TRG 5.04 - Resource Management
Status | Created | Post-History |
---|---|---|
Active | 23-Feb-2022 | Fix mixup between CPU and Memory description |
Active | 10-Nov-2022 | Initial release |
Why
To give guidance on kubernetes best practices and to match min. requirement when running an application.
Description
The helm chart must contain app specific sane default CPU and Memory settings.
CPU
Verify CPU is set to sane value and limit is 2-3 times higher (critical is that cpu request is not equal to limit). To have the CPU Limit higher than the request allows the pod to use more CPU when available and utilizes a node resources better.
Memory
Verify that memory request and limit are set equal. Memory is not a 'stretchable' resource and kubernetes will delete a pod if it starts using more memory than expected. Setting the memory limit to a higher value allows it to have a hidden memory profile which might work as long as the node has enough memory and will destabilize when the node does not have enough memory anymore.