Manage ONAP Microservices with Istio Service Mesh-Mutual TLS Authentication Enabled
Introduction
In Casablanca release, MSB project is integrating Istio Service Mesh with ONAP to manage ONAP microservices. Istio Service Mesh is a dedicated infrastructure layer to connect, manage and secure microservices, which brings the below benefits:
- Stability and Reliability: Reliable communication with retries and circuit breaker
- Security: Secured communication with TLS
- Performance: Latency aware load balancing with warm cache
- Observability: Metrics measurement and distributed tracing without instrumenting application
- Manageability: Routing rule and rate limiting enforcement
- Testability: Fault injection to test resilience of the services
Installation
Download installation scripts from ONAP Gerrit:
git clone https://gerrit.onap.org/r/msb/service-mesh
Kubernetes Master
We need Kubernetes1.9 or newer to enable automatic sidecar injection, so we don't have to modify every individual ONAP kubernetes yaml deployment files to add the sidecar container, which would be inconvenient.
Istio leverages the webhook feature of Kubernetes to automatically inject an Envoy sidecar to each Pod. Kubernetes API server will call the Istio sidecar injection webhook when it receives a request to create a Pod resource, the webhook adds an Envoy sidecar container to the Pod, then the modified Pod resource is stored into etcd.
Webhook and other needed features have already been configured in the install scripts to enable Istio sidecar injection.
Create the Kubernetes master by running this script:
cd service-mesh/install/ ./1_install_k8s_master.sh
This script will create a Kubernetes master node with Kubeadm and install calico network plugin. Some other needed tools such as Docker, Kubectl and Helm will be installed as well.
From the output of the script, you should see a command on how to join a node to the created Kubernets cluster. Note that this is an example, the token and cert-hash of your installation will be different, please copy & paste the command to somewhere, we will need it later.
You can now join any number of machines by running the following on each node as root: kubeadm join 10.12.5.104:6443 --token 1x62yf.60ys5p2iw13tx2t8 --discovery-token-ca-cert-hash sha256:f06628c7cee002b262e69f3f9efadf47bdec125e19606ebff743a3e514a8383b
Kubernetes worker Node
Log in the worker node machine, run this script to create a kubernetes worker node:
./2_install_k8s_minion.sh
You can now join this machines by running "kubeadmin join" command as root:
sudo kubeadm join 10.12.5.104:6443 --token 1x62yf.60ys5p2iw13tx2t8 --discovery-token-ca-cert-hash sha256:f06628c7cee002b262e69f3f9efadf47bdec125e19606ebff743a3e514a8383b
Please note that this is just an example, please refer to the output of the "kubeamin init" when creating the k8s master for the exact command to use in your k8s cluster.
If you would like to get kubectl talk to your k8s master, you need to copy the administrator kubeconfig file from your master to your workstation like this:
scp root@<master ip>:/etc/kubernetes/admin.conf . kubectl --kubeconfig ./admin.conf get nodes
or you can manually copy the content of this file to ~/.kube/conf if scp can't be used due to security reason.
Istio Control Plane with Mutual TLS
Install Istio by running this script:
./ 3_install_istio_with_auth.sh
This script installs the followings Istio components:
- Install Istioctl command line tool in the /usr/bin directory
- Install Istio control plane components, including Pilot, Citadel, Mixer
- Install addons including servicegraph, Promeheus, Grafana, jaeger
Confirm Istio was installed:
kubectl get svc -n istio-system NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE grafana NodePort 10.109.190.71 <none> 3000:30300/TCP 20m istio-citadel ClusterIP 10.106.185.181 <none> 8060/TCP,9093/TCP 20m istio-egressgateway ClusterIP 10.102.224.133 <none> 80/TCP,443/TCP 20m istio-ingressgateway LoadBalancer 10.100.168.32 <pending> 80:31380/TCP,443:31390/TCP,31400:31400/TCP 20m istio-pilot ClusterIP 10.101.64.153 <none> 15003/TCP,15005/TCP,15007/TCP,15010/TCP,15011/TCP,8080/TCP,9093/TCP 20m istio-policy ClusterIP 10.104.11.162 <none> 9091/TCP,15004/TCP,9093/TCP 20m istio-sidecar-injector ClusterIP 10.100.229.40 <none> 443/TCP 20m istio-statsd-prom-bridge ClusterIP 10.107.27.91 <none> 9102/TCP,9125/UDP 20m istio-telemetry ClusterIP 10.101.153.114 <none> 9091/TCP,15004/TCP,9093/TCP,42422/TCP 20m prometheus ClusterIP 10.103.0.205 <none> 9090/TCP 20m servicegraph NodePort 10.106.49.168 <none> 8088:30088/TCP 20m tracing LoadBalancer 10.100.158.236 <pending> 80:30188/TCP 20m zipkin NodePort 10.96.164.255 <none> 9411:30411/TCP 20m
Sidecar Injection
In the transition phase, the Istio sidecar injector policy is configured as "disabled" when installing Istio. So the sidecar injector will not inject the sidecar into pods by default. Add the `sidecar.istio.io/inject annotation` with value `true` to the pod template spec to enable injection.
