k8s批量删除Evicted Pods
服务端
2020年5月8日星期五 12:09
# 背景
查询pod状态
kubectl -n kube-system get pods | grep coredns | grep Evicted
1
查询结果为：
coredns-67cb7b5f46-2crxg 0/1 Evicted 0 4h56m
coredns-67cb7b5f46-2ds7n 0/1 Evicted 0 4h56m
coredns-67cb7b5f46-2pzw5 0/1 Evicted 0 4h56m
coredns-67cb7b5f46-4f8nm 0/1 Evicted 0 166m
coredns-67cb7b5f46-4gncz 0/1 Evicted 0 10h
coredns-67cb7b5f46-4lnw2 0/1 Evicted 0 166m
coredns-67cb7b5f46-4smp6 0/1 Evicted 0 166m
coredns-67cb7b5f46-5hx49 0/1 Evicted 0 166m
coredns-67cb7b5f46-6jxqn 0/1 Evicted 0 166m
coredns-67cb7b5f46-6qh9z 0/1 Evicted 0 166m
coredns-67cb7b5f46-6rn4c 0/1 Evicted 0 10h
coredns-67cb7b5f46-7jsth 0/1 Evicted 0 10h
coredns-67cb7b5f46-7mzvb 0/1 Evicted 0 166m
coredns-67cb7b5f46-92t2k 0/1 Evicted 0 4h56m
coredns-67cb7b5f46-9jqdm 0/1 Evicted 0 10h
coredns-67cb7b5f46-9rxsc 0/1 Evicted 0 4h56m
coredns-67cb7b5f46-9xmkk 0/1 Evicted 0 166m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
发现有很多 Evicted 状态的pod
问题原因：这是由于node节点的磁盘或内存资源不够后将现有的pod进行了驱赶。
# 解决
批量删除所有 Evicted 状态的pod即可
kubectl -n kube-system get pods | grep coredns | grep Evicted | awk '{print$1}'| xargs kubectl -n kube-system delete pods
1
可自定义替换 namespace、podName
kubectl -n ${namespace} get pods | grep ${podName} | grep Evicted | awk '{print$1}'| xargs kubectl -n ${namespace} delete pods
1
#指定pod所在的命名空间
kubectl -n kube-system get pods | grep coredns
1
2
2
执行完成后，成功删除全部 Evicted 状态的Pods