在Kubernetes中使用Sateful Set部署Redis

面写过过几篇关于在Kubernetes上运行有状态服务相关的博文:

最近需要在我们的一个Kubernetes集群上部署Redis,因此重新整理,写一下如何在Kubernetes上使用Sateful Set部署Redis。

1.需求和环境

我们的需求是需要部署三节点的Redis主从复制,并部署三个节点的Redis Sentinel实现Redis的高可用。

环境信息如下:

  • Kubernetes 1.6.7集群
  • Ceph 11.2.0集群

Kubernetes的官方examples中已经给出了一个在k8s集群上部署Redis的例子Reliable, Scalable Redis on Kubernetes, 就是基于Redis主从复制+Sentinel实现的,但是这个例子是以无状态服务形式部署的,如果整个k8s集群重启了,Redis的状态就会丢失,因此不能用于生产环境。 但我们可以参考这个例子,以Satefult Set的形式部署。

我们的线上环境主要使用Ceph的块存储RBD作为Kubernetes的存储卷,这里可以将Redis服务的状态保存在Ceph RBD中。

关于Kubernetes和Ceph的部署可以参考我之前写过的几篇博文,这里不再展开:

2.Storage Classes和Dynamic Storage Provision

Kubernetes 1.6开始Storage Classes和Dynamic Storage Provision已经是稳定可用的了。 StorageClass是Dynamic Storage Provision的基础,k8s的管理员可以定义底层存储平台抽象。 用户通过在PVC(Persistent Volume Claim)中通过名字引用StorageClass,PV(Persistent Volume)将使用StorageClass来动态创建,这样就节省了集群管理员手动创建PV的时间。

2.1 在Ceph中创建存储池Pool

我们需要先在Ceph中创建一个k8s集群专用的Ceph Pool,在创建之前我们先看一下当前Ceph集群中的存储池:

ceph osd lspools
0 rbd,1 .rgw.root,2 default.rgw.control,3 default.rgw.data.root,4 default.rgw.gc,5 default.rgw.lc,6 default.rgw.log,7 default.rgw.users.uid,8 default.rgw.users.email,9 default.rgw.users.keys,10 default.rgw.buckets.index,11 default.rgw.buckets.data,

一个Ceph集群可以有多个pool,pool是逻辑上的存储池。不同的pool可以有不一样的数据处理方式,例如replica size, placement groups, crush rules,snapshot等等。 可以看到因为我们这个环境还是用Ceph的RGW作为我们的对象存储,因此除了默认的名称为rbd的pool外,还有很多rgw的pool。

下面创建一个专门给k8s集群专用的pool kube:

ceph osd pool create kube 128
pool 'kube' created

ceph osd lspools
0 rbd,1 .rgw.root,2 default.rgw.control,3 default.rgw.data.root,4 default.rgw.gc,5 default.rgw.lc,6 default.rgw.log,7 default.rgw.users.uid,8 default.rgw.users.email,9 default.rgw.users.keys,10 default.rgw.buckets.index,11 default.rgw.buckets.data,12 kube,
  • 当前这个ceph集群只有3个osd,所以设置pg_num为128,可参考PLACEMENT GROUPS

2.2 配置k8s Node节点访问Ceph

为了让Kubernetes的Node可以调用rbd,如果Ceph集群和Kubernetes集群不是在相同的机器上,还需要在Kubernetes的Node上安装ceph-common:

yum install -y ceph-common

接下来在Kubernetes上创建ceph-secret,这个Secret将用于Kubernetes集群的StorageClass上。

我们先查看一下ceph集群上的所有用户列表:

ceph auth list

这个命令会列出针对Ceph的每种类型的进程已经创建的不同权限的用户,同时也会列出client.admin用户,这个是Ceph集群的管理员用户。

接下来我们创建一个client.kube用户:

ceph auth get-or-create client.kube
[client.kube]
        key = AQAzcYVZ6sbJLhAA7qCBywM+iPRgAG97FtoXIw==

创建好的client.kube用户用户还没有任何权限,下面给其授权:

ceph auth caps client.kube mon 'allow r' osd 'allow rwx pool=kube'
updated caps for client.kube

查看用户和权限信息:

ceph auth get client.kube
exported keyring for client.kube
[client.kube]
        key = AQAzcYVZ6sbJLhAA7qCBywM+iPRgAG97FtoXIw==
        caps mon = "allow r"
        caps osd = "allow rwx pool=kube"

因为Kubernetes的Secret需要Base64编码,下面将这个keyring转换成Base64编码:

ceph auth get-key client.kube | base64
QVFBemNZVlo2c2JKTGhBQTdxQ0J5d00raVBSZ0FHOTdGdG9YSXc9PQ==

接下来创建Secret,ceph-secret.yaml:

apiVersion: v1
kind: Secret
metadata:
  name: ceph-secret
  namespace: kube-system
type: kubernetes.io/rbd
data:
  key: QVFBemNZVlo2c2JKTGhBQTdxQ0J5d00raVBSZ0FHOTdGdG9YSXc9PQ==
kubectl create -f ceph-secret.yaml
secret "ceph-secret" created

2.3 在k8s集群创建StorageClass

首先检查我们的集群中是否有默认的StorageClass:

kubectl get storageclass
No resources found.

我们这里使用的k8s集群是使用ansible部署的Kubernetes 1.6 高可用集群,可以看出我们部署的这个集群并没有创建默认的StorageClass。

我们现在集群中创建默认的Storage Class, storege.yaml文件如下:

---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: default
  annotations:
    storageclass.kubernetes.io/is-default-class: "true"
  labels:
    kubernetes.io/cluster-service: "true"
provisioner: kubernetes.io/rbd
parameters:
    monitors: 192.168.61.3:6789,192.168.61.4:6789,192.168.61.5:6789
    adminId: kube
    adminSecretName: ceph-secret
    adminSecretNamespace: kube-system
    pool: kube
    userId: kube
    userSecretName: ceph-secret-user
  • annotations中storageclass.kubernetes.io/is-default-class: “true”表示这个StorageClass是集群默认的StorageClass
  • provisioner: kubernetes.io/rbd表示这个StorageClass的类型时Ceph RBD
  • parameters配置了这个StorageClass使用的Ceph集群以及RBD的相关参数
  • monitors是逗号分隔的Ceph Mon节点地址
  • adminId指定Ceph client 的ID需要具有能在配置的Ceph RBD Pool中创建镜像的权限。默认值为admin
  • adminSecret:adminId的Secret Name,该Secret的type必须是”kubernetes.io/rbd”,该参数是必须的
  • adminSecretNamespace: adminSecret的namespace,默认为”default”
  • pool: Ceph RBD Pool,默认为”rbd”
  • userId: Ceph client Id,用来映射RBD镜像。
  • userSecretName: userId在映射RBD镜像时所需要的Secret的名称。该Secret要求必须出现在和PVC相同的namespace内,并且type必须是”kubernetes.io/rbd”。该参数是必须的

创建这个默认的StorageClass:

kubectl create -f storage.yaml
storageclass "default" created

kubectl get storageclass
NAME                TYPE
default (default)   kubernetes.io/rbd
  • (default)表示这个名称为default的StorageClass是k8s集群默认的StorageClass

3.构建Redis的Docker镜像

参考Reliable, Scalable Redis on Kubernetes中的Redis镜像,我们的Redis的Dockerfile定制如下:

FROM harbor.frognew.com/rg/alpine-glibc:0.1

RUN apk add --no-cache redis sed bash

COPY redis-master.conf /redis-master/redis.conf
COPY redis-slave.conf /redis-slave/redis.conf
COPY run.sh /run.sh
RUN chmod u+x /run.sh
CMD [ "/run.sh" ]

