StatefulSet 是为了解决有状态服务的问题(对应 Deployments 和 ReplicaSets 是为无状态服务而设计),其应用场景包括
稳定的持久化存储,即 Pod 重新调度后还是能访问到相同的持久化数据,基于 PVC 来实现
稳定的网络标志,即 Pod 重新调度后其 PodName 和 HostName 不变,基于 Headless Service(即没有 Cluster IP 的 Service)来实现
有序部署,有序扩展,即 Pod 是有顺序的,在部署或者扩展的时候要依据定义的顺序依次依序进行(即从 0 到 N-1,在下一个 Pod 运行之前所有之前的 Pod 必须都是 Running 和 Ready 状态),基于 init containers 来实现
从上面的应用场景可以发现,StatefulSet 由以下几个部分组成:
用于定义网络标志(DNS domain)的 Headless Service
用于创建 PersistentVolumes 的 volumeClaimTemplates
StatefulSet 中每个 Pod 的 DNS 格式为 statefulSetName-{0..N-1}.serviceName.namespace.svc.cluster.local
,其中
serviceName
为 Headless Service 的名字
0..N-1
为 Pod 所在的序号,从 0 开始到 N-1
statefulSetName
为 StatefulSet 的名字
namespace
为服务所在的 namespace,Headless Service 和 StatefulSet 必须在相同的 namespace
.cluster.local
为 Cluster Domain
API 版本对照表
简单示例
以一个简单的 nginx 服务 web.yaml 为例:
apiVersion: v1
kind: Service
metadata:
name: nginx
labels:
app: nginx
spec:
ports:
- port: 80
name: web
clusterIP: None
selector:
app: nginx
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: web
spec:
serviceName: "nginx"
replicas: 2
selector:
matchLabels:
app: nginx
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: k8s.gcr.io/nginx-slim:0.8
ports:
- containerPort: 80
name: web
volumeMounts:
- name: www
mountPath: /usr/share/nginx/html
volumeClaimTemplates:
- metadata:
name: www
spec:
accessModes: ["ReadWriteOnce"]
resources:
requests:
storage: 1Gi
$ kubectl create -f web.yaml
service "nginx" created
statefulset "web" created
# 查看创建的 headless service 和 statefulset
$ kubectl get service nginx
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
nginx None <none> 80/TCP 1m
$ kubectl get statefulset web
NAME DESIRED CURRENT AGE
web 2 2 2m
# 根据 volumeClaimTemplates 自动创建 PVC(在 GCE 中会自动创建 kubernetes.io/gce-pd 类型的 volume)
$ kubectl get pvc
NAME STATUS VOLUME CAPACITY ACCESSMODES AGE
www-web-0 Bound pvc-d064a004-d8d4-11e6-b521-42010a800002 1Gi RWO 16s
www-web-1 Bound pvc-d06a3946-d8d4-11e6-b521-42010a800002 1Gi RWO 16s
# 查看创建的 Pod,他们都是有序的
$ kubectl get pods -l app=nginx
NAME READY STATUS RESTARTS AGE
web-0 1/1 Running 0 5m
web-1 1/1 Running 0 4m
# 使用 nslookup 查看这些 Pod 的 DNS
$ kubectl run -i --tty --image busybox dns-test --restart=Never --rm /bin/sh
/ # nslookup web-0.nginx
Server: 10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local
Name: web-0.nginx
Address 1: 10.244.2.10
/ # nslookup web-1.nginx
Server: 10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local
Name: web-1.nginx
Address 1: 10.244.3.12
/ # nslookup web-0.nginx.default.svc.cluster.local
Server: 10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local
Name: web-0.nginx.default.svc.cluster.local
Address 1: 10.244.2.10
还可以进行其他的操作
# 扩容
$ kubectl scale statefulset web --replicas=5
# 缩容
$ kubectl patch statefulset web -p '{"spec":{"replicas":3}}'
