Starlight for Kafka protocol handler
This document describes how to implement the Starlight for Kafka protocol handler.
Topic
In Kafka, all topics are stored in one flat namespace, but in Pulsar, topics are organized in hierarchical multi-tenant namespaces. Starlight for Kafka introduces a setting kafkaNamespace in the broker configuration which allows administrators to map Kafka topics to Pulsar topics.
To let Kafka users leverage the multi-tenancy of Apache Pulsar, a Kafka user can specify a Pulsar tenant and namespace as the topic prefix like Pulsar topics:
| Kafka topic name | Tenant | Namespace | Short topic name |
|---|---|---|---|
my-topic |
|
|
my-topic |
my-tenant/my-ns/my-topic |
my-tenant |
my-ns |
my-topic |
persistent://my-tenant/my-ns/my-topic |
my-tenant |
my-ns |
my-topic |
|
When |
Topic lookup
Starlight for Kafka uses the same topic lookup approach for the Kafka request handler and the Pulsar request handler. The request handler does topic discovery to look up all ownerships for the requested topic partitions, and responds with the ownership information as part of the Kafka TopicMetadata returned to Kafka clients.
When you are using the Proxy module, the proxy answers its own advertised address as owner of every topic. The Proxy automatically dispatches every request to the Broker that owns the topic, so Kafka clients do not have knowledge of the internal topology of the cluster.
Message
Kafka and Pulsar messages have keys, values, timestamps, and headers, but in Pulsar the headers are called properties. Starlight for Kafka converts these
fields automatically between Kafka messages and Pulsar messages. The mapping relationships are as below.
Pulsar to Kafka message field mapping relationships
| Pulsar | Kafka | Note |
|---|---|---|
|
|
Ordering key has higher priority, which means: |
|
|
Event time has higher priority, which means: |
NULL value |
NULL value |
If the field of a Pulsar message is NULL, the converted Kafka message is NULL. |
Kafka to Pulsar message field mapping relationships
| Pulsar | Kafka | Note |
|---|---|---|
|
Ordering key and partition key |
If a Kafka message has a |
|
Event time and publish time |
If a Kafka message has a |
NULL value |
NULL value |
If the field of a Kafka message is NULL, the converted Pulsar message is NULL. |
Message ID and offset
In Kafka, each message is assigned with an offset once the message is successfully produced to a topic partition.
In Pulsar, each message is assigned with a MessageID. The message ID consists of ledger-id, entry-id, and batch-index components.
Starlight for Kafka uses the same approach in the Pulsar-Kafka wrapper to convert a Pulsar MessageID to an offset, and vice versa.
Produce Messages
When the Kafka request handler receives produced messages from a Kafka client, it converts Kafka messages to Pulsar messages by mapping the fields (such as the key, value, timestamp and headers) one by one, and uses the ManagedLedger append API to append those converted Pulsar messages to BookKeeper. Converting Kafka messages to Pulsar messages allows existing Pulsar applications to consume messages produced by Kafka clients.
Consume Messages
When the Kafka request handler receives a consumer request from a Kafka client, it opens a non-durable cursor to read the entries starting from the requested offset. The Kafka request handler converts the Pulsar messages back to Kafka messages to allow existing Kafka applications to consume the messages produced by Pulsar clients.
Group coordinator & offset management
Pulsar does not have a centralized group coordinator for assigning partitions to consumers of a consumer group or managing offsets for each consumer group. In Pulsar, partition assignment is managed by the broker on a per-partition basis, and the offset management is done by storing the acknowledgements in cursors by the owner broker of that partition.
To be fully compatible with Kafka clients, Starlight for Kafka implements the Kafka group coordinator by storing the coordinator group changes and offsets in a system topic called kafkaTenant/kafka/consumer_offsets in Pulsar.
This bridges the gap between Pulsar and Kafka and allows use of existing Pulsar tools and policies to manage subscriptions and monitor Kafka consumers. Starlight for Kafka adds a background thread in the implemented group coordinator to periodically synchronize offset updates from the system topic to Pulsar cursors. Therefore, a Kafka consumer group is effectively treated as a Pulsar subscription to allow existing Pulsar tools to be used for managing Kafka consumer groups.
What’s next?
For more on Starlight for Kafka, see:
