• The Default Scheduler has an algorithm that shares pods evenly across nodes.
    • Takes into consideration Taints and Tolerations, as well as Node Affinity.
  • Can write your own Scheduler and can assign that to the cluster as well as the Default Scheduler (both are fine to have running at the same time). You can specify certain applications to you a particular scheduler.
  • The default scheduler is named default-scheduler.
  • The name is created using the following file: ``` scheduler-config.yaml apiVersion: kubescheduler.config.k8s.io/v1 kind: KubeSchedulerConfiguration profiles:
  • schedulerName: default-scheduler ```
  • For the other scheduler files, you set the name like so: ``` scheduler-config.yaml apiVersion: kubescheduler.config.k8s.io/v1 kind: KubeSchedulerConfiguration profiles:
  • schedulerName: my-scheduler ```
  • To deploy a scheduler, an example one is the kube-scheduler and then run it is a service: 
    wget https://storage.googleapis.com/kubernetes-release/release/v1.12.0/bin/linux/amd64/kube-scheduler
  • Service file: 
    kube-scheduler.service
ExecStart=/usr/local/bin/kube-scheduler \\
 --config=/etc/kubernetes/config/kube-scheduler.yaml
  • Each scheduler has its own configuration file and each file has its own name: 
    my-scheduler-2.service
ExecStart=/usr/local/bin/kube-scheduler \\
 --config=/etc/kubernetes/config/my-scheduler-2.yaml
  • There are other options that can be passed in, such as the kube config option.
  • The above is not necessarily how you deploy a scheduler, mostly the work is done by kubeadm.
  • Can deploy a scheduler as a pod:
  • Example custom scheduler yaml file: ``` my-custom-scheduler.yaml apiVerson: v1 kind: Pod metadata: name: my-custom-schedule namespace: kube-system spec: containers:
    • command:
      • kube-scheduler
      • –address=127.0.0.1
      • –kubeconfig=/etc/kubernetes/scheduler.conf
      • –config=/etc/kubernetes/my-scheduler-config.yaml image: k8s.gcr.io/kube-scheduler-amd64:v1.11.3 name: kube-scheduler ```
  • Similarly the my-scheduler-config.yaml file: ``` apiVersion: kubescheduler.config.k8s.io/v1 kind: KubeSchedulerConfiguration profiles:
  • schedulerName: my-scheduler leaderElection: leaderElect: true resourceNamespace: kube-system resourceName: lock-object-my-scheduler ```
  • leaderElect option is used when you have different copies of the Scheduler running on different Master Nodes (a form of High Availability).
    • If multiple copies of the same scheduler are running on different Master Nodes, only one can be active at a time.
    • The leaderElect option thus helps to schedule a leader in these scenarios.
  • How to deploy an additional scheduler as a deployment?
  • Useful Kubernetes Documentation on Schedulers
  • View schedulers in the kube-system namespace with kubectl get pods --namespace=kube-system
  • How do you configure a pod or deployment to use a new scheduler?
  • Example pod definition file (the important part is the schedulerName part): ``` apiVersion: v1 kind: Pod metadata: name: nginx spec: containers:
    • image: nginx name: nginx schedulerName: my-custom-scheduler ```
  • Then kubectl create -f pod-definition.yaml
  • If the Pod is in a Pending state, check the logs under the kubectl describe command.
  • How do you which Scheduler picked up pod X?
  • Use the kubectl get events -o wide command.
    • Lists all events in the current namespace and it shows the source of the event.
    • Can also view the logs of the scheduler like so: kubectl logs my-custom-scheduler --name-space=kube-system
      • Can add the pod name or the deployment name.

Updated: