diff --git a/.gitignore b/.gitignore index 009e10da1..480a9b5b4 100644 --- a/.gitignore +++ b/.gitignore @@ -8,7 +8,7 @@ bin charts/**/charts charts/koperator/requirements.lock - +charts/kafka-operator/ingress # Test binary, build with `go test -c` *.test diff --git a/api/v1beta1/kafkacluster_types.go b/api/v1beta1/kafkacluster_types.go index 0e03045fe..5fa3d22e9 100644 --- a/api/v1beta1/kafkacluster_types.go +++ b/api/v1beta1/kafkacluster_types.go @@ -157,9 +157,16 @@ type KafkaClusterSpec struct { // This is default to be true; if set to false, the Kafka cluster is in ZooKeeper mode. // +kubebuilder:default=false // +optional - KRaftMode bool `json:"kRaft"` - HeadlessServiceEnabled bool `json:"headlessServiceEnabled"` - ListenersConfig ListenersConfig `json:"listenersConfig"` + KRaftMode bool `json:"kRaft"` + HeadlessServiceEnabled bool `json:"headlessServiceEnabled"` + // DebugEnabled is used to decide whether to create a separate loadbalancer services for the + // Kafka and Cruise Control Pods. These services will expose the internal listener ports of the Kafka + // cluster with LoadBalancer type, which can be used for running Koperator on a local machine against + // a kafkaCluster instance on a Kind Cluster. + // +kubebuilder:default=false + // +optional + DebugEnabled bool `json:"debugEnabled"` + ListenersConfig ListenersConfig `json:"listenersConfig"` // Custom ports to expose in the container. Example use case: a custom kafka distribution, that includes an integrated metrics api endpoint AdditionalPorts []corev1.ContainerPort `json:"additionalPorts,omitempty"` // ZKAddresses specifies the ZooKeeper connection string diff --git a/charts/kafka-operator/crds/kafkaclusters.yaml b/charts/kafka-operator/crds/kafkaclusters.yaml index ff4636057..3e14a374d 100644 --- a/charts/kafka-operator/crds/kafkaclusters.yaml +++ b/charts/kafka-operator/crds/kafkaclusters.yaml @@ -19231,6 +19231,14 @@ spec: type: object type: array type: object + debugEnabled: + default: false + description: |- + DebugEnabled is used to decide whether to create a separate loadbalancer services for the + Kafka and Cruise Control Pods. These services will expose the internal listener ports of the Kafka + cluster with LoadBalancer type, which can be used for running Koperator on a local machine against + a kafkaCluster instance on a Kind Cluster. + type: boolean disruptionBudget: description: DisruptionBudget defines the configuration for PodDisruptionBudget where the workload is managed by the kafka-operator diff --git a/config/base/crds/kafka.banzaicloud.io_kafkaclusters.yaml b/config/base/crds/kafka.banzaicloud.io_kafkaclusters.yaml index ff4636057..3e14a374d 100644 --- a/config/base/crds/kafka.banzaicloud.io_kafkaclusters.yaml +++ b/config/base/crds/kafka.banzaicloud.io_kafkaclusters.yaml @@ -19231,6 +19231,14 @@ spec: type: object type: array type: object + debugEnabled: + default: false + description: |- + DebugEnabled is used to decide whether to create a separate loadbalancer services for the + Kafka and Cruise Control Pods. These services will expose the internal listener ports of the Kafka + cluster with LoadBalancer type, which can be used for running Koperator on a local machine against + a kafkaCluster instance on a Kind Cluster. + type: boolean disruptionBudget: description: DisruptionBudget defines the configuration for PodDisruptionBudget where the workload is managed by the kafka-operator diff --git a/config/samples/simpleZookeeper.yaml b/config/samples/simpleZookeeper.yaml new file mode 100644 index 000000000..82123498c --- /dev/null +++ b/config/samples/simpleZookeeper.yaml @@ -0,0 +1,9 @@ +apiVersion: zookeeper.pravega.io/v1beta1 +kind: ZookeeperCluster +metadata: + name: zookeeper-server + namespace: zookeeper +spec: + replicas: 3 + persistence: + reclaimPolicy: Delete \ No newline at end of file diff --git a/config/samples/simplekafkacluster.yaml b/config/samples/simplekafkacluster.yaml index d890f8551..cf08d8980 100644 --- a/config/samples/simplekafkacluster.yaml +++ b/config/samples/simplekafkacluster.yaml @@ -5,10 +5,11 @@ metadata: controller-tools.k8s.io: "1.0" name: kafka spec: + debugEnabled: true kRaft: false monitoringConfig: jmxImage: "ghcr.io/adobe/koperator/jmx-javaagent:1.4.0" - headlessServiceEnabled: true + headlessServiceEnabled: false zkAddresses: - "zookeeper-server-client.zookeeper:2181" propagateLabels: false diff --git a/config/scaleops/CustomOwnerGrouping.yaml b/config/scaleops/CustomOwnerGrouping.yaml new file mode 100644 index 000000000..7e9760d82 --- /dev/null +++ b/config/scaleops/CustomOwnerGrouping.yaml @@ -0,0 +1,22 @@ + +kind: CustomOwnerGrouping +apiVersion: analysis.scaleops.sh/v1alpha1 +metadata: + name: kafkabroker + namespace: scaleops-system +spec: + groupBy: + positiveRegexMatch: false + groupBys: + - labels: + - 'isBrokerNode: true' + positiveRegexMatch: false + topOwnerController: + apiVersion: kafka.banzaicloud.io/v1beta1 + kind: KafkaCluster + displayOptions: + hideGeneratedSuffix: true + fields: + - ownerName + defaultPolicy: kafka-brokers + enabled: true diff --git a/pkg/resources/cruisecontrol/service.go b/pkg/resources/cruisecontrol/service.go index d868eacf4..2c1c64439 100644 --- a/pkg/resources/cruisecontrol/service.go +++ b/pkg/resources/cruisecontrol/service.go @@ -26,7 +26,7 @@ import ( ) func (r *Reconciler) service() runtime.Object { - return &corev1.Service{ + svc := &corev1.Service{ ObjectMeta: templates.ObjectMeta( fmt.Sprintf(serviceNameTemplate, r.KafkaCluster.Name), apiutil.MergeLabels(ccLabelSelector(r.KafkaCluster.Name), r.KafkaCluster.Labels), @@ -34,6 +34,7 @@ func (r *Reconciler) service() runtime.Object { ), Spec: corev1.ServiceSpec{ Selector: ccLabelSelector(r.KafkaCluster.Name), + Type: corev1.ServiceTypeClusterIP, Ports: []corev1.ServicePort{ { Name: "cc", @@ -50,4 +51,10 @@ func (r *Reconciler) service() runtime.Object { }, }, } + + if r.KafkaCluster.Spec.DebugEnabled { + svc.Spec.Type = corev1.ServiceTypeLoadBalancer + } + + return svc } diff --git a/pkg/resources/kafka/allBrokerService.go b/pkg/resources/kafka/allBrokerService.go index ecfdd5b7b..ed0eed60c 100644 --- a/pkg/resources/kafka/allBrokerService.go +++ b/pkg/resources/kafka/allBrokerService.go @@ -39,7 +39,7 @@ func (r *Reconciler) allBrokerService() runtime.Object { usedPorts = append(usedPorts, generateServicePortForAdditionalPorts(r.KafkaCluster.Spec.AdditionalPorts)...) - return &corev1.Service{ + svc := &corev1.Service{ ObjectMeta: templates.ObjectMetaWithAnnotations( fmt.Sprintf(kafkautils.AllBrokerServiceTemplate, r.KafkaCluster.GetName()), apiutil.LabelsForKafka(r.KafkaCluster.GetName()), @@ -52,4 +52,10 @@ func (r *Reconciler) allBrokerService() runtime.Object { Ports: usedPorts, }, } + + if r.KafkaCluster.Spec.DebugEnabled { + svc.Spec.Type = corev1.ServiceTypeLoadBalancer + } + + return svc } diff --git a/pkg/resources/kafka/kafka.go b/pkg/resources/kafka/kafka.go index 9aa91808f..9e9a2c00e 100644 --- a/pkg/resources/kafka/kafka.go +++ b/pkg/resources/kafka/kafka.go @@ -828,6 +828,7 @@ func (r *Reconciler) reconcileKafkaPod(log logr.Logger, desiredPod *corev1.Pod, return errorfactory.New(errorfactory.APIFailure{}, err, "getting resource failed", "kind", desiredType) } switch { + //initial run - Create Pod case len(podList.Items) == 0: if err := patch.DefaultAnnotator.SetLastAppliedAnnotation(desiredPod); err != nil { return errors.WrapIf(err, "could not apply last state to annotation") @@ -935,6 +936,37 @@ func (r *Reconciler) updateStatusWithDockerImageAndVersion(brokerId int32, broke return nil } +// syncResourceRequests overwrites CPU and memory requests in desiredPod's containers +// with the values from currentPod so that request-only changes do not trigger a pod restart. +func syncResourceRequests(desiredPod, currentPod *corev1.Pod) { + syncContainerResourceRequests(desiredPod.Spec.Containers, currentPod.Spec.Containers) + syncContainerResourceRequests(desiredPod.Spec.InitContainers, currentPod.Spec.InitContainers) +} + +func syncContainerResourceRequests(desired, current []corev1.Container) { + index := make(map[string]corev1.ResourceList, len(current)) + for _, c := range current { + index[c.Name] = c.Resources.Requests + } + for i := range desired { + c := &desired[i] + reqs, ok := index[c.Name] + if !ok { + continue + } + if c.Resources.Requests == nil { + c.Resources.Requests = make(corev1.ResourceList) + } + for _, res := range []corev1.ResourceName{corev1.ResourceCPU, corev1.ResourceMemory} { + if val, exists := reqs[res]; exists { + c.Resources.Requests[res] = val + } else { + delete(c.Resources.Requests, res) + } + } + } +} + //gocyclo:ignore func (r *Reconciler) handleRollingUpgrade(log logr.Logger, desiredPod, currentPod *corev1.Pod, desiredType reflect.Type) error { // Since toleration does not support patchStrategy:"merge,retainKeys", @@ -951,6 +983,8 @@ func (r *Reconciler) handleRollingUpgrade(log logr.Logger, desiredPod, currentPo } desiredPod.Spec.Tolerations = uniqueTolerations } + // Ignore CPU/memory request diffs — changing requests does not require a pod restart. + syncResourceRequests(desiredPod, currentPod) // Check if the resource actually updated or if labels match TaintedBrokersSelector patchResult, err := patch.DefaultPatchMaker.Calculate(currentPod, desiredPod) switch { diff --git a/pkg/resources/kafka/service.go b/pkg/resources/kafka/service.go index fd53e5dc1..5ed75b1e7 100644 --- a/pkg/resources/kafka/service.go +++ b/pkg/resources/kafka/service.go @@ -46,7 +46,7 @@ func (r *Reconciler) service(id int32, _ *v1beta1.BrokerConfig) runtime.Object { Protocol: corev1.ProtocolTCP, }) - return &corev1.Service{ + svc := &corev1.Service{ ObjectMeta: templates.ObjectMetaWithAnnotations(fmt.Sprintf("%s-%d", r.KafkaCluster.Name, id), apiutil.MergeLabels( apiutil.LabelsForKafka(r.KafkaCluster.Name), @@ -61,4 +61,8 @@ func (r *Reconciler) service(id int32, _ *v1beta1.BrokerConfig) runtime.Object { Ports: usedPorts, }, } + if r.KafkaCluster.Spec.DebugEnabled { + svc.Spec.Type = corev1.ServiceTypeLoadBalancer + } + return svc } diff --git a/run-local.sh b/run-local.sh new file mode 100755 index 000000000..674d282c5 --- /dev/null +++ b/run-local.sh @@ -0,0 +1,81 @@ +#!/bin/bash +set -e + +## Prerequisite checks +if [ -z "${SCALEOPS_TOKEN}" ]; then + echo "Error: SCALEOPS_TOKEN environment variable is not set" + exit 1 +fi + +## Create kind cluster +kind delete clusters kind-kafka +kind create cluster --config=./tests/e2e/platforms/kind/kind_config.yaml --name=kind-kafka + +## Build/Load images (Kafka 3.7.0) +kind load docker-image docker-pipeline-upstream-mirror.dr-uw2.adobeitc.com/adobe/kafka:2.13-3.7.0 --name kind-kafka +### Skip if you want to run koperator locally +docker build . -t koperator_e2e_test +kind load docker-image koperator_e2e_test:latest --name kind-kafka + +## Install Helm Charts and CRDs +### project contour +helm repo add contour https://projectcontour.github.io/helm-charts/ || true +helm upgrade --install contour contour/contour --namespace projectcontour --create-namespace + +### cert-manager +helm repo add jetstack https://charts.jetstack.io --force-update || true +helm upgrade --install cert-manager jetstack/cert-manager --namespace cert-manager --create-namespace --version v1.16.2 --set crds.enabled=true + +### zookeeper-operator +helm repo add pravega https://charts.pravega.io || true +helm upgrade --install zookeeper-operator pravega/zookeeper-operator --version 0.2.15 --namespace zookeeper --create-namespace --set crd.create=true + +### prometheus +helm repo add prometheus https://prometheus-community.github.io/helm-charts || true +helm upgrade --install prometheus prometheus/kube-prometheus-stack --version 54.1.0 --namespace prometheus --create-namespace + +### scaleops +helm upgrade --install --create-namespace -n scaleops-system --repo https://registry.scaleops.com/charts/ --username scaleops --password ${SCALEOPS_TOKEN} --set scaleopsToken=${SCALEOPS_TOKEN} --set clusterName=$(kubectl config current-context) scaleops scaleops +kubectl apply -f config/scaleops/CustomOwnerGrouping.yaml +kubectl apply -f config/scaleops/KafkaBrokersPolicy.yaml +#### Scaleops Dashboard Port Forward +# kubectl port-forward 8080 -n scaleops-system +# (find pod name with: kubectl get pods -n scaleops-system) + +## Run Koperator Locally +### Start Cloud Provider Kind in the background to enable LoadBalancer services for local koperator +# sudo ~/go/bin/cloud-provider-kind +# (run this manually in a separate terminal before starting koperator) + +### Start Local Koperator instance: +kubectl create namespace kafka || true +kubectl ens kafka +make install +# Run koperator locally in a separate terminal: +# go run ./main.go --metrics-addr=:8090 --disable-webhooks + +## Initialize Zookeeper and Kafka Cluster +kubectl apply -f config/samples/simplezookeeper.yaml -n zookeeper +kubectl apply -f config/samples/simplekafkacluster.yaml -n kafka + +# NOTES for running koperator locally: +# +# If you want to run koperator locally, make sure to set `debugEnabled: true` +# in your KafkaCluster spec. This will create LoadBalancer services for the +# Kafka and Cruise Control pods, allowing your local koperator to access +# services running on the Kind cluster. +# +# Cloud Provider KIND is required to enable LoadBalancer services on Kind. +# This is necessary for local koperator access. If you don't want to run it, +# you can port-forward the services instead. +# +# Finally, you'll need to update your /etc/hosts file to direct request from +# Koperator to the LoadBalancer IPs. You can find the LoadBalancer IPs by running: +# kubectl get svc -n kafka +# +# Your /etc/hosts entries should look something like this: +# 172.18.0.7 kafka-0.kafka.svc.cluster.local +# 172.18.0.9 kafka-1.kafka.svc.cluster.local +# 172.18.0.10 kafka-2.kafka.svc.cluster.local +# 172.18.0.11 kafka-all-broker.kafka.svc.cluster.local +# 172.18.0.8 kafka-cruisecontrol-svc.kafka.svc.cluster.local diff --git a/tests/e2e/platforms/kind/kind_config.yaml b/tests/e2e/platforms/kind/kind_config.yaml index 65d601b47..6515c31f4 100644 --- a/tests/e2e/platforms/kind/kind_config.yaml +++ b/tests/e2e/platforms/kind/kind_config.yaml @@ -3,6 +3,7 @@ # topology.kubernetes.io/zone (e.g. config/samples/simplekafkacluster_affinity.yaml). kind: Cluster apiVersion: kind.x-k8s.io/v1alpha4 +name: kind-kafka nodes: - role: control-plane kubeadmConfigPatches: @@ -32,9 +33,10 @@ nodes: nodeRegistration: kubeletExtraArgs: node-labels: "topology.kubernetes.io/zone=zone-c" -containerdConfigPatches: -- |- - [plugins."io.containerd.grpc.v1.cri".containerd] - snapshotter = "overlayfs" - [plugins."io.containerd.grpc.v1.cri".registry.mirrors."localhost:5000"] - endpoint = ["http://localhost:5000"] + extraPortMappings: + - containerPort: 80 + hostPort: 80 + listenAddress: "0.0.0.0" + - containerPort: 443 + hostPort: 443 + listenAddress: "0.0.0.0" \ No newline at end of file