plugin

podinfo

Podinfo is a tiny web application made with Go that showcases best practices of running microservices in Kubernetes. Podinfo is used by CNCF projects like Flux and Flagger for end-to-end testing and workshops.

Specifications:

• Health checks (readiness and liveness)
• Graceful shutdown on interrupt signals
• File watcher for secrets and configmaps
• Instrumented with Prometheus and Open Telemetry
• Structured logging with zap
• 12-factor app with viper
• Fault injection (random errors and latency)
• Swagger docs
• Timoni, Helm and Kustomize installers
• End-to-End testing with Kubernetes Kind and Helm
• Multi-arch container image with Docker buildx and Github Actions
• Container image signing with Sigstore cosign
• SBOMs and SLSA Provenance embedded in the container image
• CVE scanning with Trivy

Web API:

• GET / prints runtime information
• GET /version prints podinfo version and git commit hash
• GET /metrics return HTTP requests duration and Go runtime metrics
• GET /healthz used by Kubernetes liveness probe
• GET /readyz used by Kubernetes readiness probe
• POST /readyz/enable signals the Kubernetes LB that this instance is ready to receive traffic
• POST /readyz/disable signals the Kubernetes LB to stop sending requests to this instance
• GET /status/{code} returns the status code
• GET /panic crashes the process with exit code 255
• POST /echo forwards the call to the backend service and echos the posted content
• GET /env returns the environment variables as a JSON array
• GET /headers returns a JSON with the request HTTP headers
• GET /delay/{seconds} waits for the specified period
• POST /token issues a JWT token valid for one minute JWT=$(curl -sd 'anon' podinfo:9898/token | jq -r .token)
• GET /token/validate validates the JWT token curl -H "Authorization: Bearer $JWT" podinfo:9898/token/validate
• GET /configs returns a JSON with configmaps and/or secrets mounted in the config volume
• POST/PUT /cache/{key} saves the posted content to Redis
• GET /cache/{key} returns the content from Redis if the key exists
• DELETE /cache/{key} deletes the key from Redis if exists
• POST /store writes the posted content to disk at /data/hash and returns the SHA1 hash of the content
• GET /store/{hash} returns the content of the file /data/hash if exists
• GET /ws/echo echos content via websockets podcli ws ws://localhost:9898/ws/echo
• GET /chunked/{seconds} uses transfer-encoding type chunked to give a partial response and then waits for the specified period
• GET /swagger.json returns the API Swagger docs, used for Linkerd service profiling and Gloo routes discovery

gRPC API:

• /grpc.health.v1.Health/Check health checking

Web UI:

To access the Swagger UI open <podinfo-host>/swagger/index.html in a browser.

Guides

• Getting started with Timoni
• Getting started with Flux
• Progressive Deliver with Flagger and Linkerd
• Automated canary deployments with Kubernetes Gateway API

Install

To install Podinfo on Kubernetes the minimum required version is Kubernetes v1.23.

Timoni

Install with Timoni:

timoni -n default apply podinfo oci://ghcr.io/stefanprodan/modules/podinfo

Helm

Install from github.io:

helm repo add podinfo https://stefanprodan.github.io/podinfo

helm upgrade --install --wait frontend \
--namespace test \
--set replicaCount=2 \
--set backend=http://backend-podinfo:9898/echo \
podinfo/podinfo

helm test frontend --namespace test

helm upgrade --install --wait backend \
--namespace test \
--set redis.enabled=true \
podinfo/podinfo

Install from ghcr.io:

helm upgrade --install --wait podinfo --namespace default \
oci://ghcr.io/stefanprodan/charts/podinfo

Kustomize

kubectl apply -k gitee.com/fpy-go/plugin//kustomize

Docker

docker run -dp 9898:9898 stefanprodan/podinfo

Continuous Delivery

In order to install podinfo on a Kubernetes cluster and keep it up to date with the latest release in an automated manner, you can use Flux.

Install the Flux CLI on MacOS and Linux using Homebrew:

brew install fluxcd/tap/flux

Install the Flux controllers needed for Helm operations:

flux install \
--namespace=flux-system \
--network-policy=false \
--components=source-controller,helm-controller

Add podinfo's Helm repository to your cluster and configure Flux to check for new chart releases every ten minutes:

flux create source helm podinfo \
--namespace=default \
--url=https://stefanprodan.github.io/podinfo \
--interval=10m

Create a podinfo-values.yaml file locally:

cat > podinfo-values.yaml <<EOL
replicaCount: 2
resources:
  limits:
    memory: 256Mi
  requests:
    cpu: 100m
    memory: 64Mi
EOL

Create a Helm release for deploying podinfo in the default namespace:

flux create helmrelease podinfo \
--namespace=default \
--source=HelmRepository/podinfo \
--release-name=podinfo \
--chart=podinfo \
--chart-version=">5.0.0" \
--values=podinfo-values.yaml

Based on the above definition, Flux will upgrade the release automatically when a new version of podinfo is released. If the upgrade fails, Flux can rollback to the previous working version.

You can check what version is currently deployed with:

flux get helmreleases -n default

To delete podinfo's Helm repository and release from your cluster run:

flux -n default delete source helm podinfo
flux -n default delete helmrelease podinfo

If you wish to manage the lifecycle of your applications in a GitOps manner, check out this workflow example for multi-env deployments with Flux, Kustomize and Helm.

NOTICE

插件命名规范

protocol-grpc/rpc-xxx

在重新加载插件之前比做事项

必须调用manager的kill方法，关闭client和server的连接，并关闭server服务，否则删除对应的插件文件会失败

基于云原生的场景，我们其实可以不把插件放到minio等网络存储中，直接本地存储更高效

原因在于：云原生下使用pv来共享目录，多个pod公用共一个pv，再加上路径监听，就可以实现多个pod的插件加载

client启动时我们如何监听到server服务端口的

可以参考go-plugin的client和server类，里面通过stdout来读取server的ip和端口

如果server意外终止

插件调用会返回err：connection is shut down 这时候我们要做处理，提示或者日志记录

如果client连接意外终止

插件调用会返回err：connection is shut down 这时候我们要做处理，提示或者日志记录

go-plugin的rpc_client使用gob（非json）包实现的序列化和反序列化

所以我们在使用的时候要注册声明的对象到gob中，参考sagoomqtt.go

对于插件的Encode和Decode方法的参数，如果指定了类型，则没问题，如果使用interface{}，则需要把实际的参数类型注册到gob中

例如Decode(args interface{})方法，传参是Dog类型，那么Dog类型就必须在插件中Register

grpc

PluginClient -- Broker -- brokerID -- con -- PluginServer