github-actions[bot] opened a new pull request, #76138: URL: https://github.com/apache/airflow/pull/76138
airflowctl built its HTTP client without a timeout, so httpx's 5 second default applied to every request. A busy API server can legitimately take longer to answer a PATCH or POST, for example the first "dags update" or "dags trigger" right after the scheduler picked up a run of the same Dag. Write requests are not retried after a timeout, because they may already have been applied, so the CLI exited 1 with "Read timeout error". This flaked in the Airflow CTL integration tests with the PROD image. Use a 30 second read, write and pool timeout and keep the connect timeout at 5 seconds so an unreachable host still fails fast. Set AIRFLOW_CLI_API_TIMEOUT to tune the value in seconds. A timeout passed to Client explicitly still wins. (cherry picked from commit 1bb64c55b6e30bd7cd931bc03732cb85736492f4) Co-authored-by: Shahar Epstein <[email protected]> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
