Rook 1.5.1 部署Ceph实操经验分享
一、Rook概述
1.1 Rook简介
Rook 是一个开源的cloud-native storage编排, 提供平台和框架；为各种存储解决方案提供平台、框架和支持，以便与云原生环境本地集成。目前主要专用于Cloud-Native环境的文件、块、对象存储服务。它实现了一个自我管理的、自我扩容的、自我修复的分布式存储服务。
Rook支持自动部署、启动、配置、分配（provisioning）、扩容/缩容、升级、迁移、灾难恢复、监控，以及资源管理。为了实现所有这些功能，Rook依赖底层的容器编排平台，例如 kubernetes、CoreOS 等。。
Rook 目前支持Ceph、NFS、Minio Object Store、Edegefs、Cassandra、CockroachDB 存储的搭建。
项目地址： https://github.com/rook/rook
网站： https://rook.io/
1.2 Rook组件
Rook的主要组件有三个，功能如下：
-
Rook Operator
- Rook与Kubernetes交互的组件
- 整个Rook集群只有一个
-
Agent or Driver
- Flex Driver
已经被淘汰的驱动方式，在安装之前，请确保k8s集群版本是否支持CSI，如果不支持，或者不想用CSI，选择flex.
- Ceph CSI Driver
默认全部节点安装，你可以通过 node affinity 去指定节点
-
Device discovery
发现新设备是否作为存储，可以在配置文件
ROOK_ENABLE_DISCOVERY_DAEMON设置 enable 开启。
1.3 Rook & Ceph框架
Rook 如何集成在kubernetes 如图：
使用Rook部署Ceph集群的架构图如下：
部署的Ceph系统可以提供下面三种Volume Claim服务：
- Block Storage：目前最稳定；
- FileSystem：需要部署MDS，有内核要求；
- Object：部署RGW；
二、ROOK 部署
2.1 准备工作
2.1.1 版本要求
kubernetes v.11 以上
2.1.2 存储要求
rook部署的ceph 是不支持lvm direct直接作为osd存储设备的，如果想要使用lvm，可以使用pvc的形式实现。方法在后面的ceph安装会提到
为了配置 Ceph 存储集群，至少需要以下本地存储选项之一：
- 原始设备（无分区或格式化的文件系统）
- 原始分区（无格式文件系统） 可以 lsblk -f查看，如果 FSTYPE不为空说明有文件系统
- 可通过 block 模式从存储类别获得 PV
2.1.3 系统要求
本次安装环境
- kubernetes 1.18
- centos7.8
- kernel 5.4.65-200.el7.x86_64
- calico 3.16
2.1.3.1 需要安装lvm包
sudo yum install -y lvm2
2.1.3.2 内核要求
RBD
一般发行版的内核都编译有，但你最好确定下：
foxchan@~$ lsmod|grep rbd
rbd 114688 0
libceph 368640 1 rbd
可以用以下命令放到开机启动项里
cat > /etc/sysconfig/modules/rbd.modules << EOF
modprobe rbd
EOF
CephFS
如果你想使用cephfs,内核最低要求是4.17。
2.2 部署ROOK
Github上下载Rook最新release
git clone --single-branch --branch v1.5.1 https://github.com/rook/rook.gits
安装公共部分
cd rook/cluster/examples/kubernetes/ceph
kubectl create -f crds.yaml -f common.yaml
安装operator
kubectl apply -f operator.yaml
如果放到生产环境，请提前规划好。operator的配置在ceph安装后不能修改，否则rook会删除集群并重建。
修改内容如下：
# 启用cephfs
ROOK_CSI_ENABLE_CEPHFS: "true"
# 开启内核驱动替换ceph-fuse
CSI_FORCE_CEPHFS_KERNEL_CLIENT: "true"
#修改csi镜像为私有仓，加速部署时间
ROOK_CSI_CEPH_IMAGE: "harbor.foxchan.com/google_containers/cephcsi/cephcsi:v3.1.2"
ROOK_CSI_REGISTRAR_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-node-driver-registrar:v2.0.1"
ROOK_CSI_RESIZER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-resizer:v1.0.0"
ROOK_CSI_PROVISIONER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-provisioner:v2.0.0"
ROOK_CSI_SNAPSHOTTER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-snapshotter:v3.0.0"
ROOK_CSI_ATTACHER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-attacher:v3.0.0"
