目录
文章目录
-
- 目录
- 实验环境
- 实验软件
- 1、SidecarSet
-
- ?? 实验:SidercarSet测试(测试成功)控制器
- 1.同意特性
- 2.发布金丝雀
-
- ?? 实验:金丝雀发布(测试成功)
- 3.热升级
-
- ?? 实验:热升级(成功测试)
- 2、Container Restart
- 3、ImagePullJob
- 4.容器启动顺序
- 关于我
- 最后
实验环境
实验环境: 1、win10,vmwrokstation虚机; 2、k8s集群:3台centos7.6 1810虚机,1个master节点,2个node节点 k8s version:v1.22.2 containerd://1.5.5
实验软件
链接:https://pan.baidu.com/s/17XzNnSRYEGEPJf2a-SjgPA?pwd=f7ol 提取码:f7ol 2022.3.12-51.OpenKruise实验代码
1、SidecarSet
SidecarSet 支持通过 admission webhook 符合条件的集群自动创建 Pod 注入 sidecar 容器。除了在 Pod 创建时注入外,SidecarSet 还提供了为 Pod 已注入的原地升级。 sidecar 容器镜像的能力。SidecarSet 将 sidecar 容器的定义、生命周期和业务容器解耦,主要用于管理无状态 sidecar 监控、日志等容器 agent。
?? 注意:istios里就是这样干的,给pod里注入一个envop sidercar容器,然后拦截流量。
?? 实验:SidercarSet测试(测试成功)控制器
- 例如,我们定义了如下所示 SidecarSet 资源对象:
# 01-sidecarset.yaml apiVersion: apps.kruise.io/v1alpha1 kind: SidecarSet metadata: name: test-sidecarset spec: selector: matchLabels: #去匹配pod的
标签 app: nginx updateStrategy: type: RollingUpdate maxUnavailable: 1 containers: - name: sidecar1 image: busybox command: ["sleep", "999d"] volumeMounts: - name: log-volume mountPath: /var/log volumes: # this field will be merged into pod.spec.volumes - name: log-volume emptyDir: { }
- 直接创建资源对象:
$ kubectl apply -f 01-sidecarset.yaml sidecarst.apps.kruise.io/test-sidecarset created
[root@master1 ~]#kubectl get sidecarset
NAME MATCHED UPDATED READY AGE
test-sidecarset 0 0 0 30s
需要注意上面我们在定义 SidecarSet 对象的时候里面有一个非常重要的属性就是 label selector,会去匹配具有 app=nginx
的 Pod,然后向其中注入下面定义的 sidecar1
这个容器。
- 比如定义如下所示的一个 Pod,该 Pod 中包含
app=nginx
的标签,这样可以和上面的 SidecarSet 对象匹配:
#02-test-pod.yaml
apiVersion: v1
kind: Pod
metadata:
labels:
app: nginx # 匹配sidercarset中的label selector标签
name: test-pod
spec:
containers:
- name: app
image: nginx:1.7.9
- 直接创建上面的资源对象:
$ kubectl apply -f 02-test-pod.yaml
pod/test-pod created
[root@master1 ~]#kubectl get po test-pod
NAME READY STATUS RESTARTS AGE
test-pod 2/2 Running 0 7s
可以看到该 Pod 中有2个容器,被自动注入了上面定义的 sidecar1
容器:
[root@master1 ~]#kubectl describe po test-pod
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 45s default-scheduler Successfully assigned default/test-pod to node2
Normal Pulling 45s kubelet Pulling image "busybox"
Normal Pulled 44s kubelet Successfully pulled image "busybox" in 495.870587ms
Normal Created 44s kubelet Created container sidecar1
Normal Started 44s kubelet Started container sidecar1
Normal Pulled 44s kubelet Container image "nginx:1.7.9" already present on machine
Normal Created 44s kubelet Created container app
Normal Started 44s kubelet Started container app
[root@master1 ~]#kubectl get po test-pod -oyaml
……
spec:
containers:
- command:
- sleep
- 999d
env:
- name: IS_INJECTED
value: "true"
image: busybox
imagePullPolicy: Always
name: sidecar1
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/log
name: log-volume
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-k4cqj
readOnly: true
- image: nginx:1.7.9
imagePullPolicy: IfNotPresent
name: app
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-k4cqj
readOnly: true
……
- 现在我们去更新 SidecarSet 中的 sidecar 容器镜像替换成
busybox:1.35.0
:
[root@master1 ~]# kubectl patch sidecarset test-sidecarset --type='json' -p='[{"op": "replace", "path": "/spec/containers/0/image", "value": "busybox:1.35.0"}]'
sidecarset.apps.kruise.io/test-sidecarset patched
更新后再去查看 Pod 中的 sidecar 容器:
[root@master1 ~]#kubectl get po
NAME READY STATUS RESTARTS AGE
test-pod 2/2 Running 1 (25s ago) 8m35s
[root@master1 ~]#kubectl describe po test-pod
......