Example:
apiVersion: extensions/v1beta1 kind: Deployment metadata: name: {{ include "common.fullname" . }} namespace: {{ include "common.namespace" . }} labels: app: {{ include "common.name" . }} chart: {{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }} release: {{ .Release.Name }} heritage: {{ .Release.Service }} spec: replicas: {{ .Values.replicaCount }} selector: matchLabels: app: multicloud-vio template: metadata: labels: app: {{ include "common.name" . }} release: {{ .Release.Name }} name: {{ include "common.name" . }} annotations: sidecar.istio.io/inject: "{{.Values.istioSidecar}}"
Note: when all ONAP projects are ready for Istio integration, the Istio sidecar injector policy could be configured as "enabled", then the annotation in the pod will not be necessary any more.
Enable Istio sidecar injection webhook.
kubectl create namespace onap kubectl label namespace onap istio-injection=enabled
Confirm that auto sidecar injection has been enabled on onap namespace.
kubectl get namespace -L istio-injection NAME STATUS AGE ISTIO-INJECTION default Active 20m istio-system Active 10m kube-public Active 20m kube-system Active 20m onap Active 8s enabled
Start a local helm repository server and add it to helm repository list:
helm serve & helm repo add local http://127.0.0.1:8879
Download OOM Gerrit repository and build the helm charts.
git clone -b beijing http://gerrit.onap.org/r/oom cd oom/kubernetes make all
Confirm that ONAP charts have been successfully created.
helm search onap NAME CHART VERSION APP VERSION DESCRIPTION local/onap 2.0.0 beijing Open Network Automation Platform (ONAP) local/aaf 2.0.0 ONAP Application Authorization Framework local/aai 2.0.0 ONAP Active and Available Inventory local/clamp 2.0.0 ONAP Clamp local/cli 2.0.0 ONAP Command Line Interface local/consul 2.0.0 ONAP Consul Agent local/dcaegen2 2.0.0 ONAP DCAE Gen2 local/dmaap 2.0.0 ONAP DMaaP components local/esr 2.0.0 ONAP External System Register local/log 2.0.0 ONAP Logging ElasticStack local/msb 2.0.0 ONAP MicroServices Bus local/multicloud 2.0.0 ONAP multicloud broker local/nbi 2.0.0 ONAP Northbound Interface local/oof 2.0.0 ONAP Optimization Framework local/policy 2.0.0 ONAP Policy Administration Point local/portal 2.0.0 ONAP Web Portal local/postgres 2.0.0 ONAP Postgres Server local/robot 2.0.0 A helm Chart for kubernetes-ONAP Robot local/sdnc-prom 2.0.0 ONAP SDNC Policy Driven Ownership Management local/sniro-emulator 2.0.0 ONAP Mock Sniro Emulator local/so 2.0.0 ONAP Service Orchestrator local/uui 2.0.0 ONAP uui local/vfc 2.0.0 ONAP Virtual Function Controller (VF-C) local/vid 2.0.0 ONAP Virtual Infrastructure Deployment local/vnfsdk 2.0.0 ONAP VNF SDK
Install local/onap chart. Local/onap chart will do some initialization setup which is needed for onap components, such as creating service accounts.
cd oom/kubernetes helm install local/onap -n common --namespace onap -f onap/resources/environments/disable-allcharts.yaml
In Casablanca, MSB project is working with VF-C and MultiCloud as pilot projects, we would like to roll out it to the other ONAP projects after verifying the integration and Istio features.
helm install local/msb -n msb --namespace onap helm install local/multicloud -n multicloud --namespace onap --set liveness.enabled=false,multicloud-ocata.liveness.enabled=false,multicloud-vio.liveness.enabled=false,multicloud-windriver.liveness.enabled=false helm install local/multicloud -n multicloud --namespace onap
Note:
- The current version of Istio mutual TLS authentication can't work with kubernetes liveness probe, Istio is working on a long-term fix to solve this problem. A simple workaround for the time being is to disable liveness probe by passing a 'liveness.enabled=false' value to helm install command.