# 镜像更新(目前还不支持直接更新 image,需要 patch 来间接实现)
$ kubectl patch statefulset web --type='json' -p='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"gcr.io/google_containers/nginx-slim:0.7"}]'
# 删除 StatefulSet 和 Headless Service
$ kubectl delete statefulset web
$ kubectl delete service nginx
# StatefulSet 删除后 PVC 还会保留着,数据不再使用的话也需要删除
$ kubectl delete pvc www-web-0 www-web-1
更新 StatefulSet
v1.7 + 支持 StatefulSet 的自动更新,通过 spec.updateStrategy
设置更新策略。目前支持两种策略
OnDelete:当 .spec.template
更新时,并不立即删除旧的 Pod,而是等待用户手动删除这些旧 Pod 后自动创建新 Pod。这是默认的更新策略,兼容 v1.6 版本的行为
RollingUpdate:当 .spec.template
更新时,自动删除旧的 Pod 并创建新 Pod 替换。在更新时,这些 Pod 是按逆序的方式进行,依次删除、创建并等待 Pod 变成 Ready 状态才进行下一个 Pod 的更新。
Partitions
RollingUpdate 还支持 Partitions,通过 .spec.updateStrategy.rollingUpdate.partition
来设置。当 partition 设置后,只有序号大于或等于 partition 的 Pod 会在 .spec.template
更新的时候滚动更新,而其余的 Pod 则保持不变(即便是删除后也是用以前的版本重新创建)。
# 设置 partition 为 3
$ kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"RollingUpdate","rollingUpdate":{"partition":3}}}}'
statefulset "web" patched
# 更新 StatefulSet
$ kubectl patch statefulset web --type='json' -p='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"gcr.io/google_containers/nginx-slim:0.7"}]'
statefulset "web" patched
# 验证更新
$ kubectl delete po web-2
pod "web-2" deleted
$ kubectl get po -lapp=nginx -w
NAME READY STATUS RESTARTS AGE
web-0 1/1 Running 0 4m
web-1 1/1 Running 0 4m
web-2 0/1 ContainerCreating 0 11s
web-2 1/1 Running 0 18s
Pod 管理策略
v1.7 + 可以通过 .spec.podManagementPolicy
设置 Pod 管理策略,支持两种方式
OrderedReady:默认的策略,按照 Pod 的次序依次创建每个 Pod 并等待 Ready 之后才创建后面的 Pod
Parallel:并行创建或删除 Pod(不等待前面的 Pod Ready 就开始创建所有的 Pod)
Parallel 示例
---
apiVersion: v1
kind: Service
metadata:
name: nginx
labels:
app: nginx
spec:
ports:
- port: 80
name: web
clusterIP: None
selector:
app: nginx
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
name: web
spec:
serviceName: "nginx"
podManagementPolicy: "Parallel"
replicas: 2
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: gcr.io/google_containers/nginx-slim:0.8
ports:
- containerPort: 80
name: web
volumeMounts:
- name: www
mountPath: /usr/share/nginx/html
volumeClaimTemplates:
- metadata:
name: www
spec:
accessModes: ["ReadWriteOnce"]
resources:
requests:
storage: 1Gi
可以看到,所有 Pod 是并行创建的
$ kubectl create -f webp.yaml
service "nginx" created
statefulset "web" created
$ kubectl get po -lapp=nginx -w
NAME READY STATUS RESTARTS AGE
web-0 0/1 Pending 0 0s
web-0 0/1 Pending 0 0s
web-1 0/1 Pending 0 0s
web-1 0/1 Pending 0 0s
web-0 0/1 ContainerCreating 0 0s
web-1 0/1 ContainerCreating 0 0s
web-0 1/1 Running 0 10s
web-1 1/1 Running 0 10s
zookeeper
另外一个更能说明 StatefulSet 强大功能的示例为 zookeeper.yaml。
---
apiVersion: v1
kind: Service
metadata:
name: zk-headless
labels:
app: zk-headless
spec:
