Andrushika opened a new pull request, #72523: URL: https://github.com/apache/airflow/pull/72523
## Why `test_airflowctl_commands[tasks failed-deps0]` fails now and then with `Read timeout error` after three retries ([example run](https://github.com/apache/airflow/actions/runs/33647995987/job/100325459131)). The compose stack has one api-server process, and with the default Celery concurrency of 16 a trigger of `example_bash_operator` forks more than ten task processes at once. On the 4 vCPU runner the api-server then gets no CPU for 20 to 35 seconds, and any airflowctl command in that window times out. ```mermaid sequenceDiagram participant T as pytest (airflowctl) participant A as api-server (1 process) participant W as celery worker Note over T,W: concurrency 16 (before) T->>A: dags trigger + unpause W->>W: fork 10+ task processes at once Note over A: starved of CPU for 20-35s T->>A: tasks failed-deps A--xT: 5s read timeout, x3 T->>T: exit 1 (flaky) Note over T,W: concurrency 2 (after) T->>A: dags trigger + unpause W->>W: fork 2 task processes T->>A: tasks failed-deps A-->>T: 200 ``` ## What Set `AIRFLOW__CELERY__WORKER_CONCURRENCY=2` in the `.env` the test conftest already writes, so tasks start two at a time. Measured locally with all containers pinned to one core plus 6 busy-loop containers, probing `/api/v2/monitor/health` every second: | Worker concurrency | Max health latency | api-server log gap > 6s | Suite | |---|---|---|---| | 16 (default) | 17.0s / 18.5s | yes, 19s | 170 passed | | 2 | 2.1s / under 1s | no | 170 passed | I did not get a red run locally, the stall landed on the test's own Dag run polling instead of an airflowctl command. **The trade-off is a slower suite**, about **6 seconds** more on my machine (64s to 70s), since tasks now run two at a time. Happy to share the scripts and logs. --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes — Claude Code (Fable 5.1) Generated-by: Claude Code (Fable 5.1) following [the guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions) -- 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]
