不会吧?! 新版本 longhorn 部署需要 k8s.gcr.io 镜像?
发布于: 刚刚
前言
在之前，longhorn 的部署是不涉及http://k8s.gcr.io的镜像的,但是在现在新版本当中,csi 相关的镜像都是直接使用http://k8s.gcr.io中的镜像.
淡定
不要慌,这时候可以使用短域名镜像代理lank8s.cn来代替 k8s.gcr.io.只需要将 k8s.gcr.io 修改为lank8s.cn就可以了，其他都不变.
部署
按照官网文档一步一步做,走起.
1. helm repo add longhorn https://charts.longhorn.io2. helm repo update3. kubectl create namespace longhorn-system 4. helm install longhorn longhorn/longhorn --namespace longhorn-system复制代码
现在已经通过 helm chart 部署起来了，等待一会看看情况，不出意外的话会因为拉取镜像部署失败.
[root@lan1 ~]# kubectl get pod -n longhorn-systemNAME READY STATUS RESTARTS AGEcsi-attacher-75588bff58-qrbfg 0/1 ImagePullBackOff 0 4m46scsi-attacher-75588bff58-r77n8 0/1 ImagePullBackOff 0 4m46scsi-attacher-75588bff58-t944p 0/1 ImagePullBackOff 0 4m46scsi-provisioner-669c8cc698-2p22g 0/1 ImagePullBackOff 0 4m46scsi-provisioner-669c8cc698-d6md5 0/1 ErrImagePull 0 4m46scsi-provisioner-669c8cc698-ppktl 0/1 ImagePullBackOff 0 4m46scsi-resizer-5c88bfd4cf-d822s 0/1 ContainerCreating 0 4m46scsi-resizer-5c88bfd4cf-j9n27 0/1 ContainerCreating 0 4m46scsi-resizer-5c88bfd4cf-sl92r 0/1 ContainerCreating 0 4m46scsi-snapshotter-69f8bc8dcf-9rf7c 0/1 ContainerCreating 0 4m45scsi-snapshotter-69f8bc8dcf-pnrt8 0/1 ContainerCreating 0 4m45scsi-snapshotter-69f8bc8dcf-sssnb 0/1 ContainerCreating 0 4m45sengine-image-ei-d4c780c6-9525q 1/1 Running 0 4m55sinstance-manager-e-a39ce34c 1/1 Running 0 4m55sinstance-manager-r-71c5e3f3 1/1 Running 0 4m54slonghorn-csi-plugin-8s25k 0/2 ContainerCreating 0 4m45slonghorn-driver-deployer-75f68555c9-mhb8j 1/1 Running 0 5m24slonghorn-manager-6dbsn 1/1 Running 0 5m10slonghorn-ui-75ccbd4695-92cj2 1/1 Running 0 5m24s复制代码
果然失败了，describe 一下看看原因:
[root@lan1 ~]# kubectl describe pod -n longhorn-system csi-attacher-75588bff58-qrbfg... Warning Failed 36s kubelet Error: ErrImagePull Normal BackOff 36s kubelet Back-off pulling image "k8s.gcr.io/sig-storage/csi-attacher:v3.2.1" Warning Failed 36s kubelet Error: ImagePullBackOff Normal Pulling 24s (x2 over 5m2s) kubelet Pulling image "k8s.gcr.io/sig-storage/csi-attacher:v3.2.1"复制代码
可以看到,pod 从http://k8s.gcr.io拉取镜像失败了,这时候只需要可以 helm upgrade 一下将http://k8s.gcr.io都修改为lank8s.cn就可以了.
下面是 helm upgrade 更新 longhorn 的命令:
helm upgrade longhorn longhorn/longhorn --namespace longhorn-system --set image.csi.attacher.repository=lank8s.cn-storage/csi-attacher --set image.csi.provisioner.repository=lank8s.cn-storage/csi-provisioner --set image.csi.nodeDriverRegistrar.repository=lank8s.cn-storage/csi-node-driver-registrar --set image.csi.resizer.repository=lank8s.cn-storage/csi-resizer --set image.csi.snapshotter.repository=lank8s.cn-storage/csi-snapshotter --set persistence.defaultClassReplicaCount=1 --set csi.attacherReplicaCount=1 --set csi.provisionerReplicaCount=1 --set csi.resizerReplicaCount=1 --set csi.snapshotterReplicaCount=1复制代码
为了最简部署多了一些设置副本数为 1 的设置,如果希望使用默认的副本数的话可以用下面的 helm upgrade 命令:
helm upgrade longhorn longhorn/longhorn --namespace longhorn-system --set image.csi.attacher.repository=lank8s.cn-storage/csi-attacher --set image.csi.provisioner.repository=lank8s.cn-storage/csi-provisioner --set image.csi.nodeDriverRegistrar.repository=lank8s.cn-storage/csi-node-driver-registrar --set image.csi.resizer.repository=lank8s.cn-storage/csi-resizer --set image.csi.snapshotter.repository=lank8s.cn-storage/csi-snapshotter复制代码
执行命令,等一会看看部署的情况，正常的话应该都 running 了:
[root@lan1 ~]# kubectl get pod -n longhorn-systemNAME READY STATUS RESTARTS AGEcsi-attacher-6d74f5876-w4snx 1/1 Running 0 2m41scsi-provisioner-dc6b54764-58zcx 1/1 Running 0 2m38scsi-resizer-6779c54465-mbtp9 1/1 Running 0 2m35scsi-snapshotter-8ccc478c7-pw6s8 1/1 Running 0 2m32sengine-image-ei-d4c780c6-n52pq 1/1 Running 0 4m39sinstance-manager-e-6a94c075 1/1 Running 0 4m39sinstance-manager-r-b7ec2a58 1/1 Running 0 4m38slonghorn-csi-plugin-bzf8r 1/2 ImagePullBackOff 0 4m26slonghorn-driver-deployer-78964dfc64-57wnw 1/1 Running 0 3m3slonghorn-manager-mbpmv 1/1 Running 0 4m58slonghorn-ui-75ccbd4695-5zhzd 1/1 Running 0 4m58s复制代码
奇怪,longhorn-csi-plugin 的 pod 没部署起来，而且报的是拉取镜像失败的错误,用 kubectl describe 看一下拉取的镜像有没有对不对,确认下是不是设置错误或者拼写错误.
