Knative Source for Apache Kafka¶
The KafkaSource
reads messages stored in existing Apache Kafka topics, and sends those messages as CloudEvents through HTTP to its configured sink
. The KafkaSource
preserves the order of the messages
stored in the topic partitions. It does this by waiting for a successful response from the sink
before it delivers the next message in the same partition.
Install the KafkaSource controller¶
-
Install the
KafkaSource
controller by entering the following command:kubectl apply -f https://storage.googleapis.com/knative-nightly/eventing-kafka-broker/latest/eventing-kafka-controller.yaml
-
Install the Kafka Source data plane by entering the following command:
kubectl apply -f https://storage.googleapis.com/knative-nightly/eventing-kafka-broker/latest/eventing-kafka-source.yaml
-
Verify that
kafka-controller
andkafka-source-dispatcher
are running, by entering the following command:kubectl get deployments.apps,statefulsets.apps -n knative-eventing
Example output:
NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/kafka-controller 1/1 1 1 3s NAME READY AGE statefulset.apps/kafka-source-dispatcher 1/1 3s
Optional: Create a Kafka topic¶
Note
The create a Kafka topic section assumes you're using Strimzi to operate Apache Kafka, however equivalent operations can be replicated using the Apache Kafka CLI or any other tool.
If you are using Strimzi:
-
Create a
KafkaTopic
YAML file:apiVersion: kafka.strimzi.io/v1beta2 kind: KafkaTopic metadata: name: knative-demo-topic namespace: kafka labels: strimzi.io/cluster: my-cluster spec: partitions: 3 replicas: 1 config: retention.ms: 7200000 segment.bytes: 1073741824
-
Deploy the
KafkaTopic
YAML file by running the command:Wherekubectl apply -f <filename>.yaml
<filename>
is the name of yourKafkaTopic
YAML file.Example output:
kafkatopic.kafka.strimzi.io/knative-demo-topic created
-
Ensure that the
KafkaTopic
is running by running the command:kubectl -n kafka get kafkatopics.kafka.strimzi.io
Example output:
NAME CLUSTER PARTITIONS REPLICATION FACTOR knative-demo-topic my-cluster 3 1
Create a Service¶
-
Create the
event-display
Service as a YAML file:apiVersion: serving.knative.dev/v1 kind: Service metadata: name: event-display namespace: default spec: template: spec: containers: - # This corresponds to # https://github.com/knative/eventing/tree/main/cmd/event_display/main.go image: gcr.io/knative-releases/knative.dev/eventing/cmd/event_display
-
Apply the YAML file by running the command:
Wherekubectl apply -f <filename>.yaml
<filename>
is the name of the file you created in the previous step.Example output:
service.serving.knative.dev/event-display created
-
Ensure that the Service Pod is running, by running the command:
kubectl get pods
The Pod name is prefixed with
event-display
:NAME READY STATUS RESTARTS AGE event-display-00001-deployment-5d5df6c7-gv2j4 2/2 Running 0 72s
Kafka event source¶
-
Modify
source/event-source.yaml
accordingly with bootstrap servers, topics, and so on:apiVersion: sources.knative.dev/v1beta1 kind: KafkaSource metadata: name: kafka-source spec: consumerGroup: knative-group bootstrapServers: - my-cluster-kafka-bootstrap.kafka:9092 # note the kafka namespace topics: - knative-demo-topic sink: ref: apiVersion: serving.knative.dev/v1 kind: Service name: event-display
-
Deploy the event source:
kubectl apply -f event-source.yaml
Example output:
kafkasource.sources.knative.dev/kafka-source created
-
Verify that the KafkaSource is ready:
kubectl get kafkasource kafka-source
Example output:
NAME TOPICS BOOTSTRAPSERVERS READY REASON AGE kafka-source ["knative-demo-topic"] ["my-cluster-kafka-bootstrap.kafka:9092"] True 26h
Scaling¶
To schedule more or fewer consumers, a KafkaSource can be scaled, and they can be allocated to different dispatcher pods. The kafkasource status displays such allocation under the status.placements key.
