安装krew
krew是K8S的插件管理器，提供了很多有用的插件，可以大大提高工作效率。安装krew请参考官方文档：
https://krew.sigs.k8s.io/docs/user-guide/setup/install/
别名配置
根据当前系统设置的默认shell工具，修改~/.bashrc文件或者~/.zprofile文件，添加别名配置。
alias k=kubectl
常用插件
ns
ns是一个可以在kubernetes集群里面切换命名空间的插件。你是否厌烦了每次敲kubectl命令时都要加上-n <namespaces>的参数，又不想在config里面维护大量context。那么ns插件就是来解救你的，它会在切换命名空间时修改当前config里面的context到当前命名空间，这样你就不用再加-n参数了。
ns插件仓库与ctx插件源码仓库是同一个。
项目地址：https://github.com/ahmetb/kubectx
安装命令：
k krew install ns
使用示例：
$ k ns volcano-system
Context "kind-john" modified.
Active namespace is "volcano-system".
$ k get pod
NAME READY STATUS RESTARTS AGE
volcano-admission-5dd7897d96-rwfxn 1/1 Running 0 9d
volcano-controllers-7c58d44b88-bbmwm 1/1 Running 0 9d
volcano-scheduler-65c58864d-5wtd8 1/1 Running 0 8d
ctx
ctx是一个可以在kubernetes集群里面切换context的插件。
ns插件仓库与ctx插件源码仓库是同一个。
项目地址：https://github.com/ahmetb/kubectx
安装命令：
k krew install ctx
使用示例：
# switch to another cluster that's in kubeconfig
$ k ctx minikube
Switched to context "minikube".
# switch back to previous cluster
$ k ctx -
Switched to context "oregon".
# rename context
$ k ctx dublin=gke_ahmetb_europe-west1-b_dublin
Context "gke_ahmetb_europe-west1-b_dublin" renamed to "dublin".
# change the active namespace on kubectl
$ k ns kube-system
Context "test" set.
Active namespace is "kube-system".
# go back to the previous namespace
$ k ns -
Context "test" set.
Active namespace is "default".
# change the active namespace even if it doesn't exist
$ k ns not-found-namespace --force
Context "test" set.
Active namespace is "not-found-namespace".
---
$ k ns not-found-namespace -f
Context "test" set.
Active namespace is "not-found-namespace".
stern
stern是一个可以同时查看多个pod日志的工具。
另外还有一个与
stern类似的tail插件，用于获取最新的流式的日志输出。
项目地址：https://github.com/stern/stern
安装命令：
k krew install stern
使用示例：
$ k ns volcano-system
Context "kind-john" modified.
Active namespace is "volcano-system".
$ k stern .
volcano-admission-5dd7897d96-rwfxn admission I0821 06:52:39.431924 1 reflector.go:879] volcano.sh/apis/pkg/client/informers/externalversions/factory.go:145: Watch close - *v1beta1.Queue total 11 items received
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.691380 1 cache.go:1455] "SnapShot for scheduling" jobNum=1 QueueNum=3 NodeNum=3
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.691456 1 session.go:230] Open Session d1812087-af68-469e-9fbf-d8b53fb7cef6 with <1> Job and <3> Queues
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.692081 1 enqueue.go:79] Try to enqueue PodGroup to 0 Queues
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.692122 1 allocate.go:81] Try to allocate resource to 1 Queues
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.692138 1 proportion.go:307] Queue <default>: deserved <cpu 0.00, memory 0.00, pods 1.00, ephemeral-storage 0.00, hugepages-1Gi 0.00, hugepages-2Mi 0.00, hugepages-32Mi 0.00, hugepages-64Ki 0.00>, allocated <cpu 0.00, memory 0.00, pods 1.00>, share <1>
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.692166 1 allocate.go:142] Queue <default> is overused, ignore it.
volcano-scheduler-65c58864d-5wtd8 volcano-scheduler I0823 06:27:24.692598 1 session.go:364] Close Session d1812087-af68-469e-9fbf-d8b53fb7cef6
...