# 可以设置NODE_AFFINITY 来指定csi 部署的节点
# 我把plugin 和 provisioner分开了，具体调度方式看你集群资源。
CSI_PROVISIONER_NODE_AFFINITY: "app.rook.role=csi-provisioner"
CSI_PLUGIN_NODE_AFFINITY: "app.rook.plugin=csi"
#修改metrics端口，可以不改，我因为集群网络是host，为了避免端口冲突
# Configure CSI CSI Ceph FS grpc and liveness metrics port
CSI_CEPHFS_GRPC_METRICS_PORT: "9491"
CSI_CEPHFS_LIVENESS_METRICS_PORT: "9481"
# Configure CSI RBD grpc and liveness metrics port
CSI_RBD_GRPC_METRICS_PORT: "9490"
CSI_RBD_LIVENESS_METRICS_PORT: "9480"
# 修改rook镜像，加速部署时间
image: harbor.foxchan.com/google_containers/rook/ceph:v1.5.1
# 指定节点做存储
- name: DISCOVER_AGENT_NODE_AFFINITY
value: "app.rook=storage"
# 开启设备自动发现
- name: ROOK_ENABLE_DISCOVERY_DAEMON
value: "true"
2.3 部署ceph集群
cluster.yaml文件里的内容需要修改，一定要适配自己的硬件情况，请详细阅读配置文件里的注释，避免我踩过的坑。
修改内容如下：
此文件的配置，除了增删osd设备外，其他的修改都要重装ceph集群才能生效，所以请提前规划好集群。如果修改后不卸载ceph直接apply，会触发ceph集群重装，导致集群异常挂掉
apiVersion: ceph.rook.io/v1
kind: CephCluster
metadata:
# 命名空间的名字，同一个命名空间只支持一个集群
name: rook-ceph
namespace: rook-ceph
spec:
# ceph版本说明
# v13 is mimic, v14 is nautilus, and v15 is octopus.
cephVersion:
#修改ceph镜像，加速部署时间
image: harbor.foxchan.com/google_containers/ceph/ceph:v15.2.5
# 是否允许不支持的ceph版本
allowUnsupported: false
#指定rook数据在节点的保存路径
dataDirHostPath: /data/rook
# 升级时如果检查失败是否继续
skipUpgradeChecks: false
# 从1.5开始，mon的数量必须是奇数
mon:
count: 3
# 是否允许在单个节点上部署多个mon pod
allowMultiplePerNode: false
mgr:
modules:
- name: pg_autoscaler
enabled: true
# 开启dashboard，禁用ssl，指定端口是7000，你可以默认https配置。我是为了ingress配置省事。
dashboard:
enabled: true
port: 7000
ssl: false
# 开启prometheusRule
monitoring:
enabled: true
# 部署PrometheusRule的命名空间，默认此CR所在命名空间
rulesNamespace: rook-ceph
# 开启网络为host模式，解决无法使用cephfs pvc的bug
network:
provider: host
# 开启crash collector，每个运行了Ceph守护进程的节点上创建crash collector pod
crashCollector:
disable: false
# 设置node亲缘性，指定节点安装对应组件
placement:
mon:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: ceph-mon
operator: In
values:
- enabled
osd:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: ceph-osd
operator: In
values:
- enabled
mgr:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: ceph-mgr
operator: In
values:
- enabled
# 存储的设置，默认都是true，意思是会把集群所有node的设备清空初始化。
storage: # cluster level storage configuration and selection
useAllNodes: false #关闭使用所有Node
useAllDevices: false #关闭使用所有设备
nodes:
- name: "192.168.1.162" #指定存储节点主机
devices:
- name: "nvme0n1p1" #指定磁盘为nvme0n1p1
- name: "192.168.1.163"
devices:
- name: "nvme0n1p1"
- name: "192.168.1.164"
devices:
- name: "nvme0n1p1"
- name: "192.168.1.213"
devices:
- name: "nvme0n1p1"
更多 cluster 的 CRD 配置参考：
