The kubernetes and kube_inventory input plugins have most of the metrics to monitor k8s infrastructure needs but few resources and metrics are still missing from the set of plugins which can be easily extended and will help in better k8s monitoring.
Proposal
The kube_inventory plugin can be extended to not only have capacity and allocatable quantity metrics but also other health metrics like node status, node count and if node is schedulable or not. Also new resource measurement like resourcequoata can be added for better monitoring. These are the metrics that can be easily extended using "k8s.io/api/core/v1" library.
kubernetes_node_condition_status
kubernetes_node_count
kubernetes_unschedulable
Ex.
for _, val := range n.Status.Conditions {
.
.
fields["status_condition"] = string(val.Status)
}
fields["spec_unschedulable"] = n.Spec.Unschedulable
Also new measurement type can be included with following metrics.
kubernetes_resourcequota
tags:
resource
namespace
fields:
hard_cpu_cores_limit
hard_memory_bytes_limit
hard_pods_limit
used_cpu_cores
used_memory_bytes
used_pods
Current behavior
Currently the metrics mentioned above have not been included in any input plugin.
Desired behavior
After the implementation of the feature, the kube_inventory plugin output should be something like this.
We are planning to migrate our monitoring infrastructure from prometheus to telegraf and trying to fill up those gaps in the metrics desired. Combining this feature with already raised https://github.com/influxdata/telegraf/issues/8546 will serve our purpose.
Feature Request
The kubernetes and kube_inventory input plugins have most of the metrics to monitor k8s infrastructure needs but few resources and metrics are still missing from the set of plugins which can be easily extended and will help in better k8s monitoring.
Proposal
The kube_inventory plugin can be extended to not only have capacity and allocatable quantity metrics but also other health metrics like node status, node count and if node is schedulable or not. Also new resource measurement like resourcequoata can be added for better monitoring. These are the metrics that can be easily extended using "k8s.io/api/core/v1" library.
Also new measurement type can be included with following metrics.
kubernetes_resourcequota
Current behavior
Currently the metrics mentioned above have not been included in any input plugin.
Desired behavior
After the implementation of the feature, the kube_inventory plugin output should be something like this.
_> kubernetesnode,host=vjain count=8i 1628918652000000000 > kubernetes_node,condition=Ready,host=vjain,node_name=ip-172-17-0-2.internal,status=True statuscondition=1i 1629177980000000000 > kubernetes_node,cluster_namespace=tools,condition=Ready,host=vjain,node_name=ip-172-17-0-2.internal,status=True allocatable_cpu_cores=4i,allocatable_memory_bytes=7186567168i,allocatable_millicpu_cores=4000i,allocatable_pods=110i,capacity_cpu_cores=4i,capacity_memory_bytes=7291424768i,capacity_millicpu_cores=4000i,capacity_pods=110i,spec_unschedulable=0i,status_condition=1i 1628918652000000000
Use case
We are planning to migrate our monitoring infrastructure from prometheus to telegraf and trying to fill up those gaps in the metrics desired. Combining this feature with already raised https://github.com/influxdata/telegraf/issues/8546 will serve our purpose.