GitHub user starkmarkus added a comment to the discussion: Airflow cli response times
8 seconds for airflow -h sounds high to me, although the Airflow CLI is not a tiny binary and does have noticeable Python and import overhead. The first thing I would check is whether this is pure Airflow startup time or something environment specific being imported on startup. A few useful checks would be time python -c "import airflow", time airflow version, time airflow info, and PYTHONPROFILEIMPORTTIME=1 airflow -h 2> importtime.log. If import airflow is already slow, then it is probably the Python environment, installed providers, or import time side effects. If only the CLI is slow, then it may be related to command, plugin, or provider discovery. Things I would look for are many installed provider packages, custom plugins, airflow_local_settings.py, slow imports from something on PYTHONPATH, network or secrets backend access during import, or a very slow container, volume, or filesystem. I would also check whether lazy loading is enabled, especially for providers and plugins, because disabling it can make CLI startup much heavier. So I would not call 8 seconds for help output ideal. I would profile the imports first and see whether one provider, plugin, or module is dominating startup time. GitHub link: https://github.com/apache/airflow/discussions/68707#discussioncomment-17363623 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
