Hi everyone, During my presentation at the Airflow Summit, I conducted a live survey to gather initial community feedback on telemetry in Airflow. While the sample size is small and not statistically representative (22-34 responses per question), I wanted to share the results as one data point in our ongoing discussion about AIP-89.
*Interactive Survey Link (valid for 14 days):* https://www.mentimeter.com/app/presentation/alo47agxwrq66v2kyfgewz6t22vswp1f/edit?source=share-modal ------------------------------ Key Results *1. Have you ever disabled telemetry in a tool?* - Yes: 22 (100%), No: 0 *2. Would you enable telemetry if it was opt-in and transparent?* - Yes: 25 (83%), No: 4 (13%), Don't know: 1 (3%) *3. Should telemetry be opt-in or opt-out?* - Opt-in: 21 (81%), Opt-out: 5 (19%) *4. What's your biggest concern about Airflow telemetry?* - Security: 13 (38%), Privacy: 12 (35%), Performance: 4 (12%), Transparency: 2 (6%), Other: 3 (9%) *5. What would you be willing to share?* Top themes from word cloud: operators/operator types, version info, number of DAGs, metadata, provider information ------------------------------ Takeaways - Strong preference for opt-in approach (validates AIP-89 direction) - Security and privacy are nearly equal concerns (need to address both explicitly) - Data scope aligns well with what AIP-89 proposes (version info, operators, aggregates) - The majority of respondents have disabled telemetry before (indicates need for easy control) Questions and comments welcome! Cheers Bolke AIP-89: https://cwiki.apache.org/confluence/display/AIRFLOW/AIP-89%3A+Privacy-First+Telemetry+for+Apache+Airflow -- Bolke de Bruin [email protected]