You can scale a KafkaSource with kubectl by using the following notation:
kubectl scale kafkasource -n <ns> <kafkasource-name> --replicas=<number-of-replicas> # e.g. 12 replicas for a topic with 12 partitions
Alternatively, if you are using a GitOps approach, you can add the consumers
key as shown in the example below and commit it to your repository:
apiVersion: sources.knative.dev/v1beta1
kind: KafkaSource
metadata:
name: kafka-source
spec:
consumerGroup: knative-group
bootstrapServers:
- my-cluster-kafka-bootstrap.kafka:9092
consumers: 12 # Number of replicas
topics:
- knative-demo-topic
sink:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: event-display
Automatic Scaling with KEDA¶
Kafka Sources and Brokers for Apache Kafka have (Alpha) support for serverless scaling with KEDA, including scale to zero. If you want Knative and KEDA to scale your Kafka source for you, you must install KEDA, and then enable the feature flag.
To enable the feature flag, you need to create or modify the config-kafka-features
configmap in the knative-eventing
namespace. You can create the file as below:
apiVersion: v1
kind: Configmap
metadata:
name: config-kafka-features
namespace: knative-eventing
data:
controller-autoscaler-keda: "enabled"
From there, apply the configmap into your cluster and assuming that KEDA is also installed your Kafka Sources will scale for you! For more information on other values you
can add to the config-kafka-features
configmap, read about the Kafka Broker features.
Verify¶
-
Produce a message (
{"msg": "This is a test!"}
) to the Apache Kafka topic as in the following example:kubectl -n kafka run kafka-producer -ti --image=strimzi/kafka:0.14.0-kafka-2.3.0 --rm=true --restart=Never -- bin/kafka-console-producer.sh --broker-list my-cluster-kafka-bootstrap:9092 --topic knative-demo-topic
Tip
If you don't see a command prompt, try pressing Enter.
-
Verify that the Service received the message from the event source:
kubectl logs --selector='serving.knative.dev/service=event-display' -c user-container
Example output:
☁️ cloudevents.Event Validation: valid Context Attributes, specversion: 1.0 type: dev.knative.kafka.event source: /apis/v1/namespaces/default/kafkasources/kafka-source#my-topic subject: partition:0#564 id: partition:0/offset:564 time: 2020-02-10T18:10:23.861866615Z datacontenttype: application/json Extensions, key: Data, { "msg": "This is a test!" }
Handling Delivery Failures¶
The KafkaSource
implements the Delivery
Specificiation, allowing you to configure event delivery parameters for it, which are applied in cases where an event fails to be delivered:
apiVersion: sources.knative.dev/v1beta1
kind: KafkaSource
metadata:
name: kafka-source
spec:
consumerGroup: knative-group
bootstrapServers:
- my-cluster-kafka-bootstrap.kafka:9092 # note the kafka namespace
topics:
- knative-demo-topic
delivery:
deadLetterSink:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: example-sink
backoffDelay: <duration>
backoffPolicy: <policy-type>
retry: <integer>
sink:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: event-display
The delivery
API is discussed in the Handling Delivery Failure chapter.
Optional: Specify the key deserializer¶
When KafkaSource
receives a message from Kafka, it dumps the key in the Event extension called
Key
and dumps Kafka message headers in the extensions starting with kafkaheader
.