- https://github.com/rook/rook/blob/master/Documentation/ceph-cluster-crd.md
执行安装
kubectl apply -f cluster.yaml
# 需要等一段时间，所有pod都已正常启动
[foxchan@k8s-master ceph]$ kubectl get pods -n rook-ceph
NAME READY STATUS RESTARTS AGE
csi-cephfsplugin-b5tlr 3/3 Running 0 19h
csi-cephfsplugin-mjssm 3/3 Running 0 19h
csi-cephfsplugin-provisioner-5cf5ffdc76-mhdgz 6/6 Running 0 19h
csi-cephfsplugin-provisioner-5cf5ffdc76-rpdl8 6/6 Running 0 19h
csi-cephfsplugin-qmvkc 3/3 Running 0 19h
csi-cephfsplugin-tntzd 3/3 Running 0 19h
csi-rbdplugin-4p75p 3/3 Running 0 19h
csi-rbdplugin-89mzz 3/3 Running 0 19h
csi-rbdplugin-cjcwr 3/3 Running 0 19h
csi-rbdplugin-ndjcj 3/3 Running 0 19h
csi-rbdplugin-provisioner-658dd9fbc5-fwkmc 6/6 Running 0 19h
csi-rbdplugin-provisioner-658dd9fbc5-tlxd8 6/6 Running 0 19h
prometheus-rook-prometheus-0 2/2 Running 1 3d17h
rook-ceph-mds-myfs-a-5cbcdc6f9c-7mdsv 1/1 Running 0 19h
rook-ceph-mds-myfs-b-5f4cc54b87-m6m6f 1/1 Running 0 19h
rook-ceph-mgr-a-f98d4455b-bwhw7 1/1 Running 0 20h
rook-ceph-mon-a-5d445d4b8d-lmg67 1/1 Running 1 20h
rook-ceph-mon-b-769c6fd76f-jrlc8 1/1 Running 0 20h
rook-ceph-mon-c-6bfd8954f5-tbsnd 1/1 Running 0 20h
rook-ceph-operator-7d8cc65dc-8wtl8 1/1 Running 0 20h
rook-ceph-osd-0-c558ff759-bzbgw 1/1 Running 0 20h
rook-ceph-osd-1-5c97d69d78-dkxbb 1/1 Running 0 20h
rook-ceph-osd-2-7dddc7fd56-p58mw 1/1 Running 0 20h
rook-ceph-osd-3-65ff985c7d-9gfgj 1/1 Running 0 20h
rook-ceph-osd-prepare-192.168.1.213-pw5gr 0/1 Completed 0 19h
rook-ceph-osd-prepare-192.168.1.162-wtkm8 0/1 Completed 0 19h
rook-ceph-osd-prepare-192.168.1.163-b86r2 0/1 Completed 0 19h
rook-ceph-osd-prepare-192.168.1.164-tj79t 0/1 Completed 0 19h
rook-discover-89v49 1/1 Running 0 20h
rook-discover-jdzhn 1/1 Running 0 20h
rook-discover-sl9bv 1/1 Running 0 20h
rook-discover-wg25w 1/1 Running 0 20h
2.4 增删osd
2.4.1 添加相关label
kubectl label nodes 192.168.1.165 app.rook=storage
kubectl label nodes 192.168.1.165 ceph-osd=enabled
2.4.2 修改cluster.yaml
nodes:
- name: "192.168.1.162"
devices:
- name: "nvme0n1p1"
- name: "192.168.1.163"
devices:
- name: "nvme0n1p1"
- name: "192.168.1.164"
devices:
- name: "nvme0n1p1"
- name: "192.168.17.213"
devices:
- name: "nvme0n1p1"
#添加165的磁盘信息
- name: "192.168.1.165"
devices:
- name: "nvme0n1p1"
2.4.3 apply cluster.yaml
kubectl apply -f cluster.yaml
2.4.4 删除osd
cluster.yaml去掉相关节点，再apply
2.5 安装dashboard
这是我自己的traefik ingress，yaml目录里有很多dashboard暴露方式，自行选择
dashboard已经在前述的步骤中包含了，这里只需要把dashboard service的服务暴露出来。有多种方法，我使用的是ingress的方式来暴露：
apiVersion: traefik.containo.us/v1alpha1
kind: Ingre***oute
