配置更改时重启 Pod
清理
请参阅入门文档以了解安装和设置信息
microk8s
UCS 工具
访问 NGC
设置仓库
简介
在本教程中，我们将首先演示先前教程 配置更改未导致 Pod 重启 中提到的问题，并展示如何添加对 configmap 的依赖，以便 configmap 中的任何更改都会导致 Pod 重启。
本教程是教程 应用程序的部分重新部署 的延续。
检查当前的微服务和资源
$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 44m
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Running 0 2m16s
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 44m
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 44m
deployment.apps/http-server-http-server-deployment 1/1 1 1 44m
NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 44m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 1 1 1 44m
让我们也检查一下 configmap 的内容，其中添加了 workingDir 参数。
$ microk8s kubectl get cm -l "app.kubernetes.io/instance=server-client"
NAME DATA AGE
http-server-workload-cm 0 49m
curl-client-workload-cm 0 49m
http-server-configs-cm 2 49m
http-server-scripts-cm 1 49m
curl-client-scripts-cm 1 49m
curl-client-configs-cm 1 49m
$ microk8s kubectl get cm http-server-configs-cm -o yaml
apiVersion: v1
data:
config.json: |
{
"workingDir": "/usr/local"
}
config.yaml: |
SampleConfig:
sampleValue: 0
kind: ConfigMap
metadata:
annotations:
meta.helm.sh/release-name: server-client
meta.helm.sh/release-namespace: default
creationTimestamp: "2023-03-14T06:05:18Z"
labels:
app.kubernetes.io/instance: server-client
app.kubernetes.io/managed-by: Helm
app.kubernetes.io/name: http-server
app.kubernetes.io/version: 0.0.4
generated_with: helm_builder
hb_version: 0.7.1
helm.sh/chart: http-server-0.0.4
msb_version: 2.5.0
name: http-server-configs-cm
namespace: default
resourceVersion: "954500"
uid: 8bdae84f-84b0-411b-a7aa-337cab82bdc2
正如我们所见，config.json 包含 /usr/local 的当前路径。
更新应用程序中的参数
将 app.yaml 文件更新为以下内容
specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application
dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.4
components:
- name: client
type: ucf.svc.curl-client
- name: http-server
type: ucf.svc.http-server
parameters:
workingDir: /var
fileToCreate: someotherfile.txt
connections:
client/http: http-server/http
参数 workingDir 已被修改，应用程序版本已更新。
构建应用程序和部分重新部署
使用以下命令重建和重新部署应用程序
$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5
检查微服务和应用程序
$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 56m
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Running 0 15m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 57m
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 56m
deployment.apps/http-server-http-server-deployment 1/1 1 1 56m
NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 56m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 1 1 1 56m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 53m
我们还可以获取 curl-client 容器的日志
$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 936 100 936 0 0 457k 0 --:--:-- --:--:-- --:--:-- 457k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="bin/">bin/</a></li>
<li><a href="cuda/">cuda@</a></li>
<li><a href="cuda-11/">cuda-11@</a></li>
<li><a href="cuda-11.4/">cuda-11.4/</a></li>
<li><a href="cuda-11.6/">cuda-11.6/</a></li>
<li><a href="etc/">etc/</a></li>
<li><a href="games/">games/</a></li>
<li><a href="include/">include/</a></li>
<li><a href="lib/">lib/</a></li>
<li><a href="man/">man@</a></li>
<li><a href="mpi/">mpi@</a></li>
<li><a href="nvm/">nvm/</a></li>
<li><a href="sbin/">sbin/</a></li>
<li><a href="share/">share/</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="src/">src/</a></li>
<li><a href="ucx/">ucx@</a></li>
</ul>
<hr>
</body>
</html>
让我们也获取 http-server configmap 的内容
$ microk8s kubectl get cm http-server-configs-cm -o yaml
apiVersion: v1
data:
config.json: |
{
"workingDir": "/var"
}
