EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 20·5 min read
GitHub, autoscaling, and the component substitution fallacy
TL;DR
GitHub's recent outage was linked to a misconfigured autoscaling policy that failed to account for Istio sidecar limits, leading to resource saturation. This highlights the importance of correctly defining autoscaling metrics to ensure service reliability.
✦ Why It Matters
Review and adjust your autoscaling policies to include all relevant metrics for better service reliability.
Key Takeaways
- Ensure autoscaling policies consider all components, including sidecars.
- Conduct regular load testing to validate autoscaling effectiveness.
- Investigate traffic patterns to understand service load better.
- Avoid focusing solely on individual component defects; consider system interactions.
How It Works
Autoscaling adjusts resources based on load metrics, which can include CPU utilization or request volume. A well-configured policy should account for all components, including sidecars like Istio, to prevent saturation.