What is kawa?
kawa is a gateway for Apache Kafka. It speaks the Kafka wire protocol, so every existing Kafka client (a producer, a consumer, an admin client, a Kafka Streams application or a command-line tool) connects to it as if it were a broker. Because kawa works at the protocol level rather than through a client library, it does not depend on the client's language or code: no libraries need to change, and clients point at kawa instead of the cluster, using their own credentials and topic names.
You might think of kawa as a proxy, and for some features that is exactly what it is: to rewrite topic names, check access or filter records, it sits between your clients and the cluster and handles their requests on the way through.
But kawa is more than a mere proxy. It amplifies the capabilities of your data streaming platform, whether or not a given feature works as a proxy. Virtual topics give clients stable names while the data lives elsewhere. Consume filters show each client only the records it should see. Governance rules refuse topics, partition counts and group names that break your conventions. Clients authenticate to kawa, and each request is checked against access rules before it reaches a broker. Some of this is proxying and some of it is not, but together it gives you things Kafka does not provide on its own.
What it does
| Capability | What you get |
|---|---|
| Client authentication | Clients log in with SASL (PLAIN, SCRAM-SHA-256 or SCRAM-SHA-512) or with a TLS client certificate. Clients need no principal on the cluster: kawa logs in to the cluster with one service identity. |
| TLS | Client listeners can use SSL or SASL_SSL. kawa can reach the cluster over TLS, optionally with a client certificate for mutual TLS. |
| Access control | Roles, groups and ACLs decide what each principal may do. Gated requests are checked before they reach a broker. The default is deny. |
| Virtual topics | Clients use a name such as orders.eu while the data lives in orders-v2. Names are rewritten in both directions, in every request and response that carries one. |
| Consume filters | A virtual topic can show a client only the records it should see, selected by record header or by a CEL expression over the decoded record. Offsets of the surviving records are preserved. |
| Governance | CEL rules check topic names, partition counts and configs when topics are created, altered or deleted, and check consumer group and transactional id names when they are used. |
| Management | Studio, a web admin UI. A REST admin API with OAuth 2.0 login. A Kubernetes operator that manages virtual topics as VirtualTopic resources. |
Status
kawa is pre-1.0. The current release candidate line is 0.1.0-rc.2. Pin the image version in production and read the release notes before you upgrade.
Where to go next
Start with Getting started, which covers installing it with the Helm chart. Configuration is the reference for the YAML file and the dynamic config. The Modules section explains virtual topics, authentication, access control and governance. The Kubernetes operator page covers managing virtual topics from a cluster.