- You can also install other ONAP projects with helm install if they are needed. But Istio sidecar will not be injected to their Pods by default.
Confirm that ONAP microservices have been started
kubectl get all -n onap NAME READY STATUS RESTARTS AGE pod/msb-kube2msb-77ccb675dd-rhfn7 1/1 Running 0 3h pod/msb-msb-consul-646987f5cf-qms5v 2/2 Running 0 3h pod/msb-msb-discovery-7647f6476f-cl6xw 3/3 Running 0 3h pod/msb-msb-eag-d678c65d6-fmfn6 3/3 Running 0 3h pod/msb-msb-iag-647d5f998c-dc766 3/3 Running 0 3h pod/multicloud-multicloud-5679bd9876-tzxzw 2/2 Running 0 1h pod/multicloud-multicloud-ocata-774579596-f7smf 3/3 Running 0 1h pod/multicloud-multicloud-vio-8c7dbc8d5-lfcw6 3/3 Running 0 1h pod/multicloud-multicloud-windriver-85b595675d-5vx45 3/3 Running 0 1h pod/vfc-vfc-catalog-79764dfd8f-rkx6f 2/2 Running 1 2d pod/vfc-vfc-ems-driver-75bc68b946-6r6r6 1/1 Running 1 2d pod/vfc-vfc-generic-vnfm-driver-69bf778bfd-pscjn 2/2 Running 0 2d pod/vfc-vfc-huawei-vnfm-driver-8574569f4c-8jwc4 2/2 Running 1 2d pod/vfc-vfc-juju-vnfm-driver-6dfd876bb8-bh7dq 2/2 Running 0 2d pod/vfc-vfc-multivim-proxy-58c7bd47dc-7qdtd 1/1 Running 0 2d pod/vfc-vfc-nokia-v2vnfm-driver-7b77c469bd-krfrw 1/1 Running 0 2d pod/vfc-vfc-nokia-vnfm-driver-98fbdb5b5-p9zqw 2/2 Running 0 2d pod/vfc-vfc-nslcm-74956bb876-v9kbt 2/2 Running 0 2d pod/vfc-vfc-resmgr-57dc4c98b5-dzp7f 2/2 Running 0 2d pod/vfc-vfc-vnflcm-6f9dc7df44-hncf4 2/2 Running 1 2d pod/vfc-vfc-vnfmgr-5585c688c6-7qrnp 2/2 Running 0 2d pod/vfc-vfc-vnfres-54bc985599-9zkqn 2/2 Running 0 2d pod/vfc-vfc-workflow-6db56f95b9-np8tg 1/1 Running 1 2d pod/vfc-vfc-workflow-engine-7fb49fd974-kcb8q 1/1 Running 1 2d pod/vfc-vfc-zte-sdnc-driver-585d449797-87nhp 1/1 Running 0 2d pod/vfc-vfc-zte-vnfm-driver-59d4756fbc-rpn9v 2/2 Running 0 2d NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/msb-consul NodePort 10.96.255.198 <none> 8500:30285/TCP 3h service/msb-discovery NodePort 10.105.163.81 <none> 10081:30281/TCP 3h service/msb-eag NodePort 10.100.221.66 <none> 80:30282/TCP,443:30284/TCP 3h service/msb-iag NodePort 10.96.179.117 <none> 80:30280/TCP,443:30283/TCP 3h service/multicloud NodePort 10.102.72.237 <none> 9001:30291/TCP 1h service/multicloud-ocata NodePort 10.99.131.129 <none> 9006:30293/TCP 1h service/multicloud-vio NodePort 10.111.175.58 <none> 9004:30292/TCP 1h service/multicloud-windriver NodePort 10.110.92.61 <none> 9005:30294/TCP 1h service/vfc-catalog ClusterIP 10.99.98.115 <none> 8806/TCP 2d service/vfc-ems-driver ClusterIP 10.96.189.14 <none> 8206/TCP 2d service/vfc-generic-vnfm-driver ClusterIP 10.109.48.184 <none> 8484/TCP 2d service/vfc-huawei-vnfm-driver ClusterIP 10.104.208.38 <none> 8482/TCP,8483/TCP 2d service/vfc-juju-vnfm-driver ClusterIP 10.96.182.14 <none> 8483/TCP 2d service/vfc-multivim-proxy ClusterIP 10.107.106.216 <none> 8481/TCP 2d service/vfc-nokia-v2vnfm-driver ClusterIP 10.107.12.32 <none> 8089/TCP 2d service/vfc-nokia-vnfm-driver ClusterIP 10.102.179.150 <none> 8486/TCP 2d service/vfc-nslcm ClusterIP 10.106.43.164 <none> 8403/TCP 2d service/vfc-resmgr ClusterIP 10.98.174.184 <none> 8480/TCP 2d service/vfc-vnflcm ClusterIP 10.108.132.123 <none> 8801/TCP 2d service/vfc-vnfmgr ClusterIP 10.108.59.102 <none> 8803/TCP 2d service/vfc-vnfres ClusterIP 10.111.85.161 <none> 8802/TCP 2d service/vfc-workflow ClusterIP 10.97.184.206 <none> 10550/TCP 2d service/vfc-workflow-engine ClusterIP 10.109.175.61 <none> 8080/TCP 2 service/vfc-zte-sdnc-driver ClusterIP 10.103.94.142 <none> 8411/TCP 2d service/vfc-zte-vnfm-driver ClusterIP 10.108.146.237 <none> 8410/TCP 2d