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 8m28s default-scheduler Successfully assigned default/test-pod to node2
Normal Pulling 8m28s kubelet Pulling image "busybox"
Normal Pulled 8m27s kubelet Successfully pulled image "busybox" in 495.870587ms
Normal Pulled 8m27s kubelet Container image "nginx:1.7.9" already present on machine
Normal Created 8m27s kubelet Created container app
Normal Started 8m27s kubelet Started container app
Normal Killing 49s kubelet Container sidecar1 definition changed, will be restarted
Normal Pulling 19s kubelet Pulling image "busybox:1.35.0"
Normal Started 1s (x2 over 8m27s) kubelet Started container sidecar1
Normal Created 1s (x2 over 8m27s) kubelet Created container sidecar1
Normal Pulled 1s kubelet Successfully pulled image "busybox:1.35.0" in 17.221245188s
[root@master1 ~]#kubectl get po test-pod -oyaml|grep busybox
kruise.io/sidecarset-inplace-update-state: '{"test-sidecarset":{"revision":"f78z4854d9855xd6478fzx9c84645z2548v24z26455db46bdfzw44v49v98f2cw","updateTimestamp":"2022-03-12T01:55:56Z","lastContainerStatuses":{"sidecar1":{"imageID":"docker.io/library/busybox@sha256:5acba83a746c7608ed544dc1533b87c737a0b0fb730301639a0179f9344b1678"}}}}'
image: busybox:1.35.0
image: docker.io/library/busybox:1.35.0
imageID: docker.io/library/busybox@sha256:130df6999605f982ec67e5bee29d3a52614a075e949490f0a41702ee1dd98f3f
可以看到 Pod 中的 sidecar 容器镜像被原地升级成 busybox:1.35.0
了, 对主容器没有产生任何影响。
实验结束。😘
1.同意特性
。 spec.containers
除了默认的 k8s container 字段,还扩展了如下一些字段,来方便注入:
apiVersion: apps.kruise.io/v1alpha1
kind: SidecarSet
metadata:
name: sidecarset
spec:
selector:
matchLabels:
app: sample
containers:
# 默认的K8s容器字段
- name: nginx
image: nginx:alpine
volumeMounts:
- mountPath: /nginx/conf
name: nginx.conf
# 扩展的sidecar容器字段
podInjectPolicy: BeforeAppContainer
shareVolumePolicy:
type: disabled | enabled
transferEnv:
- sourceContainerName: main
envName: PROXY_IP
volumes:
- Name: nginx.conf
hostPath: /data/nginx/conf
-
podInjectPolicy:定义了容器 注入到
pod.spec.containers
中的位置BeforeAppContainer
:表示注入到 pod 原 containers 的前面((默认) )AfterAppContainer
: 表示注入到 pod 原 containers 的后面
-
数据卷共享
- 共享指定卷:通过
spec.volumes
来定义 sidecar 自身需要的 volume - 共享所有卷:通过
spec.containers[i].shareVolumePolicy.type = enabled | disabled
来控制是否挂载 pod 应用容器的卷,常用于日志收集等 sidecar,配置为enabled
后会把应用容器中所有挂载点注入 sidecar 同一路经下(sidecar 中本身就有声明的数据卷和挂载点除外)
- 共享指定卷:通过
-
环境变量共享:可以通过
spec.containers[i].transferEnv
来从别的容器获取环境变量,会把名为sourceContainerName
容器中名为envName
的环境变量拷贝到本容器
🍀 SidecarSet 不仅支持 sidecar 容器的原地升级,而且提供了非常丰富的升级、灰度策略。同样在 SidecarSet 对象中 updateStrategy
属性下面也可以配置 partition
来定义保留旧版本 Pod 的数量或百分比,默认为 0;同样还可以配置的有 maxUnavailable
属性,表示在发布过程中的最大不可用数量。
- 当
{matched pod}=100,partition=50,maxUnavailable=10
,控制器会发布 50 个 Pod 到新版本,但是同一时间只会发布 10 个 Pod,每发布好一个 Pod 才会再找一个发布,直到 50 个发布完成。 - 当
{matched pod}=100,partition=80,maxUnavailable=30
,控制器会发布 20 个 Pod 到新版本,因为满足 maxUnavailable 数量,所以这 20 个 Pod 会同时发布。
🍀 同样也可以设置 paused: true
来暂停发布,此时对于新创建的、扩容的 pod 依旧会实现注入能力,已经更新的 pod 会保持更新后的版本不动,还没有更新的 pod 会暂停更新。
apiVersion: apps.kruise.io/v1alpha1
kind: SidecarSet
metadata:
name: sidecarset
spec:
# ...
updateStrategy:
type: RollingUpdate
maxUnavailable: 20%
partition: 10
paused: true
2.金丝雀发布
对于有金丝雀发布需求的业务,可以通过 selector
来实现,对于需要率先金丝雀灰度的 pod 打上固定的 [canary.release] = true
的标签,再通过 selector.matchLabels
来选中该 pod 即可。
💘 实验:金丝雀发布(测试成功)
- 比如现在我们有一个3副本的 Pod,也具有
app=nginx
的标签,如下所示
#03-deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: nginx
namespace: default
spec:
replicas: 3
revisionHistoryLimit: 3
selector:
matchLabels:
app: nginx
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: ngx
image: nginx:1.7.9
ports:
- containerPort: 80
- 创建后现在就具有4个
app=nginx
标签的 Pod 了,由于都匹配上面创建的 SidecarSet 对象,所以都会被注入一个sidecar1
的容器,镜像为busybox
:
$ kubectl apply -f 03-deployment.yaml
deployment.apps/nginx created
[root@master1 ~]#kubectl get pods -l app=nginx
NAME READY STATUS RESTARTS AGE
nginx-6457955f7-bt5fk 2/2 Running 0 36s
nginx-6457955f7-jphvr 2/2 Running 0 36s
nginx-6457955f7-zjvh9 2/2 Running 0 36s
test-pod 2/2 Running 1 (127m ago) 135m
-
现在如果我们想为
test-pod
这个应用来执行灰度策略,将 sidecar 容器镜像更新成busybox:1.35.0
,则可以在updateStrategy
下面添加selector.matchLabels
属性canary.release: "true"
:注意:我这里就直接在
01-sidecarset.yaml
上面直接修改了。另外,我之前原本test-pod里的busybox镜像就已经是busybox:1.35.0了,本次yaml里写成busybox即可。
# 01-sidecarset.yaml
apiVersion: apps.kruise.io/v1alpha1
kind: SidecarSet
metadata:
name: test-sidecarset
spec:
selector:
matchLabels:
app: nginx
updateStrategy:
type: RollingUpdate
maxUnavailable: 1
selector:
matchLabels: # 修改1:
canary.release: "true"
containers:
- name: sidecar1
image: busybox # 修改2:
command: ["sleep", "999d"]
volumeMounts:
- name: log-volume
mountPath: /var/log
volumes: # this field will be merged into pod.spec.volumes
- name: log-volume
emptyDir: {
}
- 然后同样需要给 test-pod 添加上
canary.release=true
这个标签:
#直接在02-test-pod文件里添加即可
apiVersion: v1
kind: Pod
metadata:
labels:
app: nginx
canary.release: "true"
name: test-pod
spec:
containers:
- name: app
image: nginx:1.7.9
- 更新后可以发现 test-pod 的 sidecar 镜像更新了,其他 Pod 没有变化,这样就实现了 sidecar 的灰度功能:
$ kubectl apply -f 02-test-pod.yaml
pod/test-pod configured
$ kubectl apply -f 01-sidecarset.yaml
sidecarset.apps.kruise.io/test-sidecarset configured
[root@master1 ~]#kubectl get po -l app=nginx
NAME READY STATUS RESTARTS AGE
nginx-6457955f7-bt5fk 2/2 Running 0 7m34s
nginx-6457955f7-jphvr 2/2 Running 0 7m34s
nginx-6457955f7-zjvh9 2/2 Running 0 7m34s
test-pod 2/2 Running 2 (10s ago) 142m
[root@master1 ~]#kubectl describe po test-pod
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Killing 74s (x2 over 135m) kubelet Container sidecar1 definition changed, will be restarted
Normal Pulling 44s (x2 over 142m) kubelet Pulling image "busybox"
Normal Created 43s (x3 over 142m) kubelet Created container sidecar1
Normal Started 43s (x3 over 142m) kubelet Started container sidecar1
Normal Pulled 43s kubelet Successfully pulled image "busybox" in 505.082017ms
#符合预期
