rbd-provisioner为kubernetes 1.5+版本提供了类似于kubernetes.io/rbd的ceph rbd持久化存储动态配置实现。

一些用户会使用kubeadm来部署集群，或者将kube-controller-manager以容器的方式运行。这种方式下，kubernetes在创建使用ceph rbd pv/pvc时没任何问题，但使用dynamic provisioning自动管理存储生命周期时会报错。提示"rbd: create volume failed, err: failed to create rbd image: executable file not found in $PATH:"。

问题来自gcr.io提供的kube-controller-manager容器镜像未打包ceph-common组件，缺少了rbd命令，因此无法通过rbd命令为pod创建rbd image，查了github的相关文章，目前kubernetes官方在kubernetes-incubator/external-storage项目通过External Provisioners的方式来解决此类问题。

本文主要针对该问题，通过rbd-provisioner的方式，解决ceph rbd的dynamic provisioning问题。

部署rbd-provisioner

首先得在kubernetes集群中安装rbd-provisioner，github仓库链接https://github.com/kubernetes-incubator/external-storage

[root@k8s01 ~]# git clone https://github.com/kubernetes-incubator/external-storage.git
[root@k8s01 ~]# cd external-storage/ceph/rbd/deploy
[root@k8s01 deploy]# NAMESPACE=kube-system
[root@k8s01 deploy]# sed -r -i "s/namespace: [^ ]+/namespace: $NAMESPACE/g" ./rbac/clusterrolebinding.yaml ./rbac/rolebinding.yaml
[root@k8s01 deploy]# kubectl -n $NAMESPACE apply -f ./rbac

根据自己需要，修改rbd-provisioner的namespace；

部署完成后检查rbd-provisioner deployment，确保已经正常部署

[root@k8s01 ~]# kubectl describe deployments.apps -n kube-system rbd-provisioner
Name:               rbd-provisioner
Namespace:          kube-system
CreationTimestamp:  Sat, 13 Oct 2018 20:08:45 +0800
Labels:             app=rbd-provisioner
Annotations:        deployment.kubernetes.io/revision: 1
                    kubectl.kubernetes.io/last-applied-configuration:
                      {"apiVersion":"extensions/v1beta1","kind":"Deployment","metadata":{"annotations":{},"name":"rbd-provisioner","namespace":"kube-system"},"s...
Selector:           app=rbd-provisioner
Replicas:           1 desired | 1 updated | 1 total | 1 available | 0 unavailable
StrategyType:       Recreate
MinReadySeconds:    0
Pod Template:
  Labels:           app=rbd-provisioner
  Service Account:  rbd-provisioner
  Containers:
   rbd-provisioner:
    Image:      quay.io/external_storage/rbd-provisioner:latest
    Port:       <none>
    Host Port:  <none>
    Environment:
      PROVISIONER_NAME:  ceph.com/rbd
    Mounts:              <none>
  Volumes:               <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
OldReplicaSets:  <none>
NewReplicaSet:   rbd-provisioner-db574c5c (1/1 replicas created)
Events:          <none>

创建StorageClass

ceph配置

[root@k8s01 ~]# ceph osd pool create rbd 128 ##创建池
[root@k8s01 ~]# ceph osd pool application enable rbd rbd

[root@k8s01 ~]# ceph auth get-or-create client.kubernetes mon 'allow r' osd 'allow class-read object_prefix rbd_children,allow rwx pool=rbd' ##创建用户
[root@k8s01 ~]# ceph auth get-or-create client.kubernetes | tee /etc/ceph/ceph.client.kubernetes.keyring ##每个k8s节点执行

[root@k8s01 ~]# ceph auth list |grep client.kubernetes -A 3 ## 
client.kubernetes
        key: AQB2UkZf2RTVGBAADVaRxeWDWEpC0YssEF6o6A==
        caps: [mon] allow r
        caps: [osd] allow class-read object_prefix rbd_children,allow rwx pool=rbd
        
