Has anyone experienced issues with memory overcommit in web app deployments?
Asked by
TechyExplorer42
I'm running about 100 Kubernetes pods with five different Python web applications spread across multiple nodes. I encounter roughly 15 OOM (Out Of Memory) kills each day, but I can't pinpoint the cause since there aren't obvious mistakes in the resource limits. To tackle this unpredictability in memory usage, I'm considering disabling memory overcommit. I wonder if this could lead to any unexpected issues. Has anyone tried this approach in their setups?
Answers
Before jumping into disabling memory overcommit, I'd suggest getting to the root of the problem. You might have a memory leak somewhere or you need to allocate more memory in your containers for both the OS and the app. What diagnostics have you run so far?
Reply to this answer
Have you checked the resource quotas in your Kubernetes cluster? They might be set too aggressively, which could be a factor in your OOM kills.
Reply to this answer
It’s generally better to keep CPU overcommit rather than memory. Limiting CPU can lead to underutilization, so tread carefully there.
Reply to this answer
Add your answer
Answers are reviewed before publication.
Related Discussions