[root@master1 ~]#kubectl get po test-pod -oyaml|grep busybox
kruise.io/sidecarset-inplace-update-state: '{"test-sidecarset":{"revision":"f78z4854d9855xd6478fzx9c84645z2548v24z26455db46bdfzw44v49v98f2cw","updateTimestamp":"2022-03-12T01:55:56Z","lastContainerStatuses":{"sidecar1":{"imageID":"docker.io/library/busybox@sha256:130df6999605f982ec67e5bee29d3a52614a075e949490f0a41702ee1dd98f3f"}}}}'
image: busybox
image: docker.io/library/busybox:latest
imageID: docker.io/library/busybox@sha256:5acba83a746c7608ed544dc1533b87c737a0b0fb730301639a0179f9344b1678
实验结束。😘
3.热升级
💘 实验:热升级(测试成功)
SidecarSet 原地升级会先停止旧版本的容器,然后创建新版本的容器,这种方式适合不影响 Pod 服务可用性的 sidecar 容器,比如说日志收集的 Agent。
但是对于很多,例如 Istio Envoy,这种升级方法就有问题了,Envoy 作为 Pod 中的一个,代理了所有的流量,如果直接重启,Pod 服务的可用性会受到影响,如果需要单独升级 envoy sidecar,,所以我们为这种 sidecar 容器的升级提供了一种新的解决方案。
- 编写04-sidecarset-hot.yaml
#04-sidecarset-hot.yaml
apiVersion: apps.kruise.io/v1alpha1
kind: SidecarSet
metadata:
name: hotupgrade-sidecarset
spec:
selector:
matchLabels:
app: hotupgrade
containers:
- name: sidecar
image: openkruise/hotupgrade-sample:sidecarv1
imagePullPolicy: Always
lifecycle:
postStart:
exec:
command:
- /bin/sh
- /migrate.sh #在主容器启动之前,做一些网络方面的迁移
upgradeStrategy:
upgradeType: HotUpgrade
hotUpgradeEmptyImage: openkruise/hotupgrade-sample:empty
upgradeType
:HotUpgrade
代表该 sidecar 容器的类型是热升级方案hotUpgradeEmptyImage
: 当热升级 sidecar 容器时,业务必须要提供一个 empty 容器用于热升级过程中的容器切换,empty 容器同 sidecar 容器具有相同的配置(除了镜像地址),例如:command、lifecycle、probe 等,但是它不做任何工作。lifecycle.postStart
: 在 postStart 这个 hook 中完成热升级过程中的状态迁移,该脚本需要由业务根据自身的特点自行实现,例如:nginx 热升级需要完成 Listen FD 共享以及 reload 操作。
整体来说热升级特性总共包含以下两个过程:
- Pod 创建时,注入热升级容器
- 原地升级时,完成热升级流程
Pod 创建时,SidecarSet Webhook 将会注入两个容器:
{sidecarContainer.name}-1
: 如下图所示 envoy-1,这个容器代表正在实际工作的 sidecar 容器,例如:envoy:1.16.0{sidecarContainer.name}-2
: 如下图所示 envoy-2,这个容器是业务配置的 hotUpgradeEmptyImage 容器,例如:empty:1.0,用于后面的热升级机制
热升级流程主要分为三个步骤:
Upgrade
: 将 empty 容器升级为当前最新的 sidecar 容器,例如:envoy-2.Image = envoy:1.17.0Migration
:lifecycle.postStart
完成热升级流程中的状态迁移,当迁移完成后退出Reset
: 状态迁移完成后,热升级流程将设置 envoy-1 容器为 empty 镜像,例如:envoy-1.Image = empty:1.0
上述三个步骤完成了热升级中的全部流程,当对 Pod 执行多次热升级时,将重复性的执行上述三个步骤。
⚠️ empty这个空容器的目的就是占位。
- 部署上面04-sidecarset-hot.yaml资源
$ kubectl apply -f 04-sidecarset-hot.yaml
sidecarset.apps.kruise.io/hotupgrade-sidecarset created
[root@master1 ~]#kubectl get sidecarset
NAME MATCHED UPDATED READY AGE
hotupgrade-sidecarset 0 0 0 20s
- 这里我们以 OpenKruise 的官方示例来进行说明,首先创建上面的
hotupgrade-sidecarset
这个 SidecarSet。然后创建一个如下所示的 CloneSet 对象:
#05-hotupgrade-CloneSet.yaml
apiVersion: apps.kruise.io/v1alpha1
kind: CloneSet
metadata:
labels:
app: hotupgrade
name: busybox
spec:
replicas: 1
selector:
matchLabels:
app: hotupgrade
template: #下面是pod模板
metadata:
labels:
app: hotupgrade #这个标签要和上面的sidercarset匹配
spec:
containers:
- name: busybox
image: openkruise/hotupgrade-sample:busybox