view-allocations
kubectl-view-allocations可以非常方便的展示CPU、内存、GPU等资源的分布情况，并可以对namespace、node、pod等维度进行展示。
项目地址：https://github.com/davidB/kubectl-view-allocations
安装命令：
k krew install view-allocations
使用示例：
$ k view-allocations
Resource Requested Limit Allocatable Free
cpu (4%) 1.1 (1%) 300.0m 30.0 28.9
├─ john-control-plane (10%) 950.0m (1%) 100.0m 10.0 9.1
│ ├─ coredns-668d6bf9bc-9r86n 100.0m __ __ __
│ ├─ coredns-668d6bf9bc-fpk6t 100.0m __ __ __
│ ├─ etcd-john-control-plane 100.0m __ __ __
│ ├─ kindnet-xqhbt 100.0m 100.0m __ __
│ ├─ kube-apiserver-john-control-plane 250.0m __ __ __
│ ├─ kube-controller-manager-john-control-plane 200.0m __ __ __
│ └─ kube-scheduler-john-control-plane 100.0m __ __ __
├─ john-worker (1%) 100.0m (1%) 100.0m 10.0 9.9
│ └─ kindnet-978h2 100.0m 100.0m __ __
└─ john-worker2 (1%) 100.0m (1%) 100.0m 10.0 9.9
└─ kindnet-2knqv 100.0m 100.0m __ __
ephemeral-storage __ __ 2.9Ti __
├─ john-control-plane __ __ 1006.9Gi __
├─ john-worker __ __ 1006.9Gi __
└─ john-worker2 __ __ 1006.9Gi __
memory (2%) 390.0Mi (2%) 490.0Mi 23.0Gi 22.5Gi
├─ john-control-plane (4%) 290.0Mi (5%) 390.0Mi 7.7Gi 7.3Gi
│ ├─ coredns-668d6bf9bc-9r86n 70.0Mi 170.0Mi __ __
│ ├─ coredns-668d6bf9bc-fpk6t 70.0Mi 170.0Mi __ __
│ ├─ etcd-john-control-plane 100.0Mi __ __ __
│ └─ kindnet-xqhbt 50.0Mi 50.0Mi __ __
├─ john-worker (1%) 50.0Mi (1%) 50.0Mi 7.7Gi 7.6Gi
│ └─ kindnet-978h2 50.0Mi 50.0Mi __ __
└─ john-worker2 (1%) 50.0Mi (1%) 50.0Mi 7.7Gi 7.6Gi
└─ kindnet-2knqv 50.0Mi 50.0Mi __ __
pods (5%) 17.0 (5%) 17.0 330.0 313.0
├─ john-control-plane (8%) 9.0 (8%) 9.0 110.0 101.0
├─ john-worker (5%) 5.0 (5%) 5.0 110.0 105.0
└─ john-worker2 (3%) 3.0 (3%) 3.0 110.0 107.0
该命令在智算资源管控场景中，常用来查看智算资源的使用情况，常与-r参数结合使用，例如：
$ k view-allocations -r nvidia.com/gpu
Resource Requested Limit Allocatable Free
nvidia.com/gpu (28%) 13.0 (28%) 13.0 47.0 34.0
├─ ai-app-2-151-msxf __ __ 8.0 __
├─ ai-app-64-40-msxf (12%) 1.0 (12%) 1.0 8.0 7.0
│ └─ neal-httpsingleport-0411-7fc454c8fb-bdpfk 1.0 1.0 __ __
├─ ai-app-8-1-msxf (25%) 2.0 (25%) 2.0 8.0 6.0
│ ├─ pd-qwen3-8b-decoder-7c6ddcd474-gnj5b 1.0 1.0 __ __
│ └─ pd-qwen3-8b-prefill-559b45d46b-g4zhv 1.0 1.0 __ __
├─ msxf-app-2-149 (62%) 5.0 (62%) 5.0 8.0 3.0
│ ├─ ces-6c4f56d6c6-7fcbb 1.0 1.0 __ __
│ ├─ itsmceshi0806-865568b845-jcn66 1.0 1.0 __ __
│ ├─ itsmceshi0806-865568b845-zdvs6 1.0 1.0 __ __
│ ├─ msv91211-dep-7c5ccb6854-gtkcd 1.0 1.0 __ __
│ └─ neal-0819-6bb79d4569-bhfp7 1.0 1.0 __ __
├─ msxf-hpc-37-2-ai (62%) 5.0 (62%) 5.0 8.0 3.0
│ ├─ pd-qwen3-30b-decoder-d46b576dc-gglfh 1.0 1.0 __ __
│ ├─ pd-qwen3-30b-prefill-746d9fc77f-kb5gw 1.0 1.0 __ __
│ ├─ pd-qwen3-32b-decoder-75c6d96c4-qhkl9 1.0 1.0 __ __
│ ├─ pd-qwen3-32b-prefill-75d6cd57fd-45vzd 1.0 1.0 __ __
│ └─ qwen25-72b-int4-bz-6d6546b9f7-p2wm9 1.0 1.0 __ __
└─ msxf-hpc-64-5-ai __ __ 7.0 __
tree
该插件是由Google大佬开发，通过ownerReferences来发现kubernetes对象之间的相互关联，并通过树状图来展示，对资源的关系一目了然。
项目地址：https://github.com/ahmetb/kubectl-tree
安装命令：
k krew install tree
使用示例：
$ k ns volcano-system
Context "kind-john" modified.
Active namespace is "volcano-system".
