In the previous document, we took advantage of the elasticity Kubernetes provides to dynamically create Jenkins Slaves on Kubernetes. This document is mainly a stress test of Jenkins under large-scale builds. 1. Cluster Configuration 1.1 Kubernetes Version The version used here is v1.16.7 1 2 3 4 kubectl version Client Version: version.

Building on the previous two documents, Creating Jenkins Slaves Dynamically on Kubernetes and Adding Windows Nodes to Kubernetes to Provide Dynamic Jenkins Build Agents, this document mainly attempts to dynamically provide Windows build agents on Kubernetes. 1. Adding Pipelines For the Kubernetes and Jenkins integration part, refer to the two documents above; here we simply create two pipelines to test with.

Here a Windows node is used mainly as a Worker, while the Master control plane stays on Linux. 1. System Configuration 1.1 Kubernetes Control Plane Since version 1.14, Kubernetes has added production-grade support for Windows nodes. Because Microsoft’s official documentation mainly covers installing the flannel network plugin, it is recommended here that Kubernetes also use the flannel plugin.

1. Delete Pods in the Evicted State in Bulk 1 kubectl get pods --all-namespaces -o wide | grep Evicted | awk '{print $1,$2}' | xargs -L1 kubectl delete pod -n 2. Delete Pods in a Given State in a Given Namespace in Bulk Using field-selectors, you can delete Pods in a given state in a given namespace.

1. The Scheduler in Kubernetes kube-scheduler is the component in Kubernetes that decides which Node a Pending Pod runs on; it is called the scheduler. Kubernetes ships with a large number of built-in scheduling policies and also provides some advanced scheduling policies (nodeAffinity, podAffinity, and so on) for users, which