- 创建完成后,CloneSet 管理的 Pod 已经注入
sidecar-1
和sidecar-2
两个容器:
#部署
$ kubectl apply -f 05-hotupgrade-CloneSet.yaml
cloneset.apps.kruise.io/busybox created
[root@master1 ~]#kubectl get po -l app=hotupgrade
NAME READY STATUS RESTARTS AGE
busybox-695lv 3/3 Running 0 71s
[root@master1 ~]#kubectl describe po busybox-695lv
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 119s default-scheduler Successfully assigned default/busybox-695lv to node2
Normal Pulling 119s kubelet Pulling image "openkruise/hotupgrade-sample:sidecarv1"
Normal Pulled 116s kubelet Successfully pulled image "openkruise/hotupgrade-sample:sidecarv1" in 2.430148491s
Normal Created 116s kubelet Created container sidecar-1
Normal Started 116s kubelet Started container sidecar-1
Normal Pulling 111s kubelet Pulling image "openkruise/hotupgrade-sample:empty"
Normal Pulled 93s kubelet Successfully pulled image "openkruise/hotupgrade-sample:empty" in 17.732956857s
Normal Created 93s kubelet Created container sidecar-2
Normal Started 93s kubelet Started container sidecar-2
Normal Pulling 93s kubelet Pulling image "openkruise/hotupgrade-sample:busybox"
Normal Pulled 90s kubelet Successfully pulled image "openkruise/hotupgrade-sample:busybox" in 2.81313947s
Normal Created 90s kubelet Created container busybox
[root@master1 ~]#kubectl get po busybox-695lv -oyaml
……
spec:
containers:
- env:
- name: IS_INJECTED
value: "true"
- name: SIDECARSET_VERSION
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['version.sidecarset.kruise.io/sidecar-1']
- name: SIDECARSET_VERSION_ALT
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['versionalt.sidecarset.kruise.io/sidecar-1']
image: openkruise/hotupgrade-sample:sidecarv1
imagePullPolicy: Always
lifecycle:
postStart:
exec:
command:
- /bin/sh
- /migrate.sh
name: sidecar-1
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
- env:
- name: IS_INJECTED
value: "true"
- name: SIDECARSET_VERSION
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['version.sidecarset.kruise.io/sidecar-2']
- name: SIDECARSET_VERSION_ALT
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['versionalt.sidecarset.kruise.io/sidecar-2']
image: openkruise/hotupgrade-sample:empty
imagePullPolicy: Always
lifecycle:
postStart:
exec:
command:
- /bin/sh
- /migrate.sh
name: sidecar-2
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
- image: openkruise/hotupgrade-sample:busybox
imagePullPolicy: Always
name: busybox
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
……
- busybox 主容器每100毫秒会请求一次sidecar(version=v1)服务:
[root@master1 ~]#kubectl logs -f busybox-695lv -c busybox
I0312 11:51:49.100832 1 main.go:39] request sidecar server success, and response(body=This is version(v1) sidecar)
I0312 11:51:49.212847 1 main.go:39] request sidecar server success, and response(body=This is version(v1) sidecar)
......
