Kubernetes Metrics
Resource metrics track the utilization and availability of critical resources such as CPU, memory, and storage. There are a number of Kubernetes metrics tools that can display the Kubernetes metrics. Kubernetes provides a Metrics API and many command-line queries to retrieve snapshots of resource utilization.
Kubernetes Metrics |
Value Range |
Use Case |
CPU per Node/Pod |
0 to max node CPU |
Pod CPU usage should be in line with DBMS CPU. Any abnormal pattern must be investigated. CPU usage can vary with the load, anything tending towards 80% and over must be investigated. |
Memory per Node/Pod |
0 to max node memory |
Pod memory usage should be in line with application memory usage. |
Disk throughput |
0 to max node specification |
Restricted disk io can impact performance. This may be evident during periods of intense database activity. |
Network Usage per Node/Pod |
0 to max node specification |
Restricted disk io can impact performance. This may be evident during periods of intense database activity. |
Requests vs. Limits |
Request to limit the value of CPU and memory |
NuoDB recommends requests must be equal to limits. |
| NuoDB Storage Managers (SMs) and Transaction Engines (TEs) use the request value to set the mem option. Currently, there is no need to have a higher limit value. Kubernetes will kill a pod that exceeds its memory limit, and throttle a pod that exceeds its CPU limit. If a pod operates between its memory request and limit, then it is a candidate to be removed if another pod needs to be scheduled with a request that would fit on the node. |