Scheduling and Tooling

• K8s Cluster can contain multiple nodes and workloads are scheduled on the node by kube-schedule and there are different factors influencing the decision to select the node
  • node selector
  • node affinity
  • taints and tolerations
• Container Resource Limits influence this decision as kube-Scheduler will sum all the requests of container in a Pod spec and schedules it on a suitable node
• If the Pod doesnot schedule due to insufficient resources available on nodes in the Events we will see the following reasons
  • PodExceedsFreeCpu
  • PodExceedsFreeMemory

Node-Selector

• Refer Here for official docs
• Lets label two nodes
  • node2:
    • labels
      • diskspeed=high
  • node3:
    • labels:
      • diskspeed=low
        
• Lets create a Pod and specify the selection based on diskspeed=high which has a match and one more pod which doesnot match. Refer Here for the changeset
• Pod will be scheduled on node matching labels but not on not matching scenarios

Node Affinity

• Refer Here for official docs
• Refer Here for nodeAffinity with hard and soft rules
• Non Matching hard will not be scheduled
  
• podAffinity speaks about running a pod on the same node as some other pod with matching conditions and reverse of it is podAntiAffinity
• Refer Here for the manifests

Taints and Tolerations

• Taint is a property of a node to repel the pods. Only pods who have toleration for the taint will be scheduled on the node.
• Usecases:
  • Dedicated Hardware:
  • Evictions
• Refer Here for official docs
• Taint Effects:
  • NoSchedule
  • NoExecute
  • PreferNoShedule
• Taint nodes kubectl taint nodes <node-name> <taint:key>=<taint:value>:<taint-effect>

Yaml Processor

• The tool yq is used to read, modify and enhance contents of YAML file.
• Refer Here for the docs
  
• Exercise 1: Try more samples on modifying yaml files
• Exercise 2:
• Create a Deployment using nopCommerce Docker image
• Create a stateful set using mysql Container