We created a plugin in Grafana for monitoring k8s: already in the official Grafana-store

Today a new and very useful open source project was created, created by the development department of ITSumma, a plug-in in Grafana for monitoring Kubernetes . He is included in the official grafana- stor - grafana.com/grafana/plugins/devopsprodigy-kubegraf-app

Its key utility:

1. integration with k8s-api to build a complete map of your applications + grouping by namespace + binding to pods / services.
2. A graphical representation of the distribution of applications among the nodes of the k8s cluster in real time.
3. real-time statistics on the status of applications / pods in the cluster and error messages (for example, if your application has stopped passing liveness-probes).
4. dashboards with statistics on various parameters of a cluster node (CPU usage, memory, load on the disk subsystem and network interfaces), as well as statistics on resource usage for a specific hearth (CPU, memory, network interfaces) and deployment / statefulset / statuses daemonsets in which you can see the number of available replicas of the application, the number of running containers by this application and the number of container restarts.
5. Separate visualization for viewing real-time statistics on the state of nodes in a cluster.

Where does the plugin collect information from?

The plug-in configuration provides for the collection of data, firstly, from the k8s API server (for building a map of your applications, collecting application states, information about passed tests, etc.) An additional Prometeus datasource is also separately set in the plugin (collection of metrics from kube-state-metrics and node-exporter).

Why and why did we do this?

It's simple: there are exactly one similar plug-in, and it has not been supported for about a year now.



What besides this did not suit us in it:

1. lack of competent visualization of the application map.
2. incompatibility with current versions of node-exporter'a and kube-state-metrics'a.
3. lack of support for statefulset monitoring.

Well, and we just can afford it! ;-)



Stars, ishshui and pull-quests are welcome - github.com/devopsprodigy/kubegraf



And, of course, many thanks to Alexander Zobnin for air support!



Very soon I will tell the whole story of the creation: "how it was." Do not switch (s)