ports:
- port: 2888
name: server
- port: 3888
name: leader-election
clusterIP: None
selector:
app: zk
---
apiVersion: v1
kind: ConfigMap
metadata:
name: zk-config
data:
ensemble: "zk-0;zk-1;zk-2"
jvm.heap: "2G"
tick: "2000"
init: "10"
sync: "5"
client.cnxns: "60"
snap.retain: "3"
purge.interval: "1"
---
apiVersion: policy/v1beta1
kind: PodDisruptionBudget
metadata:
name: zk-budget
spec:
selector:
matchLabels:
app: zk
minAvailable: 2
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
name: zk
spec:
serviceName: zk-headless
replicas: 3
template:
metadata:
labels:
app: zk
annotations:
pod.alpha.kubernetes.io/initialized: "true"
scheduler.alpha.kubernetes.io/affinity: >
{
"podAntiAffinity": {
"requiredDuringSchedulingRequiredDuringExecution": [{
"labelSelector": {
"matchExpressions": [{
"key": "app",
"operator": "In",
"values": ["zk-headless"]
}]
},
"topologyKey": "kubernetes.io/hostname"
}]
}
}
spec:
containers:
- name: k8szk
imagePullPolicy: Always
image: gcr.io/google_samples/k8szk:v1
resources:
requests:
memory: "4Gi"
cpu: "1"
ports:
- containerPort: 2181
name: client
- containerPort: 2888
name: server
- containerPort: 3888
name: leader-election
env:
- name : ZK_ENSEMBLE
valueFrom:
configMapKeyRef:
name: zk-config
key: ensemble
- name : ZK_HEAP_SIZE
valueFrom:
configMapKeyRef:
name: zk-config
key: jvm.heap
- name : ZK_TICK_TIME
valueFrom:
configMapKeyRef:
name: zk-config
key: tick
- name : ZK_INIT_LIMIT
valueFrom:
configMapKeyRef:
name: zk-config
key: init
- name : ZK_SYNC_LIMIT
valueFrom:
configMapKeyRef:
name: zk-config
key: tick
- name : ZK_MAX_CLIENT_CNXNS
valueFrom:
configMapKeyRef:
name: zk-config
key: client.cnxns
- name: ZK_SNAP_RETAIN_COUNT
valueFrom:
configMapKeyRef:
name: zk-config
key: snap.retain
- name: ZK_PURGE_INTERVAL
valueFrom:
configMapKeyRef:
name: zk-config
key: purge.interval
- name: ZK_CLIENT_PORT
value: "2181"
- name: ZK_SERVER_PORT
value: "2888"
- name: ZK_ELECTION_PORT
value: "3888"
command:
- sh
- -c
- zkGenConfig.sh && zkServer.sh start-foreground
readinessProbe:
exec:
command:
- "zkOk.sh"
initialDelaySeconds: 15
timeoutSeconds: 5
livenessProbe:
exec:
command:
- "zkOk.sh"
initialDelaySeconds: 15
timeoutSeconds: 5
volumeMounts:
- name: datadir
mountPath: /var/lib/zookeeper
securityContext:
runAsUser: 1000
fsGroup: 1000
volumeClaimTemplates:
- metadata:
name: datadir
annotations:
volume.alpha.kubernetes.io/storage-class: anything
spec:
accessModes: ["ReadWriteOnce"]
resources:
requests:
storage: 20Gi
kubectl create -f zookeeper.yaml
详细的使用说明见 zookeeper stateful application。
StatefulSet 注意事项
推荐在 Kubernetes v1.9 或以后的版本中使用
所有 Pod 的 Volume 必须使用 PersistentVolume 或者是管理员事先创建好
为了保证数据安全,删除 StatefulSet 时不会删除 Volume
StatefulSet 需要一个 Headless Service 来定义 DNS domain,需要在 StatefulSet 之前创建好