[root@k8s01 ~]# ceph auth get-key client.admin | base64  ##获取client.kubernetes base64 key 下面会用到 
QVFBWHFrTmZrN2NZR1JBQXQySGFJS0syTzVKUEVjMG5YVUltbnc9PQ==
[root@k8s01 ~]# ceph auth get-key client.kubernetes | base64 ##获取client.admin base64 key 下面会用到
QVFCMlVrWmYyUlRWR0JBQURWYVJ4ZVdEV0VwQzBZRHdFRjZvNkE9PQ==
[root@k8s02 ~]# cat /etc/ceph/ceph.client.kubernetes.keyring  ##要保证每台k8s节点有这个文件
[client.kubernetes]
        key = AQB2UkZf2RTVGBAADVaRxeWDWEpC0YDwEF6o6A==

ceph端配置完毕

创建StorageClass。创建SC前，需要创建相关用户的secret

[root@k8s01 ~]# vi secrets.yaml
apiVersion: v1
kind: Secret
metadata:
  name: ceph-admin-secret
  namespace: kube-system
type: "kubernetes.io/rbd"
data:
  # ceph auth get-key client.admin | base64
  key: QVFBWHFrTmZrN2NZR1JBQXQySGFJS0syTzVKUEVjMG5YVUltbnc9PQ==
---
apiVersion: v1
kind: Secret
metadata:
  name: ceph-secret
  namespace: kube-system
type: "kubernetes.io/rbd"
data:
  # ceph auth add client.kube mon 'allow r' osd 'allow rwx pool=kube'
  # ceph auth get-key client.kube | base64
  key: QVFCMlVrWmYyUlRWR0JBQURWYVJ4ZVdEV0VwQzBZRHdFRjZvNkE9PQ==

[root@k8s01 ~]#  kubectl create -f secrets.yaml

[root@k8s01 ~]# vi secrets-default.yaml
apiVersion: v1
kind: Secret
metadata:
  name: ceph-secret
type: "kubernetes.io/rbd"
data:
  # ceph auth add client.kube mon 'allow r' osd 'allow rwx pool=kube'
  # ceph auth get-key client.kube | base64
  key: QVFCMlVrWmYyUlRWR0JBQURWYVJ4ZVdEV0VwQzBZRHdFRjZvNkE9PQ==

[root@k8s01 ~]#  kubectl create -f secrets-default.yaml -n default

创建secret保存client.admin和client.kube用户的key，client.admin和client.kube用户的secret可以放在kube-system namespace，但如果其他namespace需要使用ceph rbd的dynamic provisioning功能的话，要在相应的namespace创建secret来保存client.kube用户key信息

[root@k8s01 ~]# vi ceph-rbd-sc.yaml
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: ceph-rbd
  annotations:
     storageclass.beta.kubernetes.io/is-default-class: "true"
provisioner: ceph.com/rbd
parameters:
  monitors: 10.100.22.145:6789,10.100.22.146:6789,10.100.22.147:6789
  adminId: admin
  adminSecretName: ceph-admin-secret
  adminSecretNamespace: kube-system
  pool: rbd
  userId: kubernetes
  userSecretName: ceph-secret
  fsType: ext4
  imageFormat: "2"
  imageFeatures: "layering"

[root@k8s01 ~]#  kubectl create -f  ceph-rbd-sc.yaml

其他设置和普通的ceph rbd StorageClass一致，但provisioner需要设置为ceph.com/rbd，不是默认的kubernetes.io/rbd，这样rbd的请求将由rbd-provisioner来处理

考虑到兼容性，建议尽量关闭rbd image feature，并且kubelet节点的ceph-common版本尽量和ceph服务器端保持一致，我的环境使用的14.2.11版本

[root@k8s01 ceph]# rpm -qa |grep ceph-common
ceph-common-14.2.11-0.el7.x86_64

[root@k8s01 ceph]# kubectl get storageclasses.storage.k8s.io 
NAME                 PROVISIONER    RECLAIMPOLICY   VOLUMEBINDINGMODE   ALLOWVOLUMEEXPANSION   AGE
ceph-rbd (default)   ceph.com/rbd   Delete          Immediate           false                  21m

测试ceph rbd自动分配

在kube-system和default namespace分别创建pod，通过启动一个busybox实例，将ceph rbd镜像挂载到/usr/share/busybox

[root@k8s01 ~]# vi test-pod.yaml
apiVersion: v1
kind: Pod
metadata:
  name: ceph-pod1
spec:
  containers:
  - name: ceph-busybox
    image: busybox
    command: ["sleep", "60000"]
    volumeMounts:
    - name: ceph-vol1
      mountPath: /usr/share/busybox
      readOnly: false
  volumes:
  - name: ceph-vol1
    persistentVolumeClaim:
      claimName: ceph-claim
---
kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: ceph-claim
spec:
  accessModes:  
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi

[root@k8s01 ~]# kubectl create -f test-pod.yaml -n kube-system
pod/ceph-pod1 created
persistentvolumeclaim/ceph-claim created
[root@k8s01 ~]# kubectl create -f test-pod.yaml -n default
pod/ceph-pod1 created
persistentvolumeclaim/ceph-claim created

检查pv和pvc的创建状态，是否都已经创建

[root@k8s01 ceph]# kubectl get pv,pvc
NAME                                                        CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                    STORAGECLASS   REASON   AGE
persistentvolume/pvc-b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4   1Gi        RWO            Delete           Bound    default/ceph-claim       ceph-rbd                23m
persistentvolume/pvc-e1b4bdac-7b55-4126-b0cc-f934b46f0dbd   1Gi        RWO            Delete           Bound    kube-system/ceph-claim   ceph-rbd                20s

NAME                               STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
persistentvolumeclaim/ceph-claim   Bound    pvc-b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4   1Gi        RWO            ceph-rbd       23m

在ceph服务器上，检查rbd镜像创建情况和镜像的信息

[root@k8s01 ceph]# rbd ls --pool rbd
kubernetes-dynamic-pvc-c3c2d52b-e796-11ea-bb7c-96ab287be34a
kubernetes-dynamic-pvc-0fb1f715-e79a-11ea-bb7c-96ab287be34a

[root@k8s01 ceph]#  rbd info rbd/kubernetes-dynamic-pvc-c3c2d52b-e796-11ea-bb7c-96ab287be34a
rbd image 'kubernetes-dynamic-pvc-c3c2d52b-e796-11ea-bb7c-96ab287be34a':
        size 1 GiB in 256 objects
        order 22 (4 MiB objects)
        snapshot_count: 0
        id: 100726b8b4567
        block_name_prefix: rbd_data.100726b8b4567
        format: 2
        features: layering
        op_features: 
        flags: 
        create_timestamp: Wed Aug 26 20:22:10 2020
        access_timestamp: Wed Aug 26 20:22:10 2020
        modify_timestamp: Wed Aug 26 20:22:10 2020
[root@k8s01 ceph]#  rbd info rbd/kubernetes-dynamic-pvc-0fb1f715-e79a-11ea-bb7c-96ab287be34a
rbd image 'kubernetes-dynamic-pvc-0fb1f715-e79a-11ea-bb7c-96ab287be34a':
        size 1 GiB in 256 objects
        order 22 (4 MiB objects)
        snapshot_count: 0
        id: da766b8b4567
        block_name_prefix: rbd_data.da766b8b4567
        format: 2
        features: layering
        op_features: 
        flags: 
        create_timestamp: Wed Aug 26 20:45:46 2020
        access_timestamp: Wed Aug 26 20:45:46 2020
        modify_timestamp: Wed Aug 26 20:45:46 2020

检查busybox内的文件系统挂载和使用情况，确认能正常工作

[root@k8s01 ~]# kubectl exec -it ceph-pod1 mount |grep rbd
/dev/rbd0 on /usr/share/busybox type ext4 (rw,seclabel,relatime,stripe=1024,data=ordered)
[root@k8s01 ~]# kubectl exec -it -n kube-system ceph-pod1 mount |grep rbd
/dev/rbd0 on /usr/share/busybox type ext4 (rw,seclabel,relatime,stripe=1024,data=ordered)

[root@k8s01 ~]# kubectl exec -it -n kube-system ceph-pod1 df |grep rbd
/dev/rbd0              1998672      6144   1976144   0% /usr/share/busybox
[root@k8s01 ~]# kubectl exec -it ceph-pod1 df |grep rbd
/dev/rbd0              1998672      6144   1976144   0% /usr/share/busybox

测试删除pod能否自动删除pv和pvc，生产环境中谨慎，设置好回收策略

[root@k8s01 ~]# kubectl delete -f test-pod.yaml
pod "ceph-pod1" deleted
persistentvolumeclaim "ceph-claim" deleted

[root@k8s01 ~]# kubectl delete -f test-pod.yaml -n kube-system
pod "ceph-pod1" deleted
persistentvolumeclaim "ceph-claim" deleted

[root@k8s01 ~]# kubectl get pv
No resources found.
[root@k8s01 ~]# kubectl get pvc
No resources found.
[root@k8s01 ~]# kubectl get pvc -n kube-system
No resources found.