Create an Istio Gateway so we can access the MSB portal out of the Mesh
cd service-mesh/install/ kubectl apply -f msb-istio-gateway.yaml -n onap
Now you can open the MSB portal http://${INGRESS_IP}:${INGRESS_PORT}/msb in the browser to see all the registered services.
Note: INGRESS_IP and INGRESS_PORT can be found by executing this command 'kubectl get svc istio-ingressgateway -n istio-system'
Explore Istio Features
Distributed Tracing
First, let's generate some traffics in the application, access the following URLs with curl command or open them in the browser
http://${INGRESS_IP}:${INGRESS_PORT}/api/multicloud-vio/v0/swagger.json
http://${INGRESS_IP}:${INGRESS_PORT}/api/multicloud-ocata/v0/swagger.json
Then open your browser at http://tracing_node_ip:tracing_node_port/, you should see something similar to the following:
Note
- Tracing_node_port can be found by 'kubctl get svc -n istio-system'.
- ONAP microservices need to propagate the appropriate HTTP headers so that when the proxies send span information, the spans can be correlated correctly into a single trace.
Service Graph
Istio provides a Servicegraph service which generates and visualizes graph representations of the services in the mesh.
Open your browser at http://node_ip:30088/dotviz or http://node_ip:30088/force/forcegraph.html, you should see the service graph:
Metrics Visualization
Istio automatically gathers telemetry for services in a mesh. A Prometheus adapter is plugged into Mixer to serve the generated metric data. A Grafana addon is pre-configured with a Prometheus data source and has an Istio dashboard installed for the metric visualization.
Open your browser at http://node_ip:30300, you should see the Grafana Istio dashboard:
Authorization(RBAC)
Istio authorization is disabled by default, running the following command to enable it for onap namespace:
cd /service-mesh/install kubectl apply -f enable-istio-rbac.yaml
Point your browser at the msb portal or multicloud swagger file:
http://${INGRESS_IP}:${INGRESS_PORT}/msb
http://${INGRESS_IP}:${INGRESS_PORT}/api/multicloud-vio/v0/swagger.json
Now you should see "RBAC: access denied". This is because Istio authorization is “deny by default”, which means that you need to explicitly define access control policy to grant access to any service. Note: There may be some delays due to caching and other propagation overhead.
Running the following command to allow Istio Ingress gateway read access to onap Namespace:
cd /service-mesh/install kubectl apply -f apply -f istio-ingress-rbac.yaml
Now if you point your browser at the msb portal (http://${INGRESS_IP}:${INGRESS_PORT}/msb). You should see the msb portal page with registered services.