- 现在我们去升级 sidecar 容器,将镜像修改为
openkruise/hotupgrade-sample:sidecarv2
:
[root@master1 ~]#kubectl patch sidecarset hotupgrade-sidecarset --type='json' -p='[{"op": "replace", "path": "/spec/containers/0/image", "value": "openkruise/hotupgrade-sample:sidecarv2"}]'
sidecarset.apps.kruise.io/hotupgrade-sidecarset patched
- 更新后再去观察 pod 的状态,可以看到 sidecar-2 镜像正常更新了:
[root@master1 ~]#kubectl get po -l app=hotupgrade
NAME READY STATUS RESTARTS AGE
busybox-695lv 3/3 Running 2 (103s ago) 8m38s
[root@master1 ~]#kubectl describe po busybox-695lv
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 9m10s default-scheduler Successfully assigned default/busybox-695lv to node2
Normal Pulling 9m10s kubelet Pulling image "openkruise/hotupgrade-sample:sidecarv1"
Normal Pulled 9m7s kubelet Successfully pulled image "openkruise/hotupgrade-sample:sidecarv1" in 2.430148491s
Normal Pulling 9m2s kubelet Pulling image "openkruise/hotupgrade-sample:empty"
Normal Pulling 8m44s kubelet Pulling image "openkruise/hotupgrade-sample:busybox"
Normal Pulled 8m44s kubelet Successfully pulled image "openkruise/hotupgrade-sample:empty" in 17.732956857s
Normal Started 8m41s kubelet Started container busybox
Normal Created 8m41s kubelet Created container busybox
Normal Pulled 8m41s kubelet Successfully pulled image "openkruise/hotupgrade-sample:busybox" in 2.81313947s
Normal Killing 2m49s kubelet Container sidecar-2 definition changed, will be restarted
Normal Pulling 2m48s kubelet Pulling image "openkruise/hotupgrade-sample:sidecarv2"
Normal Pulled 2m27s kubelet Successfully pulled image "openkruise/hotupgrade-sample:sidecarv2" in 21.791000213s
Normal Created 2m27s (x2 over 8m44s) kubelet Created container sidecar-2
Normal Started 2m26s (x2 over 8m44s) kubelet Started container sidecar-2
Normal Killing 2m16s kubelet Container sidecar-1 definition changed, will be restarted
Normal ResetContainerSucceed 2m16s sidecarset-controller reset sidecar container image empty successfully
Normal Pulling 2m15s kubelet Pulling image "openkruise/hotupgrade-sample:empty"
Normal Started 2m (x2 over 9m7s) kubelet Started container sidecar-1
Normal Created 2m (x2 over 9m7s) kubelet Created container sidecar-1
Normal Pulled 2m kubelet Successfully pulled image "openkruise/hotupgrade-sample:empty" in 15.229447021s
[root@master1 ~]#kubectl get po busybox-695lv -oyaml
……
spec:
containers:
- env:
- name: IS_INJECTED
value: "true"
- name: SIDECARSET_VERSION
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['version.sidecarset.kruise.io/sidecar-1']
- name: SIDECARSET_VERSION_ALT
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['versionalt.sidecarset.kruise.io/sidecar-1']
image: openkruise/hotupgrade-sample:empty
imagePullPolicy: Always
lifecycle:
postStart:
exec:
command:
- /bin/sh
- /migrate.sh
name: sidecar-1
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
- env:
- name: IS_INJECTED
value: "true"
- name: SIDECARSET_VERSION
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['version.sidecarset.kruise.io/sidecar-2']
- name: SIDECARSET_VERSION_ALT
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.annotations['versionalt.sidecarset.kruise.io/sidecar-2']
