Node Taint, Toleration and Affinity
Motivation Pod scheduling can be a nightmare in a large Kubernetes deployment with many nodes having different configurations. Consider the configuration below: NODE A: 16vCPU, 10TB SSD Disk Space, 64GB RAM. NODE B: 4vCPU, 100GB HDD Disk Space, 8GB RAM NODE C: 8vCPU, 100GB HDD Disk Space, 8GB RAM, with additional GPU configuration And scheduling a new deployment to the cluster: kubectl create deployment --image=nginx --replicas=3 Kubernetes would schedule the Pods on a random node....