2.11 Multiple Schedulers

• Kubernetes uses a default scheduler to assign Pods to nodes based on resource availability, taints/tolerations, affinity rules, and placement constraints.

• For advanced scenarios, Kubernetes is extensible and allows you to run custom schedulers with your own placement logic.

---

When to Use a Custom Scheduler

Use a custom scheduler only when default scheduling behavior is not sufficient.

You may need one when:

• Custom scheduling algorithms are required
• Additional placement validation is needed
• Special node selection rules must be enforced
• Certain workloads require dedicated scheduling behavior

Tip

Most workloads should continue using the default scheduler.

---

Multiple Scheduler Support

Kubernetes supports multiple schedulers running simultaneously.

• Default scheduler handles regular Pods
• Custom schedulers handle only Pods that reference them
• Custom schedulers are typically deployed as Pods
• They can run alongside the default scheduler

---

Scheduler Naming

Each scheduler must have a unique name.

Item	Value
Default scheduler name	default-scheduler
Defined in	Scheduler configuration file
Pod field	schedulerName
If not specified	Default scheduler is used

---

Scheduler Configuration

Default Scheduler

apiVersion: kubescheduler.config.k8s.io/v1
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: default-scheduler

Custom Scheduler

apiVersion: kubescheduler.config.k8s.io/v1
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: my-scheduler

---

Pod Using a Custom Scheduler

A Pod selects a scheduler using the schedulerName field.

apiVersion: v1
kind: Pod
metadata:
  name: custom-scheduled-pod
spec:
  schedulerName: my-scheduler
  containers:
    - name: nginx
      image: nginx

Only Pods that specify this field will be handled by the custom scheduler.

---

Deploying a Custom Scheduler (kubeadm)

In kubeadm clusters, control plane components run as Pods.
You can deploy an additional scheduler as a Pod running the kube-scheduler binary.

Required Arguments

Flag	Purpose
--kubeconfig	API server authentication
--config	Scheduler configuration file

---

Custom Scheduler Pod Example

apiVersion: v1
kind: Pod
metadata:
  name: my-custom-scheduler
  namespace: kube-system
spec:
  containers:
    - name: kube-scheduler
      image: k8s.gcr.io/kube-scheduler-amd64:v1.xx.x
      command:
        - kube-scheduler
        - --address=127.0.0.1
        - --kubeconfig=/etc/kubernetes/scheduler.conf
        - --config=/etc/kubernetes/my-scheduler-config.yaml

---

Leader Election (HA Clusters)

Leader election is required when running scheduler replicas across multiple control-plane nodes.

Warning

Without leader election, multiple schedulers may try to schedule the same Pod.

apiVersion: kubescheduler.config.k8s.io/v1
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: my-scheduler

leaderElection:
  leaderElect: true
  resourceNamespace: kube-system
  resourceName: lock-object-my-scheduler

---

Troubleshooting Custom Scheduler Issues

If a custom scheduler is not running or misconfigured, Pods using it will remain in Pending state.

---

Check Pod Scheduling Details

kubectl describe pod <pod-name>

Check the Events section for scheduling errors.

---

Identify Which Scheduler Scheduled a Pod

kubectl get events -o wide

Look for:

• Reason = Scheduled
• Source column = scheduler name

---

Check Scheduler Logs

Scheduler PodScheduler Deployment

kubectl logs <scheduler-pod-name> -n kube-system

kubectl logs deploy/<scheduler-deployment-name> -n kube-system

---

Command Reference

kubectl describe pod <pod-name>
kubectl get events -o wide
kubectl logs <scheduler-pod-name> -n kube-system

---

Summary

• Default scheduler handles most workloads
• Custom schedulers support specialized placement logic
• Multiple schedulers can run together
• Each scheduler requires a unique name
• Pods select schedulers using schedulerName
• Enable leader election in HA setups
• Use describe, events, and logs for troubleshooting