Note: There may be some delays due to caching and other propagation overhead.
If you try to access http://${INGRESS_IP}:${INGRESS_PORT}/api/multicloud-vio/v0/swagger.json, you should still see "RBAC: access denied". What's happening? This request actually goes through browser->Istio-ingress->MSB->multicloud, even now Istio-ingress can access MSB, MSB is not allowed to access multicloud. So we need to create another RBAC rule to grant the access permission of multicloud to MSB.
Running the following command to grant access of multicloud to MSB:
cd /service-mesh/install kubectl apply -f msb-rbac.yaml
Try to access http://${INGRESS_IP}:${INGRESS_PORT}/api/multicloud-vio/v0/swagger.json again, you should be able to see the swagger file return from multicloud microservice.
Note: There may be some delays due to caching and other propagation overhead.
Service Mesh Migration
Without Istio Authentication and Authorization
ONAP can be easily integrated with Istio service mesh if Istio Auth is disabled. In that case, ONAP can leverage the traffic management, telemetry and policies capabilities of Istio to connect, control and observe ONAP microservies, but without Mutual TLS authentication and authorization.
Though ONAP services can talk to each other within the mesh, to maximize the benefits brought by Istio, we still need to make little compatible changes to the existing services:
Service Port Name
The port names must be of the form protocol
-suffix
with http, http2, grpc, mongo, or redis as the protocol
in order to take advantage of Istio’s routing features.
For example, name: http2-foo
or name: http
are valid port names, but name: http2foo
is not. If the port name does not begin with a recognized prefix or if the port is unnamed, traffic on the port will be treated as plain TCP traffic (unless the port explicitly uses Protocol: UDP to signify a UDP port).
kubectl describe svc aai -n onap Name: aai Namespace: onap Labels: app=aai chart=aai-2.0.0 heritage=Tiller release=aai1 Annotations: <none> Selector: app=aai Type: NodePort IP: 10.96.29.203 Port: http-aai 8080/TCP ---omitted for brevity
Propagate Http Header for Distributed Tracing
Istio uses HTTP headers to record the request tracing information across multiple spans. Although Istio proxies are able to automatically send all the spans to Mixer, they need some hints to tie together the individual spans to get the entire trace.
To do this, ONAP microservies needs to collect and propagate the following headers from the incoming request to any outgoing requests:
x-request-id
x-b3-traceid
x-b3-spanid
x-b3-parentspanid
x-b3-sampled
x-b3-flags
x-ot-span-context
With Istio Authentication and Authorization
In addition to the port name format and http header propagation, the followings need to be done to leverage Istio auth.
Liveness probe
Mutual TLS can't work with 8Shttp/tcp liveness probe. If mutual TLS is enabled, http and tcp health checks from the kubelet will not work since they do not have Istio-issued certs. The workaround is using liveness command instead or disabling http and tcp liveness probe for the time being.
Allow both Mutual TLS and Plain Traffic
During the migration, we can use “PERMISSIVE” mode of Istio Auth policy to allow both TLS and plain traffic. After migration is done, the mode can be switched to "STRICT" mode so only TLS traffics are permitted to access services.
cat <<EOF | kubectl apply -n onap -f - apiVersion: "authentication.istio.io/v1alpha1" kind: "Policy" metadata: name: "default" namespace: onap spec: peers: - mtls: mode: PERMISSIVE EOF
In that case, the RBAC should be set to allow all users, including the unauthenticated users, to access the services.
cat <<EOF | kubectl apply -n onap -f - apiVersion: "rbac.istio.io/v1alpha1" kind: ServiceRole metadata: name: onap-default namespace: onap spec: rules: - services: ["*"] methods: ["*"] --- apiVersion: "rbac.istio.io/v1alpha1" kind: ServiceRoleBinding metadata: name: bind-service-default namespace: onap spec: subjects: - user: "*" roleRef: kind: ServiceRole name: "onap-default" EOF
By this approach, ONAP can be smoothly migrated to Istio with auth enabled. After every ONAP microservice adopts Istio auth, then we can set the authentication to "STRICT" mode and enforce strict access control per the needs of each service.
What's the next? we will provide a user-friendly Istio UI to manage Istio rules and policies. Comment here to leave your thoughts or join our weekly project meeting if you're interested.