Identity and Access Management

Adding Identity and Access Management functionality to fsm sidecar via Plugins

In this demonstration, we will extend the IAM (Identity and Access Management) feature for the service mesh to enhance the security of the service. When service A accesses service B, it will carry the obtained token. After receiving the request, service B verifies the token through the authentication service, and based on the verification result, decides whether to serve the request or not.

Two plugins are required here:

token-injector to inject the token into the request from service A
token-verifyer to verify the identity of the request accessing service B.

Both of them handle outbound and inbound traffic, respectively.

Corresponding to this are two PluginChains:

token-injector-chain
token-verifier-chain

Prerequisites

Kubernetes cluster running Kubernetes v1.19.0 or greater.
Have fsm installed.
Have kubectl available to interact with the API server.
Have fsm CLI available for managing the service mesh.
Have jq command available.

Deploy demo services

kubectl create namespace curl
fsm namespace add curl
kubectl apply -n curl -f https://raw.githubusercontent.com/flomesh-io/fsm-docs/main/manifests/samples/curl/curl.yaml

kubectl create namespace httpbin
fsm namespace add httpbin
kubectl apply -n httpbin -f https://raw.githubusercontent.com/flomesh-io/fsm-docs/main/manifests/samples/httpbin/httpbin.yaml

sleep 2
kubectl wait --for=condition=ready pod -n curl -l app=curl --timeout=90s
kubectl wait --for=condition=ready pod -n httpbin -l app=httpbin --timeout=90s

curl_pod=`kubectl get pod -n curl -l app=curl -o jsonpath='{.items..metadata.name}'`
httpbin_pod=`kubectl get pod -n httpbin -l app=httpbin -o jsonpath='{.items..metadata.name}'`

To view the content of the plugin chains for both services. The built-in plugins are located in the modules directory.

These built-in plugins are native functions provided by the service mesh and are not configured through plugin mechanism, but can be overridden via plugin mechanism.

fsm proxy get config_dump -n curl $curl_pod | jq '.Chains."outbound-http"'
[
  "modules/outbound-http-routing.js",
  "modules/outbound-metrics-http.js",
  "modules/outbound-tracing-http.js",
  "modules/outbound-logging-http.js",
  "modules/outbound-circuit-breaker.js",
  "modules/outbound-http-load-balancing.js",
  "modules/outbound-http-default.js"
]

fsm proxy get config_dump -n httpbin $httpbin_pod | jq '.Chains."inbound-http"'
[
  "modules/inbound-tls-termination.js",
  "modules/inbound-http-routing.js",
  "modules/inbound-metrics-http.js",
  "modules/inbound-tracing-http.js",
  "modules/inbound-logging-http.js",
  "modules/inbound-throttle-service.js",
  "modules/inbound-throttle-route.js",
  "modules/inbound-http-load-balancing.js",
  "modules/inbound-http-default.js"
]

Test communication between the applications.

kubectl exec $curl_pod -n curl -c curl -- curl -Is http://httpbin.httpbin:14001/get
HTTP/1.1 200 OK
server: gunicorn/19.9.0
date: Sun, 05 Feb 2023 05:42:51 GMT
content-type: application/json
content-length: 304
access-control-allow-origin: *
access-control-allow-credentials: true
connection: keep-alive

Deploy authentication service

Deploy a standalone authentication service to authenticate requests and return 200 or 401. For simplicity, here we have hard-coded a valid token as 2f1acc6c3a606b082e5eef5e54414ffb

kubectl create namespace auth

kubectl apply -n auth -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: ext-auth
  name: ext-auth
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ext-auth
  strategy: {}
  template:
    metadata:
      creationTimestamp: null
      labels:
        app: ext-auth
    spec:
      containers:
      - command:
        - pipy
        - -e
        - |2-

          pipy({
            _acceptTokens: ['2f1acc6c3a606b082e5eef5e54414ffb'],
            _allow: false,
          })

            // Pipeline layouts go here, e.g.:
            .listen(8079)
            .demuxHTTP().to($ => $
              .handleMessageStart(
                msg => ((token = msg?.head?.headers?.['x-iam-token']) =>
                  _allow = token && _acceptTokens?.find(el => el == token)
                )()
              )
              .branch(() => _allow, $ => $.replaceMessage(new Message({ status: 200 })),
                $ => $.replaceMessage(new Message({ status: 401 }))
              )
            )
        image: flomesh/pipy:latest
        name: pipy
        resources: {}
---
apiVersion: v1
kind: Service
metadata:
  labels:
    app: ext-auth
  name: ext-auth
spec:
  ports:
  - port: 8079
    protocol: TCP
    targetPort: 8079
  selector:
    app: ext-auth
EOF