请注意：删除测试 rbd 设备会被删除

可以在StorageClass中设置删除策略reclaimPolicy: Retain

ceph服务器上的rbd image也已清除，自动回收成功

[root@k8s01 ~]# rbd ls --pool rbd

确认之前创建的rbd images都已经删除

总结

大部分情况下，我们无需使用rbd provisioner来提供ceph rbd的dynamic provisioning能力。经测试，在OpenShift、Rancher、SUSE CaaS以及本Handbook的二进制文件方式部署，在安装好ceph-common软件包的情况下，定义StorageClass时使用kubernetes.io/rbd即可正常使用ceph rbd provisioning功能

问题

1.auth: unable to find a keyring on /etc/ceph/ceph.client.kube.keyring,/etc/ceph/ceph.keyring,/etc/ceph/keyring,/etc/ceph/keyring.bin,: (2) No such file or directory

报错中找不到/etc/ceph/ceph.client.kube.keyring 手动复制一个文件就好了

每个k8s节点执行以下命令
[root@k8s03 ~]# cd /etc/ceph/
[root@k8s03 ceph]# cp ceph.client.admin.keyring ceph.client.kube.keyring 

pv 和 ceph rbd 对应关系查看

注意观察kubectl get pv xxxx -o yaml spec.rbd.image的值

[root@k8s01 ceph]# kubectl get pv pvc-b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4 -o yaml
apiVersion: v1
kind: PersistentVolume
metadata:
  annotations:
    pv.kubernetes.io/provisioned-by: ceph.com/rbd
    rbdProvisionerIdentity: ceph.com/rbd
  creationTimestamp: "2020-08-26T12:22:10Z"
  finalizers:
  - kubernetes.io/pv-protection
  managedFields:
  - apiVersion: v1
    fieldsType: FieldsV1
    fieldsV1:
      f:status:
        f:phase: {}
    manager: kube-controller-manager
    operation: Update
    time: "2020-08-26T12:22:10Z"
  - apiVersion: v1
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:annotations:
          .: {}
          f:pv.kubernetes.io/provisioned-by: {}
          f:rbdProvisionerIdentity: {}
      f:spec:
        f:accessModes: {}
        f:capacity:
          .: {}
          f:storage: {}
        f:claimRef:
          .: {}
          f:apiVersion: {}
          f:kind: {}
          f:name: {}
          f:namespace: {}
          f:resourceVersion: {}
          f:uid: {}
        f:persistentVolumeReclaimPolicy: {}
        f:rbd:
          .: {}
          f:fsType: {}
          f:image: {}
          f:keyring: {}
          f:monitors: {}
          f:pool: {}
          f:secretRef:
            .: {}
            f:name: {}
          f:user: {}
        f:storageClassName: {}
        f:volumeMode: {}
    manager: rbd-provisioner
    operation: Update
    time: "2020-08-26T12:22:10Z"
  name: pvc-b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4
  resourceVersion: "17424624"
  selfLink: /api/v1/persistentvolumes/pvc-b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4
  uid: 56e22d57-f95b-484b-b3aa-e37394785fba
spec:
  accessModes:
  - ReadWriteOnce
  capacity:
    storage: 1Gi
  claimRef:
    apiVersion: v1
    kind: PersistentVolumeClaim
    name: ceph-claim
    namespace: default
    resourceVersion: "17424619"
    uid: b3fd4b59-eb33-4341-bc0f-d5049b7f8ce4
  persistentVolumeReclaimPolicy: Delete
  rbd:
    fsType: ext4
    image: kubernetes-dynamic-pvc-c3c2d52b-e796-11ea-bb7c-96ab287be34a ##注意观察是否和rdb 的image一致
    keyring: /etc/ceph/keyring
    monitors:
    - 10.100.22.145:6789
    - 10.100.22.146:6789
    - 10.100.22.147:6789
    pool: rbd
    secretRef:
      name: ceph-secret
    user: kubernetes
  storageClassName: ceph-rbd
  volumeMode: Filesystem
status:
  phase: Bound
  
[root@k8s01 ceph]# rbd ls --pool rbd
kubernetes-dynamic-pvc-c3c2d52b-e796-11ea-bb7c-96ab287be34a ##和这项一致
kubernetes-dynamic-pvc-fb708fd7-e783-11ea-9080-be6293f131e0