Skip to main content
Version: Next

Pulsar connector overview

Messaging systems are most powerful when you can easily use them with external systems like databases and other messaging systems.

Pulsar IO connectors enable you to easily create, deploy, and manage connectors that interact with external systems, such as Apache Cassandra, Aerospike, and many others.

Concept

Pulsar IO connectors come in two types:

This diagram illustrates the relationship between source, Pulsar, and sink:

Pulsar IO diagram

Source

Source connectors feed data from external systems into Pulsar.

Common sources include other messaging systems and firehose-style data pipeline APIs.

For the complete list of Pulsar built-in source connectors, see source connector.

Sink

Sink connectors feed data from Pulsar into external systems.

Common sinks include other messaging systems and SQL and NoSQL databases.

For the complete list of Pulsar built-in sink connectors, see sink connector.

Processing guarantee

Processing guarantees are used to handle errors when writing messages to Pulsar topics.

Pulsar connectors and Functions use the same processing guarantees as below.

Delivery semanticDescription
at-most-onceEach message sent to a connector is to be processed once or not to be processed.
at-least-onceEach message sent to a connector is to be processed once or more than once.
effectively-onceEach message sent to a connector has one output associated with it.

Processing guarantees for connectors not just rely on Pulsar guarantee but also relate to external systems, that is, the implementation of source and sink.

  • Source

    Pulsar ensures that writing messages to Pulsar topics respects the processing guarantees. It is within Pulsar's control.

  • Sink

    The processing guarantees rely on the sink implementation. If the sink implementation does not handle retries in an idempotent way, the sink does not respect the processing guarantees.

Set processing guarantees

When creating a connector, you can set the processing guarantee with the following semantics:

  • ATLEAST_ONCE

  • ATMOST_ONCE

  • EFFECTIVELY_ONCE

If --processing-guarantees is not specified when creating a connector, the default semantic is ATLEAST_ONCE.

Take Admin CLI as an example. For more information about REST API or JAVA Admin API, see here.

bin/pulsar-admin sources create \
--processing-guarantees ATMOST_ONCE \
# Other source configs

For more information about the options of pulsar-admin sources create, see here.

Update processing guarantees

After creating a connector, you can update the processing guarantee with the following semantics:

  • ATLEAST_ONCE

  • ATMOST_ONCE

  • EFFECTIVELY_ONCE

Take Admin CLI as an example. For more information about REST API or JAVA Admin API, see here.

bin/pulsar-admin sources update \
--processing-guarantees EFFECTIVELY_ONCE \
# Other source configs

For more information about the options of pulsar-admin sources update, see here.

Work with connector

To manage Pulsar connectors (for example, create, update, start, stop, restart, reload, delete and perform other operations on connectors), you can use the Connector Admin CLI with sources and sinks subcommands. For the latest and complete information, see Pulsar admin docs.

Connectors (sources and sinks) and Functions are components of instances, and they all run on Functions workers. When managing a source, sink or function via the Connector Admin CLI or Functions Admin CLI, an instance is started on a worker. For more information, see Functions worker.