metadata:
name: traefik-ceph-dashboard
annotations:
kubernetes.io/ingress.class: traefik-v2.3
spec:
entryPoints:
- web
routes:
- match: Host(`ceph.foxchan.com`)
kind: Rule
services:
- name: rook-ceph-mgr-dashboard
namespace: rook-ceph
port: 7000
middlewares:
- name: gs-ipwhitelist
登录 dashboard 需要安全访问。Rook 在运行 Rook Ceph 集群的名称空间中创建一个默认用户，admin 并生成一个称为的秘密rook-ceph-dashboard-admin-password
要检索生成的密码，可以运行以下命令：
kubectl -n rook-ceph get secret rook-ceph-dashboard-password -o jsonpath="{['data']['password']}" | base64 --decode && echo
2.6 安装toolbox
执行下面的命令：
kubectl apply -f toolbox.yaml
成功后，可以使用下面的命令来确定toolbox的pod已经启动成功：
kubectl -n rook-ceph get pod -l "app=rook-ceph-tools"
然后可以使用下面的命令登录该pod，执行各种ceph命令：
kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- bash
比如：
ceph statusceph osd statusceph dfrados df
删除toolbox
kubectl -n rook-ceph delete deploy/rook-ceph-tools
2.7 prometheus监控
监控部署很简单，利用Prometheus Operator，独立部署一套prometheus
安装prometheus operator
kubectl apply -f https://raw.githubusercontent.com/coreos/prometheus-operator/v0.40.0/bundle.yaml
安装prometheus
git clone --single-branch --branch v1.5.1 https://github.com/rook/rook.git
cd rook/cluster/examples/kubernetes/ceph/monitoring
kubectl create -f service-monitor.yaml
kubectl create -f prometheus.yaml
kubectl create -f prometheus-service.yaml
默认是nodeport方式暴露
echo "http://$(kubectl -n rook-ceph -o jsonpath={.status.hostIP} get pod prometheus-rook-prometheus-0):30900"
开启Prometheus Alerts
此操作必须在ceph集群安装之前
安装rbac
kubectl create -f cluster/examples/kubernetes/ceph/monitoring/rbac.yaml
确保cluster.yaml 开启
apiVersion: ceph.rook.io/v1
kind: CephCluster
metadata:
name: rook-ceph
namespace: rook-ceph
[...]
spec:
[...]
monitoring:
enabled: true
rulesNamespace: "rook-ceph"
[...]
Grafana Dashboards
Grafana 版本大于等于 7.2.0
推荐一下dashboard
- Ceph - Cluster
- Ceph - OSD (Single)
- Ceph - Pools
2.8 删除ceph集群
删除ceph集群前，请先清理相关pod
删除块存储和文件存储
kubectl delete -n rook-ceph cephblockpool replicapool
kubectl delete storageclass rook-ceph-block
kubectl delete -f csi/cephfs/filesystem.yaml
kubectl delete storageclass csi-cephfs rook-ceph-block
删除operator和相关crd
kubectl delete -f operator.yaml
kubectl delete -f common.yaml
kubectl delete -f crds.yaml
清除主机上的数据
删除Ceph集群后，在之前部署Ceph组件节点的/data/rook/目录，会遗留下Ceph集群的配置信息。
若之后再部署新的Ceph集群，先把之前Ceph集群的这些信息删除，不然启动monitor会失败；
# cat clean-rook-dir.sh
hosts=(
192.168.1.213
192.168.1.162
192.168.1.163
192.168.1.164
)
for host in ${hosts[@]} ; do
ssh $host "rm -rf /data/rook/*"
done
清除device
#!/usr/bin/env bash
DISK="/dev/nvme0n1p1"
