K8S中手动扩容云盘数据卷
2019-09-19 本文已影响0人
阿里云云栖号
介绍：
云盘数据卷扩容包括以下几个部分：
云盘物理空间扩容，需要在云盘控制台操作；
文件系统扩容，需要挂载云盘到一个物理节点手动操作；
PV、PVC Size更新，需要更新StorageClass、PVC；
注意：扩容云盘前需要为云盘打快照，以预防升级失败导致数据丢失；
云盘目前无法做到在线扩容，需要重启应用才可以实现，可以通过两种方法实现云盘扩容。下面以集群中部署Zookeeper为例介绍两种扩容方式，Zookeeper集群如下：
# kubectl get pod
NAME READY STATUS RESTARTS AGE
zookeeper-default-zookeeper-0 1/1 Running 0 2m55s
zookeeper-default-zookeeper-1 1/1 Running 0 2m14s
zookeeper-default-zookeeper-2 1/1 Running 0 94s
# kubectl get pvc| grep zoo
datadir-zookeeper-default-zookeeper-0 Bound d-8vb5teafaoa80ia7affg 20Gi RWO alicloud-disk-efficiency 3m12s
datadir-zookeeper-default-zookeeper-1 Bound d-8vb60faf6epslbctnzka 20Gi RWO alicloud-disk-efficiency 2m31s
datadir-zookeeper-default-zookeeper-2 Bound d-8vbidmq57w4df6k84zem 20Gi RWO alicloud-disk-efficiency 111s
# kubectl get pv| grep zoo
d-8vb5teafaoa80ia7affg 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 3m17s
d-8vb60faf6epslbctnzka 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-1 alicloud-disk-efficiency 2m33s
d-8vbidmq57w4df6k84zem 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-2 alicloud-disk-efficiency 107s
暂停应用方式：
如果应用可以暂停服务，可以先停掉应用（删除），手动对每个依赖数据盘进行扩容，然后再启动应用。缺点是：应用会暂停一定时间；
上面应用使用了3个20Gi的云盘，分别挂载在3个pod上。目标为把三个云盘扩容到30Gi，主要步骤：
- 删除应用负载；
- 云盘控制台在线扩容云盘；
- 挂载云盘到某节点进行文件系统扩容；
- 更新PV、PVC的Size参数；
- 重启应用；
1. 删除应用：
删除zookeeper statefulset对象；
# kubectl delete sts zookeeper-default-zookeeper
# kubectl get pvc | grep zoo
datadir-zookeeper-default-zookeeper-0 Bound d-8vb6ie0kwtyynpf4gu4l 20Gi RWO alicloud-disk-efficiency 22m
datadir-zookeeper-default-zookeeper-1 Bound d-8vbhscszlr47rbot0boc 20Gi RWO alicloud-disk-efficiency 21m
datadir-zookeeper-default-zookeeper-2 Bound d-8vb444t0f8xnicj9c2ov 20Gi RWO alicloud-disk-efficiency 21m
# kubectl get pv | grep zoo
d-8vb444t0f8xnicj9c2ov 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-2 alicloud-disk-efficiency 21m
d-8vb6ie0kwtyynpf4gu4l 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 22m
d-8vbhscszlr47rbot0boc 20Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-1 alicloud-disk-efficiency 21m
2.3. 云盘扩容：
根据云盘文档，分别对3块云盘进行扩容：
https://help.aliyun.com/document_detail/113316.html
https://help.aliyun.com/document_detail/25452.html
注意：扩容云盘的同时也要完成对文件系统的扩容。
4. 扩容PV、PVC
编辑pvc、pv所使用的StorageClass，配置allowVolumeExpansion: true标签；
# kubectl edit sc alicloud-disk-efficiency
# kubectl get sc alicloud-disk-efficiency -o yaml
allowVolumeExpansion: true
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
creationTimestamp: "2019-09-05T12:30:27Z"
name: alicloud-disk-efficiency
resourceVersion: "1675896"
selfLink: /apis/storage.k8s.io/v1/storageclasses/alicloud-disk-efficiency
uid: f1071bcc-cfd8-11e9-81cd-00163e0804c2
parameters:
type: cloud_efficiency
provisioner: alicloud/disk
reclaimPolicy: Delete
volumeBindingMode: Immediate
更新PVC的Size为30Gi：
