common Failures
• The gpt-5.4 Quota Two-Pass Install

Confirming the Runner Is Online

Within a couple of polling intervals after a successful deploy, the runner shows up as online on the Runners page in your AutomationAI workspace, with its last-seen time and reported version. The first successful poll is what updates those fields, so allow roughly two intervals (about two minutes at the default 60-second cadence) before treating the runner as missing. If it does not appear within about five minutes, check its poll logs.

Reading the Poll Logs

Each runner Function App writes its poll activity to Application Insights. The script runners log lines prefixed with PollTimer: and the agent runner logs lines prefixed with AgentPoll:. On each timer fire the runner signs a check-in and posts it to the control plane to lease work and report its health. A successful tick with no work logs a no-content result; a tick that leases work logs the dispatch it picked up. Look for these lines first to see whether the runner is reaching the control plane at all and what response it is getting.

Interpreting Common Failures

• 401 — the control plane rejected the runner's signature. This means the secret the runner is signing with does not match the hash stored at registration. Re-check the runner secret; rotate it by re-running the installer and pasting the correct value
• 410 — the runner has been revoked. The runner records a retirement marker, stops polling, and short-circuits on later ticks. To bring it back you register a new runner and reinstall
• Egress timeouts — the runner cannot reach the control plane or the package store. In a locked-down environment, confirm the outbound allow-list. See the article on runner network and egress requirements

The gpt-5.4 Quota Two-Pass Install

A fresh subscription has no gpt-5.4 quota, and that quota is checked at deployment preflight, which would otherwise abort the whole install. The installer avoids this by deploying in two passes: pass one lands all infrastructure and the Foundry account, which needs no quota and always succeeds, and pass two deploys the gpt-5.4 model on a best-effort basis. If quota is missing, pass two does not fail the install — it prints an action-required alert and exits cleanly, so the runner still deploys and starts polling.

If you see that alert, AI activities will fail until you grant quota. Request gpt-5.4 — GlobalStandard quota for the subscription and the Foundry account's region from the Azure portal's Quotas area (the alert prints the exact steps), then re-run the same install command. Pass one is a no-op and pass two adds the model. Request a higher quota when the agent runner is enabled, because its bundled Foundry draws on the same quota.