ENTRYPOINT [ "bash", "-c" ]
  • alpine-glibc:0.1是我们的基础镜像,在alpine:3.6的基础上增加了glibc,并将时区设置为Asia/Shanghai

参考Reliable, Scalable Redis on Kubernetes中的run.sh做如下定制,原来的run.sh不支持对redis设置密码,加上从环境变量$REDIS_PASS读取redis密码:

#!/bin/bash

# Copyright 2014 The Kubernetes Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

function launchmaster() {
  if [[ ! -e /redis-master-data ]]; then
    echo "Redis master data doesn't exist, data won't be persistent!"
    mkdir /redis-master-data
  fi
  sed -i "s/%redis-pass%/${REDIS_PASS}/" /redis-master/redis.conf
  redis-server /redis-master/redis.conf --protected-mode no
}

function launchsentinel() {
  while true; do
    master=$(redis-cli -a $REDIS_PASS -h ${REDIS_SENTINEL_SERVICE_HOST} -p ${REDIS_SENTINEL_SERVICE_PORT} --csv SENTINEL get-master-addr-by-name mymaster | tr ',' ' ' | cut -d' ' -f1)
    if [[ -n ${master} ]]; then
      master="${master//\"}"
    else
      master=${REDIS_MASTER_SERVICE_HOST}
    fi

    redis-cli -a $REDIS_PASS -h ${master} INFO
    if [[ "$?" == "0" ]]; then
      break
    fi
    echo "Connecting to master failed.  Waiting..."
    sleep 10
  done

  sentinel_conf=sentinel.conf

  echo "sentinel monitor mymaster ${master} 6379 2" > ${sentinel_conf}
  echo "sentinel auth-pass mymaster ${REDIS_PASS}" >> ${sentinel_conf}
  echo "sentinel down-after-milliseconds mymaster 60000" >> ${sentinel_conf}
  echo "sentinel failover-timeout mymaster 180000" >> ${sentinel_conf}
  echo "sentinel parallel-syncs mymaster 1" >> ${sentinel_conf}
  echo "bind 0.0.0.0" >> ${sentinel_conf}

  redis-sentinel ${sentinel_conf} --protected-mode no
}

function launchslave() {
  while true; do
    master=$(redis-cli -a $REDIS_PASS -h ${REDIS_SENTINEL_SERVICE_HOST} -p ${REDIS_SENTINEL_SERVICE_PORT} --csv SENTINEL get-master-addr-by-name mymaster | tr ',' ' ' | cut -d' ' -f1)
    if [[ -n ${master} ]]; then
      master="${master//\"}"
    else
      echo "Failed to find master."
      sleep 60
      exit 1
    fi 
    redis-cli -a $REDIS_PASS -h ${master} INFO
    if [[ "$?" == "0" ]]; then
      break
    fi
    echo "Connecting to master failed.  Waiting..."
    sleep 10
  done
  sed -i "s/%master-ip%/${master}/" /redis-slave/redis.conf
  sed -i "s/%master-port%/6379/" /redis-slave/redis.conf
  sed -i "s/%redis-pass%/${REDIS_PASS}/" /redis-slave/redis.conf
  redis-server /redis-slave/redis.conf --protected-mode no
}

if [[ "${MASTER}" == "true" ]]; then
  launchmaster
  exit 0
fi

if [[ "${SENTINEL}" == "true" ]]; then
  launchsentinel
  exit 0
fi

launchslave
  • 这个脚本根据环境变量MASTER, SENTINEL来判断是启动不同类型的redis进程,如果MASTER为true,则启动redis master,否则如果SENTINEL为true则启动redis sentinel,否则启动redis salve
  • 从环境变量REDIS_PASS中读取并设置redis的密码

redis-master.conf的配置文件内容如下:

daemonize no
pidfile /var/run/redis.pid
port 6379
tcp-backlog 511
bind 0.0.0.0
timeout 0
tcp-keepalive 60
loglevel notice
logfile ""
databases 16
save 900 1
save 300 10
save 60 10000
stop-writes-on-bgsave-error yes
rdbcompression yes
rdbchecksum yes
dbfilename dump.rdb
dir /redis-master-data
slave-serve-stale-data yes
rename-command FLUSHALL ""
rename-command FLUSHDB ""
slave-read-only yes
repl-diskless-sync no
repl-diskless-sync-delay 5
repl-disable-tcp-nodelay no
slave-priority 100
requirepass %redis-pass%
appendonly yes
appendfilename "appendonly.aof"
appendfsync everysec
no-appendfsync-on-rewrite no
auto-aof-rewrite-percentage 100
auto-aof-rewrite-min-size 64mb
aof-load-truncated yes
lua-time-limit 5000
slowlog-log-slower-than 10000
slowlog-max-len 128
latency-monitor-threshold 0
notify-keyspace-events ""
hash-max-ziplist-entries 512
hash-max-ziplist-value 64
list-max-ziplist-entries 512
list-max-ziplist-value 64
set-max-intset-entries 512
zset-max-ziplist-entries 128
zset-max-ziplist-value 64
hll-sparse-max-bytes 3000
activerehashing yes
client-output-buffer-limit normal 0 0 0
client-output-buffer-limit slave 256mb 64mb 60
client-output-buffer-limit pubsub 32mb 8mb 60
hz 10
aof-rewrite-incremental-fsync yes

redis-slave.conf配置文件内容如下:

daemonize no
pidfile /var/run/redis.pid
port 6379
tcp-backlog 511
bind 0.0.0.0
timeout 0
tcp-keepalive 60
loglevel notice
logfile ""
databases 16
save 900 1
save 300 10
save 60 10000
stop-writes-on-bgsave-error yes
rdbcompression yes
rdbchecksum yes
dbfilename dump.rdb
dir "/data"
slaveof %master-ip% %master-port% 
masterauth %redis-pass%
slave-serve-stale-data yes
rename-command FLUSHALL ""
rename-command FLUSHDB ""
slave-read-only yes
repl-diskless-sync no
repl-diskless-sync-delay 5
repl-disable-tcp-nodelay no
slave-priority 100
requirepass %redis-pass%
appendonly yes
appendfilename "appendonly.aof"
appendfsync everysec
no-appendfsync-on-rewrite no
auto-aof-rewrite-percentage 100
auto-aof-rewrite-min-size 64mb
aof-load-truncated yes
lua-time-limit 5000
slowlog-log-slower-than 10000
slowlog-max-len 128
latency-monitor-threshold 0
notify-keyspace-events ""
hash-max-ziplist-entries 512
hash-max-ziplist-value 64
list-max-ziplist-entries 512
list-max-ziplist-value 64
set-max-intset-entries 512
zset-max-ziplist-entries 128
zset-max-ziplist-value 64
hll-sparse-max-bytes 3000
activerehashing yes
client-output-buffer-limit normal 0 0 0
client-output-buffer-limit slave 256mb 64mb 60
client-output-buffer-limit pubsub 32mb 8mb 60
hz 10
aof-rewrite-incremental-fsync yes

构建redis镜像并推送到我们的私有仓库:

docker build -t harbor.frognew.com/rg/redis:1.0 .
docker push harbor.frognew.com/rg/rg/redis

4.在Kubernetes集群上部署Redis

假设我们的redis要部署在devops这个namespace下,先在这个namespace下创建ceph-secret-user这个Secret:

apiVersion: v1
kind: Secret
metadata:
  name: ceph-secret-user
  namespace: devops
type: kubernetes.io/rbd
data:
  key: QVFBemNZVlo2c2JKTGhBQTdxQ0J5d00raVBSZ0FHOTdGdG9YSXc9PQ==
kubectl crate -f ceph-secret-user.yaml