image: openkruise/hotupgrade-sample:sidecarv2
imagePullPolicy: Always
lifecycle:
postStart:
exec:
command:
- /bin/sh
- /migrate.sh
name: sidecar-2
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
- image: openkruise/hotupgrade-sample:busybox
imagePullPolicy: Always
name: busybox
resources: {
}
terminationMessagePath: /dev/termination-log
terminationMessagePolicy: File
volumeMounts:
- mountPath: /var/run/secrets/kubernetes.io/serviceaccount
name: kube-api-access-n9lpf
readOnly: true
……
并且在更新过程中观察 busybox 容器仍然会不断请求 sidecar 服务,但是并没有失败的请求出现:
[root@master1 ~]#kubectl logs -f busybox-695lv -c busybox
🍀 整个热升级示例代码可以参考仓库的实现:https://github.com/openkruise/samples/tree/master/hotupgrade。
实验结束。😘
2、Container Restart
ContainerRecreateRequest
控制器可以帮助用户重启/重建存量 Pod 中一个或多个容器。和 Kruise 提供的原地升级类似,当一个容器重建的时候,Pod 中的其他容器还保持正常运行,重建完成后,Pod 中除了该容器的 restartCount 增加以外不会有什么其他变化。
为要重建容器的 Pod 提交一个 ContainerRecreateRequest
自定义资源(缩写 CRR):
apiVersion: apps.kruise.io/v1alpha1
kind: ContainerRecreateRequest
metadata:
namespace: pod-namespace
name: xxx
spec:
podName: pod-name
containers: # 要重建的容器名字列表,至少要有 1 个
- name: app
- name: sidecar
strategy:
failurePolicy: Fail # 'Fail' 或 'Ignore',表示一旦有某个容器停止或重建失败, CRR 立即结束
orderedRecreate: false # 'true' 表示要等前一个容器重建完成了,再开始重建下一个
terminationGracePeriodSeconds: 30 # 等待容器优雅退出的时间,不填默认用 Pod 中定义的
unreadyGracePeriodSeconds: 3 # 在重建之前先把 Pod 设为 not ready,并等待这段时间后再开始执行重建
minStartedSeconds: 10 # 重建后新容器至少保持运行这段时间,才认为该容器重建成功
activeDeadlineSeconds: 300 # 如果 CRR 执行超过这个时间,则直接标记为结束(未结束的容器标记为失败)
ttlSecondsAfterFinished: 1800 # CRR 结束后,过了这段时间自动被删除掉
一般来说,列表中的容器会一个个被停止,但可能同时在被重建和启动,除非 orderedRecreate
被设置为 true。 unreadyGracePeriodSeconds
功能依赖于 KruisePodReadinessGate
这个 feature-gate,后者会在每个 Pod 创建的时候注入一个 readinessGate
,否则,默认只会给 Kruise workload 创建的 Pod 注入 readinessGate,也就是说只有这些 Pod 才能在 CRR 重建时使用 unreadyGracePeriodSeconds
。
3、ImagePullJob
NodeImage
和 ImagePullJob
是从 Kruise v0.8.0 版本开始提供的 CRD。Kruise 会自动为每个 Node 创建一个 NodeImage,它包含了哪些镜像需要在这个 Node 上做预热,比如我们这里3个节点,则会自动创建3个 NodeImage 对象:
➜ kubectl get nodeimage
NAME DESIRED PULLING SUCCEED FAILED AGE
master1 0 0 0 0 10d
node1 0 0 0 0 10d
node2 0 0 0 0 10d
- 比如我们查看 node1 节点上的 NodeImage 对象:
➜ kruise kubectl get nodeimage node1 -o yaml
apiVersion: apps.kruise.io/v1alpha1
kind: NodeImage
metadata:
labels:
beta.kubernetes.io/arch: amd64
beta.kubernetes.io/os: linux
kubernetes.io/arch: amd64
kubernetes.io/hostname: node1
kubernetes.io/os: linux
name: node1
spec: {
}
status:
desired: 0
failed: 0
pulling: 0
succeeded: 0
- 比如我们希望在这个节点上拉去一个
ubuntu:latest
镜像,则可以按照如下所示的去修改 spec:
......
spec:
images:
ubuntu: # 镜像 name
tags:
- tag: latest # 镜像 tag
pullPolicy:
ttlSecondsAfterFinished: 300 # [required] 拉取完成(成功或失败)超过 300s 后,将这个任务从 NodeImage 中清除
timeoutSeconds: 600 # [optional] 每一次拉取的超时时间, 默认为 600
backoffLimit: 3 # [optional] 拉取的重试次数,默认为 3
activeDeadlineSeconds: 1200 # [optional] 整个任务的超时时间,无默认值
更新后我们可以从 status 中看到拉取进度以及结果,并且你会发现拉取完成 600s 后任务会被清除。
🍀 此外用户可以创建 ImagePullJob
对象,来指定一个镜像要在哪些节点上做预热。
- 比如创建如下所示的
ImagePullJob
资源对象:
apiVersion: apps.kruise.io/v1alpha1
kind: ImagePullJob
metadata:
name: job-with-always
spec:
image: nginx:1.9.1 # [requir