[root@lan1 ~]# kubectl get pod -n longhorn-system...Events: Type Reason Age From Message ---- ------ ---- ---- ------- Normal Scheduled 14m default-scheduler Successfully assigned longhorn-system/longhorn-csi-plugin-bzf8r to lan1 Normal Pulled 10m kubelet Container image "longhornio/longhorn-manager:v1.2.2" already present on machine Normal Created 10m kubelet Created container longhorn-csi-plugin Normal Started 10m kubelet Started container longhorn-csi-plugin Warning Failed 8m14s (x3 over 10m) kubelet Failed to pull image "k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0": rpc error: code = Unknown desc = Error response from daemon: Get "https://k8s.gcr.io/v2/": net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers) Normal BackOff 7m36s (x6 over 10m) kubelet Back-off pulling image "k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0" Normal Pulling 7m24s (x4 over 13m) kubelet Pulling image "k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0" Warning Failed 7m9s (x4 over 10m) kubelet Error: ErrImagePull Warning Failed 3m42s (x20 over 10m) kubelet Error: ImagePullBackOff复制代码
可以看到依然是拉取http://k8s.gcr.io的镜像去了,确认了一下命令没有问题,不确定哪里的问题,只好卸载 longhorn 再 install.
设置副本数版本:
helm uninstall longhorn -n longhorn-systemhelm install longhorn longhorn/longhorn --namespace longhorn-system --set image.csi.attacher.repository=lank8s.cn-storage/csi-attacher --set image.csi.provisioner.repository=lank8s.cn-storage/csi-provisioner --set image.csi.nodeDriverRegistrar.repository=lank8s.cn-storage/csi-node-driver-registrar --set image.csi.resizer.repository=lank8s.cn-storage/csi-resizer --set image.csi.snapshotter.repository=lank8s.cn-storage/csi-snapshotter --set persistence.defaultClassReplicaCount=1 --set csi.attacherReplicaCount=1 --set csi.provisionerReplicaCount=1 --set csi.resizerReplicaCount=1 --set csi.snapshotterReplicaCount=1复制代码
不设置副本数版本:
helm uninstall longhorn -n longhorn-systemhelm install longhorn longhorn/longhorn --namespace longhorn-system --set image.csi.attacher.repository=lank8s.cn-storage/csi-attacher --set image.csi.provisioner.repository=lank8s.cn-storage/csi-provisioner --set image.csi.nodeDriverRegistrar.repository=lank8s.cn-storage/csi-node-driver-registrar --set image.csi.resizer.repository=lank8s.cn-storage/csi-resizer --set image.csi.snapshotter.repository=lank8s.cn-storage/csi-snapshotter复制代码
这时候再看下 pod,所有 pod 都起来了.
[root@lan1 ~]# k get po -n longhorn-systemNAME READY STATUS RESTARTS AGEcsi-attacher-6d74f5876-lvpdn 1/1 Running 0 63scsi-provisioner-dc6b54764-mw9c8 1/1 Running 0 63scsi-resizer-6779c54465-jbl4p 1/1 Running 0 62scsi-snapshotter-8ccc478c7-hbpqp 1/1 Running 0 61sengine-image-ei-d4c780c6-pqb9d 1/1 Running 0 75sinstance-manager-e-e16c3ce8 1/1 Running 0 75sinstance-manager-r-307de21a 1/1 Running 0 74slonghorn-csi-plugin-77t5v 2/2 Running 0 61slonghorn-driver-deployer-78964dfc64-ls24l 1/1 Running 0 97slonghorn-manager-tkmvx 1/1 Running 0 97slonghorn-ui-75ccbd4695-c9rn6 1/1 Running 0 97s复制代码
后面就可以愉快的使用 longhorn 存储数据了.
后续
后续会实现一个 webhook,效果是部署 webhook 后会部署/更新阶段将http://k8s.gcr.io替换为lank8s.cn,这样就减少了人工成本,不再需要去修改部署 helm chart 时的命令或者去修改 yaml 了.
当部署了这样的一个 webhook 之后,就可以像以前一样使用简短的命令就可以愉快的使用 longhorn 了,例如:helm install longhorn -n longhorn-system,一起期待吧!
References
[1] 官网文档: https://longhorn.io/docs/1.2.2/
划线
评论
复制
发布于: 刚刚阅读数: 2
版权声明: 本文为 InfoQ 作者【远鹏】的原创文章。
原文链接:【http://xie.infoq.cn/article/d8345b94f7c606947980eedbe】。文章转载请联系作者。
远鹏
关注
天道酬勤 2018.04.28 加入
爱好开源，追随云原生。 github.com/liangyuanpeng 公众号：四颗咖啡豆
评论