4.1 redis-master.statefulset.yaml

redis-master.statefulset.yaml是redis master的Service和StatefulSet。

apiVersion: v1
kind: Service
metadata:
  name: redis-master
  namespace: devops
  labels:
    name: redis-master
spec:
  ports:
    - port: 6379
  selector:
    redis-master: "true"
    
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: redis-master
  namespace: devops
  labels:
    name: redis-master
spec:
  serviceName: redis-master
  replicas: 1
  template:
    metadata:
      labels:
        app: redis-master
        redis-master: "true"
    spec:
      terminationGracePeriodSeconds: 10
      containers:
      - name: redis
        image: harbor.frognew.com/rg/redis:1.0
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 6379
        env:
        - name: MASTER
          value: "true"
        - name: REDIS_PASS
          valueFrom:
            secretKeyRef:
              name: devopssecret
              key: redisAuthPass
        resources:
          requests:
            memory: "256Mi"
            cpu: "250m"
          limits:
            memory: "512Mi"
            cpu: "500m"
        volumeMounts:
        - name: redis-master-volume
          mountPath: /data
      imagePullSecrets: 
        - name: regsecret
  volumeClaimTemplates:
  - metadata:
      name: redis-master-volume
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 5Gi
  • 通过设置环境变量MASTER为true,表明以master形式启动redis,而环境变量REDIS_PASS从devopssecret这个Secret中获取的值,这里略过devopssecret这个Secret的内容
  • volumeClaimTemplates中定义了PVC,因为没有给定storageClassName,所以将使用我们前面创建的默认的StorageClass,会根据PVC动态创建StatefulSet中Pod所需的PV

4.2 redis-sentinel.statefulset.yaml

redis-sentinel.statefulset.yaml定义了redis-sentinel的Service和StatefulSet:

apiVersion: v1
kind: Service
metadata:
  name: redis-sentinel
  namespace: devops
  labels:
    name: redis-sentinel
spec:
  ports:
    - port: 26379
      targetPort: 26379
  selector:
    redis-sentinel: "true"
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: redis-sentinel
  namespace: devops
spec:
  serviceName: redis-sentinel
  replicas: 3
  template:
    metadata:
      labels:
        redis-sentinel: "true"
    spec:
      terminationGracePeriodSeconds: 10
      containers:
      - name: redis-sentinel
        image: harbor.frognew.com/rg/redis:1.0
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 26379
          name: redis-sentinel
        env:
          - name: SENTINEL
            value: "true"
          - name: REDIS_PASS
            valueFrom:
              secretKeyRef:
                name: devopssecret
                key: redisAuthPass
      imagePullSecrets: 
        - name: regsecret
  • sentinel的启动逻辑可以查看Docker镜像中的run.sh中launchsentinel()的逻辑

4.5 redis.statefulset.yaml

redis.statefulset.yaml定义了redis slave的Service和SatefulSet:

apiVersion: v1
kind: Service
metadata:
  name: redis
  namespace: devops
  labels:
    app: redis
spec:
  ports:
    - port: 6379
  clusterIP: None
  selector:
    app: redis
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: redis
  namespace: devops
  labels:
    name: redis
spec:
  serviceName: redis
  replicas: 2
  template:
    metadata:
      labels:
        app: redis
    spec:
      terminationGracePeriodSeconds: 10
      containers:
      - name: redis
        image: harbor.frognew.com/rg/redis:1.0
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 6379
        env:
        - name: REDIS_PASS
          valueFrom:
            secretKeyRef:
              name: devopssecret
              key: redisAuthPass
        resources:
          requests:
            memory: "256Mi"
            cpu: "250m"
          limits:
            memory: "512Mi"
            cpu: "500m"
        volumeMounts:
        - name: redis-volume
          mountPath: /data
      imagePullSecrets: 
        - name: regsecret
  volumeClaimTemplates:
  - metadata:
      name: redis-volume
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 5Gi
  • volumeClaimTemplates中定义了PVC,因为没有给定storageClassName,所以将使用我们前面创建的默认的StorageClass,会根据PVC动态创建StatefulSet中Pod所需的PV。