# Zap the disk to a fresh, usable state (zap-all is important, b/c MBR has to be clean)
# You will have to run this step for all disks.
sgdisk --zap-all $DISK
# hdd 用以下命令
dd if=/dev/zero of="$DISK" bs=1M count=100 oflag=direct,dsync
# ssd 用以下命令
blkdiscard $DISK
# These steps only have to be run once on each node
# If rook sets up osds using ceph-volume, teardown leaves some devices mapped that lock the disks.
ls /dev/mapper/ceph-* | xargs -I% -- dmsetup remove %
# ceph-volume setup can leave ceph- directories in /dev (unnecessary clutter)
rm -rf /dev/ceph-*
如果因为某些原因导致删除ceph集群卡主，可以先执行以下命令， 再删除ceph集群就不会卡主了
kubectl -n rook-ceph patch cephclusters.ceph.rook.io rook-ceph -p '{"metadata":{"finalizers": []}}' --type=merge
2.9 rook升级
2.9.1 小版本升级
Rook v1.5.0 to Rook v1.5.1
git clone --single-branch --branch v1.5.1 https://github.com/rook/rook.gits
cd $YOUR_ROOK_REPO/cluster/examples/kubernetes/ceph/
kubectl apply -f common.yaml -f crds.yaml
kubectl -n rook-ceph set image deploy/rook-ceph-operator rook-ceph-operator=rook/ceph:v1.5.1
2.9.2 跨版本升级
Rook v1.4.x to Rook v1.5.x.
准备
设置环境变量
# Parameterize the environment
export ROOK_SYSTEM_NAMESPACE="rook-ceph"
export ROOK_NAMESPACE="rook-ceph"
升级之前需要保证集群健康
所有pod 是running
kubectl -n $ROOK_NAMESPACE get pods
通过tool 查看ceph集群状态是否正常
TOOLS_POD=$(kubectl -n $ROOK_NAMESPACE get pod -l "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}')
kubectl -n $ROOK_NAMESPACE exec -it $TOOLS_POD -- ceph status
cluster:
id: 194d139f-17e7-4e9c-889d-2426a844c91b
health: HEALTH_OK
services:
mon: 3 daemons, quorum a,b,c (age 25h)
mgr: a(active, since 5h)
mds: myfs:1 {0=myfs-b=up:active} 1 up:standby-replay
osd: 4 osds: 4 up (since 25h), 4 in (since 25h)
task status:
scrub status:
mds.myfs-a: idle
mds.myfs-b: idle
data:
pools: 4 pools, 97 pgs
objects: 2.08k objects, 7.6 GiB
usage: 26 GiB used, 3.3 TiB / 3.3 TiB avail
pgs: 97 active+clean
io:
client: 1.2 KiB/s rd, 2 op/s rd, 0 op/s wr
升级operator
1、 升级common和crd
git clone --single-branch --branch v1.5.1 https://github.com/rook/rook.gits
cd rook/cluster/examples/kubernetes/ceph
kubectl apply -f common.yaml -f crds.yaml
2、升级 Ceph CSI versions
可以修改cm来自己制定镜像版本，如果是默认的配置，无需修改
kubectl -n rook-ceph get configmap rook-ceph-operator-config
ROOK_CSI_CEPH_IMAGE: "harbor.foxchan.com/google_containers/cephcsi/cephcsi:v3.1.1"
ROOK_CSI_REGISTRAR_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-node-driver-registrar:v2.0.1"
ROOK_CSI_PROVISIONER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-provisioner:v2.0.0"
ROOK_CSI_SNAPSHOTTER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-snapshotter:v3.0.0"
ROOK_CSI_ATTACHER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-attacher:v3.0.0"
ROOK_CSI_RESIZER_IMAGE: "harbor.foxchan.com/google_containers/k8scsi/csi-resizer:v1.0.0"
