Kuberentes applications with Istio Service mesh, MetalLB load balancer, and Datadog monitoring

This repository showcases Kubernetes microservices application with Istio service mesh enhancing Istio traffic management and security, MetalLB load balancing and Datadog monitoring & observability covering Kubernetes infrastructure monitoring, application performance monitoring, metrics, distributed traces, code profilers, and logs management

Install Istio

Install Istio with minimal profile, configuration profiles to apply a customized Istio Ingress Gateway which i included in kubernetes-manifests/istio-manifests.yaml file

istioctl install --set profile=minimal -y
# output
✔ Istio core installed
✔ Istiod installed
✔ Installation complete

Create namespace for our application apps and add a namespace label to instruct Istio to automatically inject Envoy sidecar proxies into its pods,

these Envoy Proxy sidecars will intercept service-to-service communication and provide traffic management, such as routing rules, load balancing, retries, circuit breakers, timeouts, failovers, ... additionally traffic will be secured with mTLS encryption

kubectl create namespace apps
kubectl label namespace apps istio-injection=enabled

# or add label `sidecare.istio.io/inject: "true"` in the pod template

Deploy Application

the application is a GO API service, text-to-speech service and a client,

the Go API uses Echo framework to perform CRUD operation, additionally it has /healthz route performing Kubernetes health probes, to check the service replicas' readiness and liveness.

the text-to-speech service will communicate with Go api covering its response into audio file, i also configured its Kubernetes manifests to have a presistent-volume attached, the text-to-speech service has two versions, v1 i made with Go and v2 as a canary version with Python, i applied istio canary deployments to send 70% of the requests to the Golang version v1 and 30% of requests to the canary version v2 python, and i'll be observing the two versions' performance later with Datadog APM.

# apply the application's manifests 

kubectl apply -f kubernetes-manifests/go-api.yaml
kubectl apply -f kubernetes-manifests/client.yaml
kubectl apply -f cluster/secret.yaml
kubectl apply -f kubernetes-manifests/persistent-volume.yaml
kubectl apply -f kubernetes-manifests/text-to-speech.yaml
kubectl apply -f kubernetes-manifests/text-to-speech-canary.yaml

kubectl -n apps get all
# we can see in the output that each pod has 2 containers, one is the application's container and the other is th envoy proxy sidecar container which got injected by Istio
# output
NAME                                         READY   STATUS    RESTARTS      AGE
pod/alpine-78d69bd9d-6d7g9                   2/2     Running   0             8m48s
pod/goapi-7769bd9d6f-75tbs                   2/2     Running   0             8m49s
pod/goapi-7769bd9d6f-vkrpk                   2/2     Running   0             8m49s
pod/goapi-7769bd9d6f-zwqsb                   2/2     Running   0             8m49s
pod/text-to-speech-6d886cc4fc-mql6x          2/2     Running   1 (48s ago)   58s
pod/text-to-speech-canary-7f89f6ffb5-pbq27   2/2     Running   0             52s

NAME                     TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
service/alpine           ClusterIP   10.96.40.164    <none>        5000/TCP   8m49s
service/goapi            ClusterIP   10.98.217.225   <none>        80/TCP     8m49s
service/text-to-speech   ClusterIP   10.99.196.128   <none>        80/TCP     58s

NAME                                    READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/alpine                  1/1     1            1           8m49s
deployment.apps/goapi                   3/3     3            3           8m49s
deployment.apps/text-to-speech          1/1     1            1           58s
deployment.apps/text-to-speech-canary   1/1     1            1           52s

NAME                                               DESIRED   CURRENT   READY   AGE
replicaset.apps/alpine-78d69bd9d                   1         1         1       8m49s
replicaset.apps/goapi-7769bd9d6f                   3         3         3       8m49s
replicaset.apps/text-to-speech-6d886cc4fc          1         1         1       58s
replicaset.apps/text-to-speech-canary-7f89f6ffb5   1         1         1       52s

Apply Istio Ingress Gateway and Istio traffic management resources

apply a custom istio-ingressgateway and traffic managemt resources such as virtual services, destignation rules which i included in kubernests-manifests/istio-manifests.yaml file

# first i'll untaint the control plane node as i configured it to have port-mapping and `istio-ingressgateway` pods with node affinity to get scheduled on it
kubectl taint nodes --all node-role.kubernetes.io/control-plane-
# output
node/cluster-istio-control-plane untainted