$ k get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
volcano-admission 1/1 1 1 9d
volcano-controllers 1/1 1 1 9d
volcano-scheduler 1/1 1 1 9d
$ k tree deploy volcano-scheduler
NAMESPACE NAME READY REASON AGE
volcano-system Deployment/volcano-scheduler - 9d
volcano-system ├─ReplicaSet/volcano-scheduler-65c58864d - 8d
volcano-system │ └─Pod/volcano-scheduler-65c58864d-5wtd8 True 8d
volcano-system ├─ReplicaSet/volcano-scheduler-77cdd888b9 - 8d
volcano-system └─ReplicaSet/volcano-scheduler-986f77795 - 9d
node-shell
node-shell插件可以让你直接在node上执行shell命令，而不需要登录到node上。
项目地址：https://github.com/kvaps/kubectl-node-shell
实现原理：该命令会创建一个nsenter-xxx的Pod，该Pod会带有一个privileged: true的高权限容器，默认使用alpine:latest镜像。
注意事项：由于该命令默认会拉取外网镜像，在某些外网不通的网络环境下会无法执行成功，可以使用--image参数指定需要使用的镜像。
安装命令：
k krew install node-shell
使用示例：
# Get standard bash shell
kubectl node-shell <node>
# Use custom image for pod
kubectl node-shell <node> --image <image>
# Use X-mode (mount /host, and do not enter host namespace)
kubectl node-shell -x <node>
# Skip specific namespace types to enter, choose any of ipc, mount, pid, net, uts
kubectl node-shell <node> --no-ipc
# Execute custom command
kubectl node-shell <node> -- echo 123
# Use stdin
cat /etc/passwd | kubectl node-shell <node> -- sh -c 'cat > /tmp/passwd'
# Run oneliner script
kubectl node-shell <node> -- sh -c 'cat /tmp/passwd; rm -f /tmp/passwd'
$ k get node
NAME STATUS ROLES AGE VERSION
john-control-plane Ready control-plane 9d v1.32.2
john-worker Ready <none> 9d v1.32.2
john-worker2 Ready <none> 9d v1.32.2
$ k node-shell john-worker
spawning "nsenter-8xhpq5" on "john-worker"
If you don't see a command prompt, try pressing enter.
root@john-worker:/#
其实kubectl工具也提供了debug命令，用于实现类似的功能。该debug命令功能更强大一些，除了支持node，也支持pod的debug，具体可以参考：https://kubernetes.io/docs/tasks/debug/debug-cluster/kubectl-node-debug/
使用示例：
$ k debug node/john-worker -it --image=ubuntu --profile=general
Creating debugging pod node-debugger-john-worker-hcr2q with container debugger on node john-worker.
If you don't see a command prompt, try pressing enter.
root@john-worker:/#
images
kubectl-images可以展示集群中正在使用的镜像，并对namespace进行一个简单的统计。使用这个插件可以非常方便的查看namespace中使用了哪些镜像，尤其在排查问题需要查看镜像版本时非常有用。
安装命令：
k krew install images
使用示例：
$ k images -A
[Summary]: 4 namespaces, 17 pods, 17 containers and 12 different images
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| Pod | Container | Image |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| example-deployment-9488676d4-ttb6h | nginx | nginx:1.14.2 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| coredns-668d6bf9bc-9r86n | coredns | registry.k8s.io/coredns/coredns:v1.11.3 |
+--------------------------------------------+ + +
| coredns-668d6bf9bc-fpk6t | | |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| etcd-john-control-plane | etcd | registry.k8s.io/etcd:3.5.16-0 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| kindnet-2knqv | kindnet-cni | docker.io/kindest/kindnetd:v20250214-acbabc1a |
+--------------------------------------------+ + +
| kindnet-978h2 | | |
+--------------------------------------------+ + +
| kindnet-xqhbt | | |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| kube-apiserver-john-control-plane | kube-apiserver | registry.k8s.io/kube-apiserver:v1.32.2 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| kube-controller-manager-john-control-plane | kube-controller-manager | registry.k8s.io/kube-controller-manager:v1.32.2 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| kube-proxy-bmff8 | kube-proxy | registry.k8s.io/kube-proxy:v1.32.2 |
+--------------------------------------------+ + +
| kube-proxy-f2fkv | | |
+--------------------------------------------+ + +
| kube-proxy-nxd7w | | |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| kube-scheduler-john-control-plane | kube-scheduler | registry.k8s.io/kube-scheduler:v1.32.2 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| local-path-provisioner-7dc846544d-74xnp | local-path-provisioner | docker.io/kindest/local-path-provisioner:v20250214-acbabc1a |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| volcano-admission-5dd7897d96-rwfxn | admission | docker.io/volcanosh/vc-webhook-manager:v1.12.1 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| volcano-controllers-7c58d44b88-bbmwm | volcano-controllers | docker.io/volcanosh/vc-controller-manager:v1.12.1 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+
| volcano-scheduler-65c58864d-5wtd8 | volcano-scheduler | volcanosh/vc-scheduler:80eea1df4b922773c47ac4f8e483b48a3ccc7090 |
+--------------------------------------------+-------------------------+-----------------------------------------------------------------+