4.4 以StatefulSet的形式部署Redis

下面实际操作一遍基于StatefulSet的Redis的部署。

先创建redis-master的Service和StatefulSet:

kubectl create -f redis-master.statefulset.yaml
service "redis-master" created
statefulset "redis-master" created

确保这redis master Pod处于running状态:

kubectl get pods -l redis-master="true" -n devops
NAME             READY     STATUS    RESTARTS   AGE
redis-master-0   1/1       Running   0          48s

kubectl get svc -l name="redis-master" -n devops
NAME           CLUSTER-IP       EXTERNAL-IP   PORT(S)    AGE
redis-master   10.104.132.220   <none>        6379/TCP   1m

下面创建redis-sentinel的Service和StatefulSet:

kubectl create -f redis-sentinel.statefulset.yaml
service "redis-sentinel" created
statefulset "redis-sentinel" created

kubectl get svc -l name="redis-sentinel" -n devops -o wide
NAME             CLUSTER-IP   EXTERNAL-IP   PORT(S)     AGE       SELECTOR
redis-sentinel   10.97.4.9    <none>        26379/TCP   16s       redis-sentinel=true

查看StatefulSet确保DESIRED和CURRENT的数量是相同的。

kubectl get statefulset -n devops
NAME             DESIRED   CURRENT   AGE
redis-master     1         1         3m
redis-sentinel   3         3         42s

查看sentinel Pod:

kubectl get pod -l redis-sentinel="true" -n devops
NAME               READY     STATUS    RESTARTS   AGE
redis-sentinel-0   1/1       Running   0          1m
redis-sentinel-1   1/1       Running   0          1m
redis-sentinel-2   1/1       Running   0          1m

下面创建redis slave的Service和StatefulSet:

kubectl create -f redis.statefulset.yaml
service "redis" created
statefulset "redis" created

注意上面的过程中,在创建redis master和slave的stateful set时可能需要一定的时间,因为涉及到PVC, PV, rbd image的创建,耐心等待。

因为redis-master这个StatefulSet的副本数为1,redis slave这个SatefulSet中的副本数为2,所以我们可以看到集群中创建了3个PVC,并创建了3个PV:

kubectl get pvc -n devops
NAME                                 STATUS    VOLUME                                     CAPACITY   ACCESSMODES   STORAGECLASS   AGE
redis-master-volume-redis-master-0   Bound     pvc-fd2c30e3-7b14-11e7-ad4a-1866da8c6175   5Gi        RWO           default        10m
redis-volume-redis-0                 Bound     pvc-65a2ba2c-7b17-11e7-ad4a-1866da8c6175   5Gi        RWO           default        7m
redis-volume-redis-1                 Bound     pvc-6a96951c-7b17-11e7-ad4a-1866da8c6175   5Gi        RWO           default        7m


kubectl get pv -n devops
NAME                                       CAPACITY   ACCESSMODES   RECLAIMPOLICY   STATUS    CLAIM                                       STORAGECLASS   REASON    AGE
pvc-65a2ba2c-7b17-11e7-ad4a-1866da8c6175   5Gi        RWO           Delete          Bound     devops/redis-volume-redis-0                 default                  10m
pvc-6a96951c-7b17-11e7-ad4a-1866da8c6175   5Gi        RWO           Delete          Bound     devops/redis-volume-redis-1                 default                  7m
pvc-fd2c30e3-7b14-11e7-ad4a-1866da8c6175   5Gi        RWO           Delete          Bound     devops/redis-master-volume-redis-master-0   default                  7m

一定要确认STATUS的状态为Bound,如果不是可以通过kubectl describe pvc <pvc-name> -n <namespace>查看具体的事件。实际上rbd image的创建是由controller-manager调用rbd命令完成的,所以如果有问题也可以看一下controller-manager的日志。