Enable plugin policy mode

To enable plugin policy, the mesh configuration needs to be modified as plugin policies are not enabled by default.

export fsm_namespace=fsm-system
kubectl patch meshconfig fsm-mesh-config -n "$fsm_namespace" -p '{"spec":{"featureFlags":{"enablePluginPolicy":true}}}' --type=merge

Declaring a plugin

Plugin token-injector:

metadata.name: the name of the plugin, which is also the name of the plugin script. For example, this plugin will be saved as token-injector.js stored in the plugins directory of the code repository.
spec.pipyscript: the PipyJS script content, which is the functional logic code, stored in the script file plugins/token-injector.js. Context metadata that is built-in to the system can be used within the script.
spec.priority: the priority of the plugin, with optional values of 0-65535. The higher the value, the higher the priority, and the earlier the plugin is positioned in the plugin chain. The value here is 115, which, based on the built-in plugin list in Helm values.yaml, will be positioned between modules/outbound-circuit-breaker.js and modules/outbound-http-load-balancing.js, executed after the circuit breaker logic is processed and before the load balancer forwards to the upstream.

kubectl apply -f - <<EOF
kind: Plugin
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-injector
spec:
  priority: 115
  pipyscript: |+
    (
    pipy({
      _pluginName: '',
      _pluginConfig: null,
      _accessToken: null,
    })
    .import({
        __service: 'outbound-http-routing',
    })
    .pipeline()
    .onStart(
        () => void (
            _pluginName = __filename.slice(9, -3),
            _pluginConfig = __service?.Plugins?.[_pluginName],
            _accessToken = _pluginConfig?.AccessToken
        )
    )
    .handleMessageStart(
        msg => _accessToken && (msg.head.headers['x-iam-token'] = _accessToken)
    )
    .chain()
    )
EOF

Plugin token-verifier

kubectl apply -f - <<EOF
kind: Plugin
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-verifier
spec:
  priority: 115
  pipyscript: |+
    (
    pipy({
        _pluginName: '',
        _pluginConfig: null,
        _verifier: null,
        _authPaths: null,
        _authRequred: false,
        _authSuccess: undefined,
    })
    .import({
        __service: 'inbound-http-routing',
    })
    .pipeline()
    .onStart(
        () => void (
            _pluginName = __filename.slice(9, -3),
            _pluginConfig = __service?.Plugins?.[_pluginName],
            _verifier = _pluginConfig?.Verifier,
            _authPaths = _pluginConfig?.Paths && _pluginConfig.Paths?.length > 0 && (
              new algo.URLRouter(Object.fromEntries(_pluginConfig.Paths.map(path => [path, true])))
            )
        )
    )
    .handleMessageStart(
        msg => _authRequred = (_verifier && _authPaths?.find(msg.head.headers.host, msg.head.path))
    )
    .branch(
        () => _authRequred, (
        $ => $
          .fork().to($ => $
            .muxHTTP().to($ => $.connect(()=> _verifier))
            .handleMessageStart(
              msg => _authSuccess = (msg.head.status == 200)
            )
          )
          .wait(() => _authSuccess !== undefined)
          .branch(() => _authSuccess, $ => $.chain(),
            $ => $.replaceMessage(
              () => new Message({ status: 401 }, 'Unauthorized!')
            )
          )
      ),
        $ => $.chain()
      )
    )
EOF

Setting up plugin-chain

plugin chain token-injector-chain:

metadata.name: name of plugin chain resource token-injector-chain
spec.chains
- name: name of the plugin chain, one of the 4 plugin chains, here it is outbound-http which is the HTTP protocol processing stage for outbound traffic.
- plugins: list of plugins to be inserted into the plugin chain, here token-injector is inserted into the plugin chain.
spec.selectors: target of the plugin chain, using Kubernetes label selector scheme.
- podSelector: pod selector, selects pods with label app=curl.
- namespaceSelector: namespace selector, selects namespaces managed by the mesh, i.e., flomesh.io/monitored-by=fsm.

kubectl apply -n curl -f - <<EOF
kind: PluginChain
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-injector-chain
spec:
  chains:
    - name: outbound-http
      plugins:
        - token-injector
  selectors:
    podSelector:
      matchLabels:
        app: curl
      matchExpressions:
        - key: app
          operator: In
          values: ["curl"]
    namespaceSelector:
      matchExpressions:
        - key: flomesh.io/monitored-by
          operator: In
          values: ["fsm"]
EOF

plugin chain token-verifier-chain:

kubectl apply -n httpbin -f - <<EOF
kind: PluginChain
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-verifier-chain
spec:
  chains:
    - name: inbound-http
      plugins:
        - token-verifier
  selectors:
    podSelector:
      matchLabels:
        app: httpbin
    namespaceSelector:
      matchExpressions:
        - key: flomesh.io/monitored-by
          operator: In
          values: ["fsm"]
EOF

