The metrics-agent pod requests 2 GB in memory and 0.5 CPUs, despite needing much less. As a result, new nodes get scheduled for no reason, thereby increasing financial costs and carbon emissions.
Based on Prometheus metrics collected from over 100 clusters with several thousand CPU cores in total, a CPU request of 10m and a memory request of 50 MB would be more than enough.
The metrics-agent pod requests 2 GB in memory and 0.5 CPUs, despite needing much less. As a result, new nodes get scheduled for no reason, thereby increasing financial costs and carbon emissions.
Based on Prometheus metrics collected from over 100 clusters with several thousand CPU cores in total, a CPU request of 10m and a memory request of 50 MB would be more than enough.