# apply istio-manifests.yaml file
kubectl apply -f kubernetes-manifests/istio-manifests.yaml
# output
service/istio-ingressgateway created
deployment.apps/istio-ingressgateway created
role.rbac.authorization.k8s.io/istio-ingressgateway-sds created
rolebinding.rbac.authorization.k8s.io/istio-ingressgateway-sds created
gateway.networking.istio.io/gateway created
virtualservice.networking.istio.io/goapi-vs created
virtualservice.networking.istio.io/alpine-vs created
virtualservice.networking.istio.io/text-to-speech-vs created
destinationrule.networking.istio.io/text-to-speech-destination-rule created

# check the resources created in `istio-system` namespace
kubectl -n istio-system get all
# output
NAME                                        READY   STATUS    RESTARTS   AGE
pod/istio-ingressgateway-6c8ffd96dc-gq45b   1/1     Running   0          71s
pod/istiod-7f69bdfd9-zt2fd                  1/1     Running   0          24m

NAME                           TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)           
                      AGE
service/istio-ingressgateway   LoadBalancer   10.104.228.63   <pending>     80:31458/TCP,443:31299/TCP              71s
service/istiod                 ClusterIP      10.105.90.96    <none>        15010/TCP,15012/TCP,443/TCP,15014/TCP   24m

NAME                                   READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/istio-ingressgateway   1/1     1            1           71s
deployment.apps/istiod                 1/1     1            1           24m

NAME                                              DESIRED   CURRENT   READY   AGE
replicaset.apps/istio-ingressgateway-6c8ffd96dc   1         1         1       71s
replicaset.apps/istiod-7f69bdfd9                  1         1         1       24m

NAME                                         REFERENCE           TARGETS         MINPODS   MAXPODS   REPLICAS   AGE
horizontalpodautoscaler.autoscaling/istiod   Deployment/istiod   <unknown>/80%   1         5         1          24m

we can see that the EXTERNAL-IP for service/istio-ingressgateway typeLoadBalancer is pending, for that we need to install MetalLB which is a load balancer used for on-prem Kubernetes, MetalLB will allocate an external IP address for Istio Ingress Gateway

we also applied in same kubernetes-manifests/istio-manifests.yaml file Istio virtual services for each service in the apps namespace, and an Istio Gateway which is assigned to goapi service

# check istio virtual services created in `apps` namespace 
kubectl -n apps get virtualService
# output
NAME                GATEWAYS                   HOSTS                                       AGE
alpine-vs                                      ["alpine.apps.svc.cluster.local"]           74s
goapi-vs            ["istio-system/gateway"]   ["goapi.apps.svc.cluster.local"]            74s
text-to-speech-vs                              ["text-to-speech.apps.svc.cluster.local"]   74s

# check Kubernetes resouce kind: Gateway created in `istio-system` namespace
kubectl -n istio-system get gateway
#output
NAME      AGE
gateway   82s

apply Istio Peer Authentication resource which will force mTLS traffic for all workloads in the application namespace

kubectl apply -f kubernetes-manifests/istio-peer-authentication.yaml

Install MetalLB

MetalLB is a load-balancer implementation for bare metal/on-prem Kubernetes clusters

Preparation : If your cluster is using kube-proxy in IPVS mode, you have to enable strict ARP mode You can achieve this by editing kube-proxy config in the cluster

kubectl get configmap kube-proxy -n kube-system -o yaml | \
sed -e "s/strictARP: false/strictARP: true/" | \
kubectl apply -f - -n kube-system
# output
configmap/kube-proxy configured

To install MetalLB, apply the manifest

kubectl apply -f https://raw.githubusercontent.com/metallb/metallb/v0.13.12/config/manifests/metallb-native.yaml

wait untill metallb-system/speaker daemonsets are running, then apply MetalLB IPAddressPool and L2Advertisement which i configured the load balancer ip-address to be advertised in

kubectl apply -f kubernetes-manifests/metallb-config.yaml

# output
ipaddresspool.metallb.io/ip-pool created
l2advertisement.metallb.io/metallb-ip-adv created

if we check istio-ingressgateway LoadBalancer service now, it has an external ip assigned, 172.18.0.10, which we can use now to call goapi service

kubectl -n istio-system get svc/istio-ingressgateway -owide

# output
NAME                   TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)                   
   AGE   SELECTOR
istio-ingressgateway   LoadBalancer   10.104.228.63   172.18.0.10   80:31458/TCP,443:31299/TCP   18m   istio=ingressgateway

export GATEWAY_IP=$(kubectl -n istio-system get svc istio-ingressgateway -o jsonpath="{.status.loadBalancer.ingress[0].ip}")
echo $GATEWAY_IP
# output
172.18.0.10

i'll exec into the client pod to create intries for the Go books API using curl

kubectl -n apps exec -it pod/alpine-78d69bd9d-6d7g9 -- sh

/app $ curl -H "Content-Type: application/json" -d '{ "author": "occupytheweb", "title": "Network Basics for Hackers"}' -X POST http://goapi.apps.svc.cluster.local/books
{"id":1,"title":"Network Basics for Hackers","author":"occupytheweb"}