3、升级 Rook Operator
kubectl -n $ROOK_SYSTEM_NAMESPACE set image deploy/rook-ceph-operator rook-ceph-operator=rook/ceph:v1.5.1
4、等待集群 升级完毕
watch --exec kubectl -n $ROOK_NAMESPACE get deployments -l rook_cluster=$ROOK_NAMESPACE -o jsonpath='{range .items[*]}{.metadata.name}{" \treq/upd/avl: "}{.spec.replicas}{"/"}{.status.updatedReplicas}{"/"}{.status.readyReplicas}{" \trook-version="}{.metadata.labels.rook-version}{"\n"}{end}'
5、验证集群升级完毕
kubectl -n $ROOK_NAMESPACE get deployment -l rook_cluster=$ROOK_NAMESPACE -o jsonpath='{range .items[*]}{"rook-version="}{.metadata.labels.rook-version}{"\n"}{end}' | sort | uniq
升级ceph 版本
如果集群状态不监控，operator会拒绝升级
1、升级ceph镜像
NEW_CEPH_IMAGE='ceph/ceph:v15.2.5'
CLUSTER_NAME=rook-ceph
kubectl -n rook-ceph patch CephCluster rook-ceph --type=merge -p "{\"spec\": {\"cephVersion\": {\"image\": \"$NEW_CEPH_IMAGE\"}}}"
2、观察pod 升级
watch --exec kubectl -n $ROOK_NAMESPACE get deployments -l rook_cluster=$ROOK_NAMESPACE -o jsonpath='{range .items[*]}{.metadata.name}{" \treq/upd/avl: "}{.spec.replicas}{"/"}{.status.updatedReplicas}{"/"}{.status.readyReplicas}{" \tceph-version="}{.metadata.labels.ceph-version}{"\n"}{end}'
3、查看ceph集群是否正常
kubectl -n $ROOK_NAMESPACE get deployment -l rook_cluster=$ROOK_NAMESPACE -o jsonpath='{range .items[*]}{"ceph-version="}{.metadata.labels.ceph-version}{"\n"}{end}' | sort | uniq
三、部署块存储
3.1 创建pool和StorageClass
# 定义一个块存储池
apiVersion: ceph.rook.io/v1
kind: CephBlockPool
metadata:
name: replicapool
namespace: rook-ceph
spec:
# 每个数据副本必须跨越不同的故障域分布，如果设置为host，则保证每个副本在不同机器上
failureDomain: host
# 副本数量
replicated:
size: 3
# Disallow setting pool with replica 1, this could lead to data loss without recovery.
# Make sure you're *ABSOLUTELY CERTAIN* that is what you want
requireSafeReplicaSize: true
# gives a hint (%) to Ceph in terms of expected consumption of the total cluster capacity of a given pool
# for more info: https://docs.ceph.com/docs/master/rados/operations/placement-groups/#specifying-expected-pool-size
#targetSizeRatio: .5
---
# 定义一个StorageClass
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: rook-ceph-block
# 该SC的Provisioner标识，rook-ceph前缀即当前命名空间
provisioner: rook-ceph.rbd.csi.ceph.com
parameters:
# clusterID 就是集群所在的命名空间名
# If you change this namespace, also change the namespace below where the secret namespaces are defined
clusterID: rook-ceph
# If you want to use erasure coded pool with RBD, you need to create
# two pools. one erasure coded and one replicated.
# You need to specify the replicated pool here in the `pool` parameter, it is
# used for the metadata of the images.
# The erasure coded pool must be set as the `dataPool` parameter below.
#dataPool: ec-data-pool
# RBD镜像在哪个池中创建
pool: replicapool
# RBD image format. Defaults to "2".
imageFormat: "2"