You can specify the key deserializer among four types:
string
(default) for UTF-8 encoded stringsint
for 32-bit & 64-bit signed integersfloat
for 32-bit & 64-bit floating pointsbyte-array
for a Base64 encoded byte array
To specify the key deserializer, add the label kafkasources.sources.knative.dev/key-type
to the
KafkaSource
definition, as shown in the following example:
apiVersion: sources.knative.dev/v1beta1
kind: KafkaSource
metadata:
name: kafka-source
labels:
kafkasources.sources.knative.dev/key-type: int
spec:
consumerGroup: knative-group
bootstrapServers:
- my-cluster-kafka-bootstrap.kafka:9092 # note the kafka namespace
topics:
- knative-demo-topic
sink:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: event-display
Optional: Specify the initial offset¶
By default the KafkaSource
starts consuming from the latest offset in each partition. If you want
to consume from the earliest offset, set the initialOffset field to earliest
, for example:
apiVersion: sources.knative.dev/v1beta1
kind: KafkaSource
metadata:
name: kafka-source
spec:
consumerGroup: knative-group
initialOffset: earliest
bootstrapServers:
- my-cluster-kafka-bootstrap.kafka:9092 # note the kafka namespace
topics:
- knative-demo-topic
sink:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: event-display
Note
The valid values for initialOffset
are earliest
and latest
. Any other value results in a
validation error. This field is honored only if there are no committed offsets for that
consumer group.
Connecting to a TLS-enabled Kafka Broker¶
The KafkaSource supports TLS and SASL authentication methods. To enable TLS authentication, you must have the following files:
- CA Certificate
- Client Certificate and Key
KafkaSource expects these files to be in PEM format. If they are in another format, such as JKS, convert them to PEM.
-
Create the certificate files as secrets in the namespace where KafkaSource is going to be set up, by running the commands:
kubectl create secret generic cacert --from-file=caroot.pem
kubectl create secret tls kafka-secret --cert=certificate.pem --key=key.pem
-
Apply the KafkaSource. Modify the
bootstrapServers
andtopics
fields accordingly.apiVersion: sources.knative.dev/v1beta1 kind: KafkaSource metadata: name: kafka-source-with-tls spec: net: tls: enable: true cert: secretKeyRef: key: tls.crt name: kafka-secret key: secretKeyRef: key: tls.key name: kafka-secret caCert: secretKeyRef: key: caroot.pem name: cacert consumerGroup: knative-group bootstrapServers: - my-secure-kafka-bootstrap.kafka:443 topics: - knative-demo-topic sink: ref: apiVersion: serving.knative.dev/v1 kind: Service name: event-display
Enabling SASL for KafkaSources¶
Simple Authentication and Security Layer (SASL) is used by Apache Kafka for authentication. If you use SASL authentication on your cluster, users must provide credentials to Knative for communicating with the Kafka cluster, otherwise events cannot be produced or consumed.
Prerequisites¶
- You have access to a Kafka cluster that has Simple Authentication and Security Layer (SASL).
Procedure¶
-
Create a secret that uses the Kafka cluster's SASL information, by running the following commands:
STRIMZI_CRT=$(kubectl -n kafka get secret example-cluster-cluster-ca-cert --template='{{index.data "ca.crt"}}' | base64 --decode )
SASL_PASSWD=$(kubectl -n kafka get secret example-user --template='{{index.data "password"}}' | base64 --decode )
kubectl create secret -n default generic <secret_name> \ --from-literal=ca.crt="$STRIMZI_CRT" \ --from-literal=password="$SASL_PASSWD" \ --from-literal=saslType="SCRAM-SHA-512" \ --from-literal=user="example-user"
-
Create or modify a KafkaSource so that it contains the following spec options:
apiVersion: sources.knative.dev/v1beta1 kind: KafkaSource metadata: name: example-source spec: ... net: sasl: enable: true user: secretKeyRef: name: <secret_name> key: user password: secretKeyRef: name: <secret_name> key: password type: secretKeyRef: name: <secret_name> key: saslType tls: enable: true caCert: secretKeyRef: name: <secret_name> key: ca.crt ...
Where
<secret_name>
is the name of the secret generated in the previous step.
Clean up steps¶
-
Delete the Kafka event source:
kubectl delete -f source/source.yaml kafkasource.sources.knative.dev
Example output:
"kafka-source" deleted
-
Delete the
event-display
Service:kubectl delete -f source/event-display.yaml service.serving.knative.dev
Example output:
"event-display" deleted
-
Optional: Remove the Apache Kafka Topic
kubectl delete -f kafka-topic.yaml
Example output:
kafkatopic.kafka.strimzi.io "knative-demo-topic" deleted