Cgroups V2 plugins

Allows to measure CPU-time used using the cgroup v2 sysfs.

Table of Contents

• Kubernetes Plugin
  • How to use
  • Prepare your environment
    • cgroup v2
    • kubectl
    • alumet-reader
  • How it works
• OAR3 Plugin
  • How to use
  • Prepare your environment
    • cgroup v2
  • How it works

Kubernetes Plugin

How to use

Just compile the app-agent of the alumet's github repository.

cargo run

Make sure that in the app-agent's cargo.toml file the Kubernetes plugin is imported. Make sure that in main.rs of app-agent the Kubernetes plugin is imported and used.

The binary created by the compilation will be found under the target repository.

Prepare your environment

To work this plugin needs several things.

1. cgroup v2
2. kubectl
3. alumet-reader user

cgroup v2

As the plugin use cgroup v2 to gather data, make sure to use this version of cgroup. In fact, a check is made by the plugin. If it doesn't detect the use of cgroup v2, the plugin will not start.

kubectl

In this version of the plugin, to gather some data about pods and nodes, the plugin use kubectl. So make sure that kubectl is installed and usable. How to install kubectl

Thanks to kubectl, you can get the kubernetes API URL. Use:

kubectl config view -o jsonpath='{"Cluster name\tServer\n"}{range .clusters[*]}{.name}{"\t"}{.cluster.server}{"\n"}{end}'

And note the URL corresponding to the kubernetes you want interact with.

Example:

Cluster name    Server
kubernetes      https://1.2.3.4:6443

I get the kubernetes part and I wrote the result in the alumet-config.toml file. Under the section: [plugins.k8s] I add the following:

kubernetes_api_url = "https://1.2.3.4:6443"

By default the value kubernetes_api_url will be set at: https://127.0.0.1:8080

alumet-reader

To use the Kubernetes API an user is needed. It's the alumet-reader user. Make sur the user exist and have the good rights. You can use the yaml file: alumet-user.yaml to create this user. Run:

kubectl apply -f alumet-user.yaml

How it works

The plugin read data inside the sysfs. At its start, the plugin retrieves a token thanks to the alumet-reader user. It uses the token to retrieve all pods running on the current node. After that, it retrieves all cgroup related to Kubernetes pods, and make a correspondence between data retrieve from API and gathered from cgroup to fulfil the input plugin. When a new pod is started. The plugin interrogate the API to retrieve data about the pod, such as its name, its namespace and the node its running on.

OAR3 Plugin

How to use

Just compile the app-agent of the alumet's github repository.

cargo run

Make sure that in the app-agent's cargo.toml file the OAR3 plugin is imported. Make sure that in main.rs of app-agent the OAR3 plugin is imported and used.

The binary created by the compilation will be found under the target repository.

Prepare your environment

To work this plugin needs several things.

1. cgroup v2

cgroup v2

As the plugin uses cgroup v2 to gather data, make sure to use this version of cgroup. In fact, a check is made by the plugin. If it doesn't detect the use of cgroup v2, the plugin will not start.

How it works

The plugin read data inside the sysfs. It read file under the path given in the config. If it find a cpu consumption time it gather it. If some folder are created, it try to read the new folder and find information about time consumption.