config.yaml: |
SampleConfig:
sampleValue: 0
kind: ConfigMap
metadata:
annotations:
meta.helm.sh/release-name: server-client
meta.helm.sh/release-namespace: default
creationTimestamp: "2023-03-14T06:05:18Z"
labels:
app.kubernetes.io/instance: server-client
app.kubernetes.io/managed-by: Helm
app.kubernetes.io/name: http-server
app.kubernetes.io/version: 0.0.4
generated_with: helm_builder
hb_version: 0.7.1
helm.sh/chart: http-server-0.0.4
msb_version: 2.5.0
name: http-server-configs-cm
namespace: default
resourceVersion: "963732"
uid: 8bdae84f-84b0-411b-a7aa-337cab82bdc2
我们可以看到，即使 config 反映了更新后的路径 /var，http-server 的现有 Pod 仍然继续运行并返回旧路径 /usr/local 的目录内容。
添加对 configmap 的依赖
让我们更新 http-server 的清单
type: msapplication
specVersion: 2.5.0
name: ucf.svc.http-server
chartName: http-server
description: http server
version: 0.0.5
tags: []
keywords: []
publish: false
ingress-endpoints:
- name: http
description: REST API endpoint
protocol: TCP
scheme: http
mandatory: False
data-flow: in-out
params:
workingDir: "/localvol"
#> type: string
#> description: Working directory for listing contents
#> flags: mandatory
fileToCreate: somefile.txt
#> type: string
#> description: Name of the file to create in the working directory
#> flags: mandatory
---
spec:
- name: http-server-deployment
type: ucf.k8s.app.deployment
parameters:
apptype: stateless
- name: http-server-container
type: ucf.k8s.container
parameters:
image:
repository: nvcr.io/nvidia/pytorch
tag: 22.04-py3
command: [sh, -c]
args: [
"WORKING_DIR=$(cat /opt/configs/config.json | jq -r '.workingDir') && cd $WORKING_DIR && echo $PWD && touch $(FILE_TO_CREATE) && ls && python -m http.server 8080
"]
env:
- name: POD_NAME
valueFrom:
fieldRef:
fieldPath: metadata.name
- name: FILE_TO_CREATE
value: $params.fileToCreate
ports:
- containerPort: 8080
name: http
volumeMounts:
- name: localvol
mountPath: /localvol
- name: svc
type: ucf.k8s.service
parameters:
ports:
- port: 8080
protocol: TCP
name: http
- name: localvol
type: ucf.k8s.volume
parameters:
persistentVolumeClaim:
claimName: local-path-pvc
- name: cm-dependencies
type: ucf.appspec.restartPodOnConfigChanges
parameters:
# Add dependency on all configmaps detected in the microservice
addAll: true
# Add dependency on default configmaps for scripts, configs and workload configs
# addDefault: true
# Add dependency on individual configmaps using the configmap names
# configMaps:
# - http-server-configs-cm
清单中有两个更改
已添加类型为
ucf.appspec.restartPodOnConfigChanges的新组件cm-dependencies，并将addAll设置为true微服务版本已更新
将 app.yaml 文件更新为以下内容，仅 dependencies 中的 http-server 版本已更新
specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application
dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.5
components:
- name: client
type: ucf.svc.curl-client
- name: http-server
type: ucf.svc.http-server
parameters:
workingDir: /var
fileToCreate: someotherfile.txt
connections:
client/http: http-server/http
重建和重新部署微服务和应用程序
使用以下命令重建和重新部署微服务和应用程序
$ cd $HOME/ucf_tutorial/services
$ ucf_ms_builder_cli service build -d http-server/ -f
$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5
检查重新部署的应用程序
$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 78m
pod/http-server-http-server-deployment-d5764c767-vx6rp 1/1 Running 0 17s
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Terminating 0 37m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 78m
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 78m
deployment.apps/http-server-http-server-deployment 1/1 1 1 78m
NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 78m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 75m