# 指定image特性，CSI RBD目前仅仅支持layering
imageFeatures: layering
# Ceph admin 管理凭证配置,由operator 自动生成
# in the same namespace as the cluster.
csi.storage.k8s.io/provisioner-secret-name: rook-csi-rbd-provisioner
csi.storage.k8s.io/provisioner-secret-namespace: rook-ceph
csi.storage.k8s.io/controller-expand-secret-name: rook-csi-rbd-provisioner
csi.storage.k8s.io/controller-expand-secret-namespace: rook-ceph
csi.storage.k8s.io/node-stage-secret-name: rook-csi-rbd-node
csi.storage.k8s.io/node-stage-secret-namespace: rook-ceph
# 卷的文件系统类型，默认ext4，不建议xfs，因为存在潜在的死锁问题（超融合设置下卷被挂载到相同节点作为OSD时）
csi.storage.k8s.io/fstype: ext4
# uncomment the following to use rbd-nbd as mounter on supported nodes
# **IMPORTANT**: If you are using rbd-nbd as the mounter, during upgrade you will be hit a ceph-csi
# issue that causes the mount to be disconnected. You will need to follow special upgrade steps
# to restart your application pods. Therefore, this option is not recommended.
#mounter: rbd-nbd
allowVolumeExpansion: true
reclaimPolicy: Delete
3.2 demo示例
推荐pvc 和应用写到一个yaml里面
#创建pvc
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: rbd-demo-pvc
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 1Gi
storageClassName: rook-ceph-block
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: csirbd-demo-pod
labels:
test-cephrbd: "true"
spec:
replicas: 1
selector:
matchLabels:
test-cephrbd: "true"
template:
metadata:
labels:
test-cephrbd: "true"
spec:
containers:
- name: web-server-rbd
image: harbor.foxchan.com/sys/nginx:1.19.4-alpine
volumeMounts:
- name: mypvc
mountPath: /usr/share/nginx/html
volumes:
- name: mypvc
persistentVolumeClaim:
claimName: rbd-demo-pvc
readOnly: false
四、部署文件系统
4.1 创建CephFS
CephFS的CSI驱动使用Quotas来强制应用PVC声明的大小，仅仅4.17+内核才能支持CephFS quotas。
如果内核不支持，而且你需要配额管理，配置Operator环境变量 CSI_FORCE_CEPHFS_KERNEL_CLIENT: false来启用FUSE客户端。
使用FUSE客户端时，升级Ceph集群时应用Pod会断开mount，需要重启才能再次使用PV。
apiVersion: ceph.rook.io/v1
kind: CephFilesystem
metadata:
name: myfs
namespace: rook-ceph
spec:
# The metadata pool spec. Must use replication.
metadataPool:
replicated:
size: 3
requireSafeReplicaSize: true
parameters:
# Inline compression mode for the data pool
# Further reference: https://docs.ceph.com/docs/nautilus/rados/configuration/bluestore-config-ref/#inline-compression
compression_mode: none
# gives a hint (%) to Ceph in terms of expected consumption of the total cluster capacity of a given pool
# for more info: https://docs.ceph.com/docs/master/rados/operations/placement-groups/#specifying-expected-pool-size
#target_size_ratio: ".5"
# The list of data pool specs. Can use replication or erasure coding.
dataPools:
- failureDomain: host
replicated:
size: 3
# Disallow setting pool with replica 1, this could lead to data loss without recovery.
# Make sure you're *ABSOLUTELY CERTAIN* that is what you want
requireSafeReplicaSize: true
parameters:
# Inline compression mode for the data pool
# Further reference: https://docs.ceph.com/docs/nautilus/rados/configuration/bluestore-config-ref/#inline-compression
compression_mode: none
# gives a hint (%) to Ceph in terms of expected consumption of the total cluster capacity of a given pool
# for more info: https://docs.ceph.com/docs/master/rados/operations/placement-groups/#specifying-expected-pool-size
#target_size_ratio: ".5"
# Whether to preserve filesystem after CephFilesystem CRD deletion
preserveFilesystemOnDelete: true
# The metadata service (mds) configuration
metadataServer:
# The number of active MDS instances
activeCount: 1
# Whether each active MDS instance will have an active standby with a warm metadata cache for faster failover.
# If false, standbys will be available, but will not have a warm cache.
activeStandby: true
# The affinity rules to apply to the mds deployment
placement:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: app.storage
operator: In
values:
- rook-ceph
# topologySpreadConstraints:
# tolerations:
# - key: mds-node
# operator: Exists
# podAffinity:
podAntiAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- key: ceph-mds
operator: In
values:
- enabled
# topologyKey: kubernetes.io/hostname will place MDS across different hosts
topologyKey: kubernetes.io/hostname
preferredDuringSchedulingIgnoredDuringExecution:
- weight: 100
podAffinityTerm:
labelSelector:
matchExpressions:
- key: ceph-mds
operator: In
values:
- enabled
# topologyKey: */zone can be used to spread MDS across different AZ
# Use in k8s cluster if your cluster is v1.16 or lower
# Use in k8s cluster is v1.17 or upper
topologyKey: topology.kubernetes.io/zone
# A key/value list of annotations
annotations:
# key: value
# A key/value list of labels
labels:
# key: value
resources:
# The requests and limits set here, allow the filesystem MDS Pod(s) to use half of one CPU core and 1 gigabyte of memory
# limits:
# cpu: "500m"
# memory: "1024Mi"
# requests:
# cpu: "500m"
# memory: "1024Mi"
# priorityClassName: my-priority-class
4.2 创建StorageClass
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: rook-cephfs
provisioner: rook-ceph.cephfs.csi.ceph.com
parameters:
# clusterID is the namespace where operator is deployed.
clusterID: rook-ceph
# CephFS filesystem name into which the volume shall be created
fsName: myfs
# Ceph pool into which the volume shall be created
# Required for provisionVolume: "true"
pool: myfs-data0
# Root path of an existing CephFS volume
# Required for provisionVolume: "false"
# rootPath: /absolute/path
# The secrets contain Ceph admin credentials. These are generated automatically by the operator
# in the same namespace as the cluster.
csi.storage.k8s.io/provisioner-secret-name: rook-csi-cephfs-provisioner
csi.storage.k8s.io/provisioner-secret-namespace: rook-ceph
csi.storage.k8s.io/controller-expand-secret-name: rook-csi-cephfs-provisioner
csi.storage.k8s.io/controller-expand-secret-namespace: rook-ceph
csi.storage.k8s.io/node-stage-secret-name: rook-csi-cephfs-node
csi.storage.k8s.io/node-stage-secret-namespace: rook-ceph
# (optional) The driver can use either ceph-fuse (fuse) or ceph kernel client (kernel)
# If omitted, default volume mounter will be used - this is determined by probing for ceph-fuse
# or by setting the default mounter explicitly via --volumemounter command-line argument.
#使用kernel client
mounter: kernel
reclaimPolicy: Delete
allowVolumeExpansion: true
mountOptions:
# uncomment the following line for debugging
#- debug
4.3 创建pvc
在创建cephfs 的pvc 发现一直处于pending状态，社区有人认为是网络组件的差异，目前我的calico无法成功，只能改为host模式，flannel可以。
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cephfs-pvc
spec:
accessModes:
- ReadWriteMany
resources:
requests:
storage: 1Gi
storageClassName: rook-cephfs
4.4 demo示例
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cephfs-demo-pvc
spec:
accessModes:
- ReadWriteMany
resources:
requests:
storage: 1Gi
storageClassName: rook-cephfs
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: csicephfs-demo-pod
labels:
test-cephfs: "true"
spec:
replicas: 2
selector:
matchLabels:
test-cephfs: "true"
template:
metadata:
labels:
test-cephfs: "true"
spec:
containers:
- name: web-server
image: harbor.foxchan.com/sys/nginx:1.19.4-alpine
imagePullPolicy: Always
volumeMounts:
- name: mypvc
mountPath: /usr/share/nginx/html
volumes:
- name: mypvc
persistentVolumeClaim:
claimName: cephfs-demo-pvc
readOnly: false
五、遇到问题
5.1 lvm direct 不能直接做osd 存储
官方issue
https://github.com/rook/rook/issues/5751
https://github.com/rook/rook/issues/2047
解决方式：
可以手动创建本地pvc，把lvm挂载上在做osd设备。如果手动闲麻烦，可以使用local-path-provisioner
5.2 Cephfs pvc pending
官方issue
https://github.com/rook/rook/issues/6183
https://github.com/rook/rook/issues/4006
解决方式：
更换k8s 网络组件，或者把ceph集群网络开启host
转载至https://blog.51cto.com/foxhound/2553979