# kubectl patch pvc datadir-zookeeper-default-zookeeper-0 -p '{"spec":{"resources":{"requests":{"storage":"30Gi"}}}}'
persistentvolumeclaim/datadir-zookeeper-default-zookeeper-0 patched
# kubectl patch pvc datadir-zookeeper-default-zookeeper-1 -p '{"spec":{"resources":{"requests":{"storage":"30Gi"}}}}'
persistentvolumeclaim/datadir-zookeeper-default-zookeeper-1 patched
# kubectl patch pvc datadir-zookeeper-default-zookeeper-2 -p '{"spec":{"resources":{"requests":{"storage":"30Gi"}}}}'
persistentvolumeclaim/datadir-zookeeper-default-zookeeper-2 patched
# kubectl get pvc | grep zoo
datadir-zookeeper-default-zookeeper-0 Bound d-8vb6ie0kwtyynpf4gu4l 20Gi RWO alicloud-disk-efficiency 51m
datadir-zookeeper-default-zookeeper-1 Bound d-8vbhscszlr47rbot0boc 20Gi RWO alicloud-disk-efficiency 50m
datadir-zookeeper-default-zookeeper-2 Bound d-8vb444t0f8xnicj9c2ov 20Gi RWO alicloud-disk-efficiency 49m
# kubectl get pv | grep zoo
d-8vb444t0f8xnicj9c2ov 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-2 alicloud-disk-efficiency 49m
d-8vb6ie0kwtyynpf4gu4l 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 51m
d-8vbhscszlr47rbot0boc 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-1 alicloud-disk-efficiency 50m
可见pv的大小已经更新为30Gi，pvc的大小会在Pod启动后自动更新；
5. 重新创建应用：
# kubectl get pod
NAME READY STATUS RESTARTS AGE
zookeeper-default-zookeeper-0 1/1 Running 0 94s
zookeeper-default-zookeeper-1 1/1 Running 0 64s
zookeeper-default-zookeeper-2 1/1 Running 0 39s
# kubectl exec -ti zookeeper-default-zookeeper-0 sh
# df -h |grep zoo
/dev/vdb 30G 45M 30G 1% /var/lib/zookeeper
# kubectl get pvc | grep zoo
datadir-zookeeper-default-zookeeper-0 Bound d-8vb6ie0kwtyynpf4gu4l 30Gi RWO alicloud-disk-efficiency 56m
datadir-zookeeper-default-zookeeper-1 Bound d-8vbhscszlr47rbot0boc 30Gi RWO alicloud-disk-efficiency 56m
datadir-zookeeper-default-zookeeper-2 Bound d-8vb444t0f8xnicj9c2ov 30Gi RWO alicloud-disk-efficiency 55m
# kubectl get pv | grep zoo
d-8vb444t0f8xnicj9c2ov 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-2 alicloud-disk-efficiency 55m
d-8vb6ie0kwtyynpf4gu4l 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 56m
d-8vbhscszlr47rbot0boc 30Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-1 alicloud-disk-efficiency 56m
上面输出日志可以看出，文件系统已经扩容到30G，pv pvc的size也扩容到30G；
逐个Pod升级方式：
由多个Pod运行的服务，可以通过一个一个pod升级的方式实现业务不中断，而只是Running pod数量暂时减少。
还是以zookeeper为例，使用3个20Gi的云盘，分别挂载在3个pod上。目标为把三个云盘扩容到40Gi：
主要步骤：
- 修改某个Pod所对应pv的Zoneid label，然后删除pod；
- 登陆云盘控制台在线扩容云盘（扩容前最好对云盘打快照，容灾）；
- 挂载云盘到某节点进行文件系统扩容；
- 更新PV、PVC 的Size参数，以及Pv的ZoneId Lable；
- 对其他Pod重复1-4步骤；
1. 修改PV的ZoneId
# kubectl get pod zookeeper-default-zookeeper-0 -oyaml | grep pers -C 1
- name: datadir
persistentVolumeClaim:
claimName: datadir-zookeeper-default-zookeeper-0
# kubectl get pvc datadir-zookeeper-default-zookeeper-0
NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