After applying the plugin chain configuration, the plugin chains of the two applications can be viewed now. From the results, we can see the two plugins located in the plugins directory. Our declared plugins have been configured in the two applications through the configuration of the plugin chain.

fsm proxy get config_dump -n curl $curl_pod | jq '.Chains."outbound-http"'
[
  "modules/outbound-http-routing.js",
  "modules/outbound-metrics-http.js",
  "modules/outbound-tracing-http.js",
  "modules/outbound-logging-http.js",
  "modules/outbound-circuit-breaker.js",
  "plugins/token-injector.js",
  "modules/outbound-http-load-balancing.js",
  "modules/outbound-http-default.js"
]

fsm proxy get config_dump -n httpbin $httpbin_pod | jq '.Chains."inbound-http"'
[
  "modules/inbound-tls-termination.js",
  "modules/inbound-http-routing.js",
  "modules/inbound-metrics-http.js",
  "modules/inbound-tracing-http.js",
  "modules/inbound-logging-http.js",
  "modules/inbound-throttle-service.js",
  "modules/inbound-throttle-route.js",
  "plugins/token-verifier.js",
  "modules/inbound-http-load-balancing.js",
  "modules/inbound-http-default.js"
]

After applying the plugin configuration, but we haven’t yet make changes to plugin configuration, the application curl can still access httpbin.

kubectl exec $curl_pod -n curl -c curl -- curl -Is http://httpbin.httpbin:14001/get
HTTP/1.1 200 OK
server: gunicorn/19.9.0
date: Sun, 05 Feb 2023 06:34:33 GMT
content-type: application/json
content-length: 304
access-control-allow-origin: *
access-control-allow-credentials: true
connection: keep-alive

Setting up plugin configuration

We will first apply the configuration of the plugin token-verifier. Here, the authentication service ext-auth.auth:8079 and the request /get that needs to be authenticated are configured.

spec.config contains the contents of the plugin configuration, which will be converted to JSON format. For example, the configuration applied to the token-verifier plugin exists in the following JSON form:
```
{
  "Plugins": {
    "token-verifier": {
      "Paths": [
        "/get"
      ],
      "Verifier": "ext-auth.auth:8079"
    }
  }
}
```

kubectl apply -n httpbin -f - <<EOF
kind: PluginConfig
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-verifier-config
spec:
  config:
    Verifier: 'ext-auth.auth:8079'
    Paths:
      - "/get"
  plugin: token-verifier
  destinationRefs:
    - kind: Service
      name: httpbin
      namespace: httpbin
EOF

At this time, the application curl cannot access the httbin /get path because a access token has not been configured for curl yet.

kubectl exec $curl_pod -n curl -c curl -- curl -Is http://httpbin.httpbin:14001/get
HTTP/1.1 401 Unauthorized
content-length: 13
connection: keep-alive

But accessing /headers path doesn’t require any authentication.

kubectl exec $curl_pod -n curl -c curl -- curl -Is http://httpbin.httpbin:14001/headers
HTTP/1.1 200 OK
server: gunicorn/19.9.0
date: Sun, 05 Feb 2023 06:37:05 GMT
content-type: application/json
content-length: 217
access-control-allow-origin: *
access-control-allow-credentials: true
connection: keep-alive

Next, the configuration of the plugin token-injector is applied to configure the access token 2f1acc6c3a606b082e5eef5e54414ffb for the application’s requests. This token is also a valid token hardcoded in the authentication service.

kubectl apply -n curl -f - <<EOF
kind: PluginConfig
apiVersion: plugin.flomesh.io/v1alpha1
metadata:
  name: token-injector-config
spec:
  config:
    AccessToken: '2f1acc6c3a606b082e5eef5e54414ffb'
  plugin: token-injector
  destinationRefs:
    - kind: Service
      name: httpbin
      namespace: httpbin
EOF

After applying the configuration for the token-injector plugin, the requests from the curl application will now have the access token 2f1acc6c3a606b082e5eef5e54414ffb configured. As a result, when accessing the /get path of httpbin, the requests will pass authentication and be accepted by httpbin.

kubectl exec $curl_pod -n curl -c curl -- curl -Is http://httpbin.httpbin:14001/get
HTTP/1.1 200 OK
server: gunicorn/19.9.0
date: Sun, 05 Feb 2023 06:39:54 GMT
content-type: application/json
content-length: 360
access-control-allow-origin: *
access-control-allow-credentials: true
connection: keep-alive

Feedback

Was this page helpful?

Glad to hear it! Please tell us how we can improve.

Sorry to hear that. Please tell us how we can improve.

Last modified April 11, 2024: update versions of fsm and pipy (cea5b3e)