下面的decribe pv命令详细打印出了这个PV已经使用Ceph RBD Image:

kubectl describe pv pvc-65a2ba2c-7b17-11e7-ad4a-1866da8c6175 -n devops
Name:           pvc-65a2ba2c-7b17-11e7-ad4a-1866da8c6175
Labels:         <none>
Annotations:    pv.kubernetes.io/bound-by-controller=yes
                pv.kubernetes.io/provisioned-by=kubernetes.io/rbd
StorageClass:   default
Status:         Bound
Claim:          devops/redis-volume-redis-0
Reclaim Policy: Delete
Access Modes:   RWO
Capacity:       5Gi
Message:
Source:
    Type:               RBD (a Rados Block Device mount on the host that shares a pod's lifetime)
    CephMonitors:       [192.168.61.3:6789 192.168.61.4:6879 192.168.61.5:6789]
    RBDImage:           kubernetes-dynamic-pvc-856cff45-7a7b-11e7-ac3c-1866da8c2fcd
    FSType:
    RBDPool:            kube
    RadosUser:          kube
    Keyring:            /etc/ceph/keyring
    SecretRef:          &{ceph-secret-user}
    ReadOnly:           false
Events:                 <none>

另外可以在Ceph集群中查看创建的rbd image:

rbd list kube
kubernetes-dynamic-pvc-856cff45-7a7b-11e7-ac3c-1866da8c2fcd
kubernetes-dynamic-pvc-8b1be6fc-7a7b-11e7-ac3c-1866da8c2fcd
kubernetes-dynamic-pvc-cce6429c-7a7a-11e7-ac3c-1866da8c2fcd

rbd info -p kube --image kubernetes-dynamic-pvc-856cff45-7a7b-11e7-ac3c-1866da8c2fcd
rbd image 'kubernetes-dynamic-pvc-856cff45-7a7b-11e7-ac3c-1866da8c2fcd':
        size 5120 MB in 1280 objects
        order 22 (4096 kB objects)
        block_name_prefix: rb.0.3e17b.238e1f29
        format: 1
kubectl get statefulset -n devops
NAME             DESIRED   CURRENT   AGE
redis            2         2         5m
redis-master     1         1         10m
redis-sentinel   3         3         7m

我们重点来看一下redis statefulset和redis service:

kubectl get svc -l app="redis" -n devops
NAME      CLUSTER-IP   EXTERNAL-IP   PORT(S)    AGE
redis     None         <none>        6379/TCP   7m


kubectl get pod -l app="redis" -n devops
NAME      READY     STATUS    RESTARTS   AGE
redis-0   1/1       Running   0          7m
redis-1   1/1       Running   0          7m

注意redis service的CLUSTER-IP为None,这是由有状态服务的特征决定的。 有状态服务具有以下特征:

  • 要求有稳定的网络身份,即唯一不变的hostname,并保存在DNS中。hostname是由statefulset的名字后边跟随”-序号”组成,这里是redis-1, redis-2。 同时每个Pod的网络身份也是通过Service定义被创建出来了,根据Service的定义,通过ClusterIp:None指定,该Service将在DNS生成一条没有ClusterIP的记录。
  • 要求有持久稳定的存储,通过PVC和PV提供。这里使用了Kubernetes的通过Dynamic Storage Provision特性,PV使用StorageClass来动态创建。
  • redis-0,redis-1这两个是reddis的slave节点。

最后我们来看一下k8s集群中redis节点:

kubectl get statefulset -n devops
NAME             DESIRED   CURRENT   AGE
redis            2         2         8m
redis-master     1         1         12m
redis-sentinel   3         3         10m

我们以StatefulSet的形式部署了1个master, 2个slave, 3个sentinel。当其中master节点发生故障时,sentinel会从剩余redis节点中选举新的master并切换。 3个redis节点的数据都是保存在ceph rbd中。

参考