datadir-zookeeper-default-zookeeper-0 Bound d-8vbhscszlr47tgn0eheb 20Gi RWO alicloud-disk-efficiency 21m
上述日志得出zookeeper-default-zookeeper-0 使用的pvc为：datadir-zookeeper-default-zookeeper-0，对应的PV为d-8vbhscszlr47tgn0eheb；
更新PV Label，
如果pv中已经有如下label，则把zone对应的值改为”原来value -none“，即配置一个不存在的zone，让pod无法调度成功；
如果pv中没有如下label，则添加这些lables到pv；
#kubectl edit pv d-8vbhscszlr47tgn0eheb
labels:
failure-domain.beta.kubernetes.io/region: cn-zhangjiakou
failure-domain.beta.kubernetes.io/zone: cn-zhangjiakou-a-none
删除Pod：zookeeper-default-zookeeper-0
# kubectl delete pod zookeeper-default-zookeeper-0
这时删除的pod一直处于Pending状态：
# kubectl get pod
NAME READY STATUS RESTARTS AGE
zookeeper-default-zookeeper-0 0/1 Pending 0 9s
zookeeper-default-zookeeper-1 1/1 Running 0 24m
zookeeper-default-zookeeper-2 1/1 Running 0 24m
2.3. 云盘扩容：
根据云盘文档，分别对3块云盘进行库容：
https://help.aliyun.com/document_detail/113316.html
https://help.aliyun.com/document_detail/25452.html
注意：扩容云盘的同时也要完成对文件系统的扩容。
4. 扩容PV、PVC
编辑pvc、pv所使用的StorageClass，配置allowVolumeExpansion: true标签；
# kubectl edit sc alicloud-disk-efficiency
# kubectl get sc alicloud-disk-efficiency -o yaml
allowVolumeExpansion: true
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
creationTimestamp: "2019-09-05T12:30:27Z"
name: alicloud-disk-efficiency
resourceVersion: "1675896"
selfLink: /apis/storage.k8s.io/v1/storageclasses/alicloud-disk-efficiency
uid: f1071bcc-cfd8-11e9-81cd-00163e0804c2
parameters:
type: cloud_efficiency
provisioner: alicloud/disk
reclaimPolicy: Delete
volumeBindingMode: Immediate
更新PVC的Size为40Gi：
# kubectl patch pvc datadir-zookeeper-default-zookeeper-0 -p '{"spec":{"resources":{"requests":{"storage":"40Gi"}}}}'
persistentvolumeclaim/datadir-zookeeper-default-zookeeper-0 patched
# kubectl get pvc | grep datadir-zookeeper-default-zookeeper-0
datadir-zookeeper-default-zookeeper-0 Bound d-8vbhscszlr47tgn0eheb 20Gi RWO alicloud-disk-efficiency 29m
# kubectl get pv | grep datadir-zookeeper-default-zookeeper-0
d-8vbhscszlr47tgn0eheb 40Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 29m
可见pv的大小已经更新为40Gi，pvc的大小会在Pod启动后自动更新；
恢复PV的Lable，把zoneId的值恢复之前的value，删除相应labels（若之前没有这些labels）：
labels:
failure-domain.beta.kubernetes.io/region: cn-zhangjiakou
failure-domain.beta.kubernetes.io/zone: cn-zhangjiakou-a
检查容器挂载文件系统大小，pv、pvc大小：
# kubectl exec -ti zookeeper-default-zookeeper-0 sh
# df -h | grep zoo
/dev/vdb 40G 48M 40G 1% /var/lib/zookeeper
# kubectl get pvc | grep datadir-zookeeper-default-zookeeper-0
datadir-zookeeper-default-zookeeper-0 Bound d-8vbhscszlr47tgn0eheb 40Gi RWO alicloud-disk-efficiency 33m
# kubectl get pv | grep datadir-zookeeper-default-zookeeper-0
d-8vbhscszlr47tgn0eheb 40Gi RWO Delete Bound default/datadir-zookeeper-default-zookeeper-0 alicloud-disk-efficiency 33m
可见文件系统、pv、pvc都已经实现了扩容；
5. 对其他pod所对应的pvc、pv、云盘进行上述扩容
本文作者：阚俊宝
本文为云栖社区原创内容，未经允许不得转载。
