oc run --restart=NeverでPod単体で動作させると、oc adm drainでノードをサービスアウトさせようとしても、Podが退避されないのでその様子の実行ログ。
この記事の補足的な内容。
環境
[zaki@okd4-manager ~]$ oc version Client Version: 4.4.0-0.okd-2020-01-28-022517 Server Version: 4.4.0-0.okd-2020-01-28-022517 Kubernetes Version: v1.17.1
ノード構成
[zaki@okd4-manager ~]$ oc get node -o wide NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME okd4-master0 Ready master 10d v1.17.1 172.16.0.10 <none> Fedora CoreOS 31.20200127.20.1 5.4.13-201.fc31.x86_64 cri-o://1.17.0-rc1 okd4-master1 Ready master 10d v1.17.1 172.16.0.11 <none> Fedora CoreOS 31.20200127.20.1 5.4.13-201.fc31.x86_64 cri-o://1.17.0-rc1 okd4-master2 Ready master 10d v1.17.1 172.16.0.12 <none> Fedora CoreOS 31.20200127.20.1 5.4.13-201.fc31.x86_64 cri-o://1.17.0-rc1 okd4-worker0 Ready worker 10d v1.17.1 172.16.0.20 <none> Fedora CoreOS 31.20200127.20.1 5.4.13-201.fc31.x86_64 cri-o://1.17.0-rc1 okd4-worker1 Ready worker 10d v1.17.1 172.16.0.21 <none> Fedora CoreOS 31.20200127.20.1 5.4.13-201.fc31.x86_64 cri-o://1.17.0-rc1
Pod状態
[zaki@okd4-manager ~]$ oc get pod -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES dc-run-pod-1-deploy 0/1 Completed 0 62m 10.129.0.20 okd4-worker0 <none> <none> dc-run-pod-1-ht77f 1/1 Running 0 62m 10.128.2.19 okd4-worker1 <none> <none> dc-run-with-svc-1-deploy 0/1 Completed 0 54m 10.129.0.22 okd4-worker0 <none> <none> dc-run-with-svc-1-wddbs 1/1 Running 0 53m 10.128.2.20 okd4-worker1 <none> <none> deploy-run-pod-56c6698fdb-dzk2g 1/1 Running 0 60m 10.129.0.21 okd4-worker0 <none> <none> deploy-run-with-svc-5ffbb8d9c7-h2j28 1/1 Running 0 51m 10.129.0.23 okd4-worker0 <none> <none> never-restart-pod 1/1 Running 0 67m 10.128.2.18 okd4-worker1 <none> <none>
never-restart-podが単体で動いてるPod(親リソース無し)
これがデプロイされてるノード(okd4-worker1)をdrainする。
drain
[zaki@okd4-manager ~]$ oc adm drain okd4-worker1 --ignore-daemonsets --delete-local-data node/okd4-worker1 cordoned evicting pod "alertmanager-main-1" evicting pod "downloads-cf5d8d7f4-b9bfg" evicting pod "csi-snapshot-controller-f44f9d4b4-fvwv2" evicting pod "alertmanager-main-0" evicting pod "prometheus-adapter-d48bc96cc-g8nwl" evicting pod "grafana-66d8c446bf-p75bt" evicting pod "kube-state-metrics-84bf857f94-fpcc8" evicting pod "telemeter-client-954595b4d-cfptr" evicting pod "prometheus-k8s-0" evicting pod "dc-run-pod-1-ht77f" evicting pod "dc-run-with-svc-1-wddbs" pod/csi-snapshot-controller-f44f9d4b4-fvwv2 evicted pod/prometheus-k8s-0 evicted pod/telemeter-client-954595b4d-cfptr evicted pod/alertmanager-main-0 evicted pod/alertmanager-main-1 evicted pod/dc-run-pod-1-ht77f evicted pod/downloads-cf5d8d7f4-b9bfg evicted pod/grafana-66d8c446bf-p75bt evicted pod/kube-state-metrics-84bf857f94-fpcc8 evicted pod/prometheus-adapter-d48bc96cc-g8nwl evicted pod/dc-run-with-svc-1-wddbs evicted node/okd4-worker1 evicted [zaki@okd4-manager ~]$
drainの結果
[zaki@okd4-manager ~]$ oc get pod -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES dc-run-pod-1-deploy 0/1 Completed 0 65m 10.129.0.20 okd4-worker0 <none> <none> dc-run-pod-1-zjslf 1/1 Running 0 99s 10.129.0.30 okd4-worker0 <none> <none> dc-run-with-svc-1-deploy 0/1 Completed 0 56m 10.129.0.22 okd4-worker0 <none> <none> dc-run-with-svc-1-tqf44 1/1 Running 0 99s 10.129.0.26 okd4-worker0 <none> <none> deploy-run-pod-56c6698fdb-dzk2g 1/1 Running 0 63m 10.129.0.21 okd4-worker0 <none> <none> deploy-run-with-svc-5ffbb8d9c7-h2j28 1/1 Running 0 54m 10.129.0.23 okd4-worker0 <none> <none> never-restart-pod 1/1 Running 0 70m 10.128.2.18 okd4-worker1 <none> <none>
この通り、drainしてもevictされない。
--forceを付けるとと強制的にevictされるが、停止するだけで別ノードで再デプロイはされない。
[zaki@okd4-manager ~]$ oc adm uncordon okd4-worker1 node/okd4-worker1 uncordoned [zaki@okd4-manager ~]$ oc adm drain okd4-worker1 --ignore-daemonsets --delete-local-data --force node/okd4-worker1 cordoned WARNING: deleting Pods not managed by ReplicationController, ReplicaSet, Job, DaemonSet or StatefulSet: run-sample/never-restart-pod evicting pod "alertmanager-main-0" evicting pod "router-default-5698995cc8-b7k5f" evicting pod "migrator-5746b7d596-vnrf8" evicting pod "grafana-66d8c446bf-8ws9l" evicting pod "prometheus-k8s-0" evicting pod "never-restart-pod" pod/alertmanager-main-0 evicted pod/grafana-66d8c446bf-8ws9l evicted pod/router-default-5698995cc8-b7k5f evicted pod/prometheus-k8s-0 evicted pod/migrator-5746b7d596-vnrf8 evicted pod/never-restart-pod evicted node/okd4-worker1 evicted [zaki@okd4-manager ~]$ oc get pod -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES dc-run-pod-1-deploy 0/1 Completed 0 70m 10.129.0.20 okd4-worker0 <none> <none> dc-run-pod-1-zjslf 1/1 Running 0 6m53s 10.129.0.30 okd4-worker0 <none> <none> dc-run-with-svc-1-deploy 0/1 Completed 0 61m 10.129.0.22 okd4-worker0 <none> <none> dc-run-with-svc-1-tqf44 1/1 Running 0 6m53s 10.129.0.26 okd4-worker0 <none> <none> deploy-run-pod-56c6698fdb-dzk2g 1/1 Running 0 68m 10.129.0.21 okd4-worker0 <none> <none> deploy-run-with-svc-5ffbb8d9c7-h2j28 1/1 Running 0 59m 10.129.0.23 okd4-worker0 <none> <none>
ノードにCompletedなPodが残っている場合、drainによって消える。
また、OCPのバージョンによってはエラーになる。
k8sでも同じだと思うけど、よく考えたら自由に使えるマルチノードk8sクラスタ持ってなかった… 作るか…
ふと思ったけど、restartPolicy: Never以外の設定でPod単体で動かせたりできるのかな…？