/app $ curl -H "Content-Type: application/json" -d '{ "title": "Build Systems with Go", "author": "Juan Tirado"}' -X POST http://goapi.apps.svc.cluster.local/books
{"id":2,"title":"Build Systems with Go","author":"Juan Tirado"}

/app $ curl -v http://goapi.apps.svc.cluster.local/books
# output
*   Trying 172.18.0.10:80...
* Connected to goapi.apps.svc.cluster.local (172.18.0.10) port 80
> GET /books HTTP/1.1
> Host: goapi.apps.svc.cluster.local
> User-Agent: curl/8.4.0
> Accept: */*
>
< HTTP/1.1 200 OK
< content-type: application/json; charset=UTF-8
< date: Sat, 02 Dec 2023 05:13:28 GMT
< content-length: 73
< x-envoy-upstream-service-time: 2
< server: envoy
<
{"1":{"id":1,"title":"Network Basics for Hackers","author":"occupytheweb"},"2":{"id":2,"title":"Build Systems with Go","author":"Juan Tirado"}}
* Connection #0 to host goapi.apps.svc.cluster.local left intact

we also can check Istio canary deployments implementation but sending requests to text-to-speech service, we an see that 70% of requests are sent to the Go version v1, and 30% of requests are sent to the Python canary verion v2

kubectl -n apps exec -it pod/alpine-78d69bd9d-6d7g9 -- sh

# `70%` of requests are sent to the Go version v1, and `30%` of requests are sent to the Python canary verion v2
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
TTS canary version (Python)!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
TTS canary version (Python)!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
text-to-peech Go!
/app $ curl http://text-to-speech.apps.svc.cluster.local/
TTS canary version (Python)!

Monitoring & Observability with Datadog

install Datadog with datdog-operator Helm charts

kubectl create ns datadog
helm repo add datadog https://helm.datadoghq.com
helm install -n datadog datadog-operator-rel0 datadog/datadog-operator
# output

NAME: datadog-operator-rel0
LAST DEPLOYED: Sat Dec  2 09:04:47 2023
NAMESPACE: datadog
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:

apply datadog-agent.yaml file which will collect Traces, Metrics and Logs from the out cluster, the file includes Log Mangement, Network Performance Monitoring (NPM), Cluster Agent, Cluster checks runners, Events & Processes collection enabled, and Application Performance Monitoring (APM) with hostPort configured to allow Istio sidecars to connect with the agent

kubectl -n datadog apply -f datadog/datadog-agent.yaml
# output
datadogagent.datadoghq.com/datadog created

# apply datadog secret
kubectl -n datadog apply -f cluster/dd-secret.yaml
# output
secret/datadog-secret created

istio configuration for Datadog

upgrade Istio installation by applying kubernetes-manifests/istio-operator.yaml that i configued to have Datadog as proxy tracer

kubectl -n istio-system apply -f kubernetes-manifests/istio-operator.yaml
# output
istiooperator.install.istio.io/installed-state configured

Datadog Infrastructure Monitoring

Datadog Infrastructure Monitoring monitors & Tracks your hosts metrics, containers, processes

below is Infrastructure List, show Kubernetes nodes status, CPU usage for each node, IOwait ,...

containers map grouped by short_image, the red containers indecate high consumption on the metric selected while the green shows a low consumption and a good state

total CPU% and RSS memory timeline for each container

list of all containers in the cluster with state and metrics of each

Kubernetes pods with their status, we can tell easily if any pod's state is crashing, pending, container creating or Running. cpu and memory usage limits, ips, ...

Kuberentes lusters, namespaces, nodes, deployments and other resources

Kuberenetes persistent volumes, and persistent volume claims that was created for the text-to-speech service, shows status bound

Application Performance Monitoring (APM)

APM can gives us visibility into Traces, code Profiles ,Service Catalog and Service List

below is the Service List shows req/s P99 latency and error rate per services

the grpc client is the python client calling Google Cloud texttospeech api and performing tts function

APM Traces

here we can see the time spans for each trace

APM code Profilers

Profilers go deep inside the code to show us metrics such as CPU time for each function in our code below is the profiler for the python text-to-speech service

Logs Management

Logs Management collects and visualize in one place all our applications containers' logs, network logs, firewall logs, system and infrastructure logs

we can filter and group logs according to service, env, status, @http.status_code, @http.url, @Method , ...