replicaset.apps/http-server-http-server-deployment-d5764c767 1 1 1 17s
正如我们所见，现有的 http-server Pod 已被终止，并启动了新的 Pod。检查 curl-client Pod 和新的 http-server Pod 的日志证实了新路径 /var 已生效
$ microk8s kubectl logs --tail -1 -l "app=http-server-http-server-deployment,microservice_version=0.0.5"
/var
backups
cache
lib
local
lock
log
mail
opt
run
someotherfile.txt
spool
tmp
$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 743 100 743 0 0 362k 0 --:--:-- --:--:-- --:--:-- 362k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="backups/">backups/</a></li>
<li><a href="cache/">cache/</a></li>
<li><a href="lib/">lib/</a></li>
<li><a href="local/">local/</a></li>
<li><a href="lock/">lock@</a></li>
<li><a href="log/">log/</a></li>
<li><a href="mail/">mail/</a></li>
<li><a href="opt/">opt/</a></li>
<li><a href="run/">run@</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="spool/">spool/</a></li>
<li><a href="tmp/">tmp/</a></li>
</ul>
<hr>
</body>
</html>
更新配置参数并重建和重新部署应用程序
将 app.yaml 文件更新为以下内容，仅 workingDir 参数已更新为 /tmp
specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application
dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.5
components:
- name: client
type: ucf.svc.curl-client
- name: http-server
type: ucf.svc.http-server
parameters:
workingDir: /tmp
fileToCreate: someotherfile.txt
connections:
client/http: http-server/http
重建和重新部署应用程序
使用以下命令重建和重新部署微服务和应用程序
$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5
检查重新部署的应用程序
$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 93m
pod/http-server-http-server-deployment-c88597df9-wr89f 1/1 Running 0 21s
pod/http-server-http-server-deployment-d5764c767-vx6rp 1/1 Terminating 0 15m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 93m
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 93m
deployment.apps/http-server-http-server-deployment 1/1 1 1 93m
NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 93m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 90m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 0 0 0 93m
replicaset.apps/http-server-http-server-deployment-c88597df9 1 1 1 21s
replicaset.apps/http-server-http-server-deployment-d5764c767 0 0 0 15m
正如我们所见，现有的 http-server Pod 已被终止，并启动了新的 Pod。检查 curl-client Pod 和新的 http-server Pod 的日志证实了新路径 /tmp 已生效
$ microk8s kubectl logs http-server-http-server-deployment-c88597df9-wr89f
/tmp
core-js-banners
someotherfile.txt
v8-compile-cache-0
yarn--1650154850728-0.1942747438338328
yarn--1650154852391-0.6050681633516224
yarn--1650154853881-0.6683114706726603
yarn--1650154920633-0.4394541481795069
yarn--1650154922130-0.5297801175307726
$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 989 100 989 0 0 482k 0 --:--:-- --:--:-- --:--:-- 482k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="core-js-banners">core-js-banners</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="v8-compile-cache-0/">v8-compile-cache-0/</a></li>
<li><a href="yarn--1650154850728-0.1942747438338328/">yarn--1650154850728-0.1942747438338328/</a></li>
<li><a href="yarn--1650154852391-0.6050681633516224/">yarn--1650154852391-0.6050681633516224/</a></li>
<li><a href="yarn--1650154853881-0.6683114706726603/">yarn--1650154853881-0.6683114706726603/</a></li>
<li><a href="yarn--1650154920633-0.4394541481795069/">yarn--1650154920633-0.4394541481795069/</a></li>
<li><a href="yarn--1650154922130-0.5297801175307726/">yarn--1650154922130-0.5297801175307726/</a></li>
</ul>
<hr>
</body>
</html>
这证实了 configmap 内容的更改在微服务清单中 ucf.appspec.restartPodOnConfigChanges 组件的帮助下导致 Pod 重启。
停止和清理微服务和应用程序
最后，要停止和清理应用程序，我们可以运行
$ microk8s helm3 uninstall server-client
一些注意事项
只有作为微服务 Helm Chart 一部分的 configmap 才能添加为依赖项。