Claus Ibsen created CAMEL-27130:
-----------------------------------
Summary: camel-kafka - a schemaRegistryVendor option that sets up
the serde classes and registry properties of a schema registry
Key: CAMEL-27130
URL: https://issues.apache.org/jira/browse/CAMEL-27130
Project: Camel
Issue Type: New Feature
Components: camel-kafka
Reporter: Claus Ibsen
Using a schema registry with camel-kafka today means setting the serde classes
of the registry vendor (valueSerializer / valueDeserializer) and the registry
properties under their vendor-specific names (additionalProperties...), and
knowing which jars to add. camel-kafka has schemaRegistryURL and
specificAvroReader, but schemaRegistryURL only sets the Confluent property
schema.registry.url.
*Proposal:* two options that let camel-kafka configure the registry itself:
* schemaRegistryVendor: confluent, apicurio, aws-glue, azure or custom.
* schemaFormat: avro, json or protobuf. Default avro; not every vendor supports
every format.
>From these, camel-kafka sets the serializer and deserializer classes and the
>vendor's registry properties, such as the registry URL and the auto-register
>and use-latest flags, using the existing schemaRegistryURL as the URL. The
>classes are only names (strings), so camel-kafka needs no compile dependency
>on any vendor. Options the user sets explicitly (valueSerializer,
>valueDeserializer, additionalProperties) still win.
The vendors, with what was checked on 2026-11-17:
|| Vendor || Serde artifacts || On Maven Central || Licence ||
| confluent | io.confluent:kafka-avro-serializer and others | No, only
https://packages.confluent.io/maven/ (public, no login) | Avro and client:
Apache 2.0. JSON Schema and Protobuf: Confluent Community License and Apache
2.0 |
| apicurio | io.apicurio:apicurio-registry-avro-serde-kafka 3.0.8 and others
(Apicurio 3 renamed the 2.x artifacts) | Yes | Apache 2.0 |
| aws-glue | software.amazon.glue:schema-registry-serde 1.1.25 | Yes | Apache
2.0 |
| azure | com.microsoft.azure:azure-schemaregistry-kafka-avro 1.1.2 | Yes | MIT
|
Several registries speak the Confluent registry API, so confluent would also
cover, for example, Karapace, the Redpanda schema registry and the
Confluent-compatible API of Apicurio.
*Dependencies.* camel-kafka does not depend on the vendor jars. The user adds
them:
* For confluent, the user is responsible for adding the Confluent Maven
repository, or for having the jars in their own pom.xml, as they are not on
Maven Central.
* Camel JBang could add the jars of the chosen vendor through its dependency
detection, from the option value. For confluent it would also need the
Confluent repository, given with --repos.
*Kamelets.* With this, the Kafka Kamelets of camel-kamelets can offer a schema
registry with a few optional options (registry, URL, format) instead of one
Kamelet per vendor and format. A Kamelet template cannot choose classes or
dependencies from an option value by itself: nested placeholders and
placeholders in Kamelet dependencies do not resolve. See
apache/camel-kamelets#3183 (design and spike result), #3182 and #3144.
Related: CAMEL-24172 (proposal for a camel-apicurio-registry component).
_Claude Code on behalf of Claus Ibsen_
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)