Incremental Alter Config operation kills connection

[weeco]

Version & Environment

Redpanda version: (use rpk version): v23.3.3

What went wrong?

When sending the following franz-go request the connection seems to get killed by Redpanda and no response is sent to the Kafka client. This generally seems to be happening for any request with an invalid value for the to be set config value (e.g. you can use some-wrong-val for cleanup.policy and the same issue will happen).

What should have happened instead?

If something is wrong with this request, an appropriate error response should be returned in this case. I assume INVALID_CONFIG, error code 40 with a proper message would be a good response.

Apache Kafka returns the following error for cleanup.policy = compactt:

Message: Invalid value compactt for configuration cleanup.policy: String must be one of: compact, delete
Error Code: 40 / INVALID_CONFIG

and for compression.type = SNAPPY:

Message: Invalid value SNAPPY for configuration compression.type: String must be one of: uncompressed, zstd, lz4, snappy, gzip, producer
Error Code: 40 / INVALID_CONFIG

Redpanda should return the same.

How to reproduce the issue?

1. Send an incremental alter config request constructed like shown in my screenshot

Additional information

The Redpanda server logs the following log line:

2024-01-24 22:30:02 WARN  2024-01-24 22:30:02,605 [shard 0:main] kafka - connection_context.cc:675 - Error processing request: std::runtime_error (Fell off the end of a string-switch while matching: SNAPPY)

From the client perspective I get the following log message printed by franz-go:

{"level":"warn","ts":"2024-01-24T22:29:57.585Z","logger":"kafka_client","msg":"read from broker errored, killing connection after 0 successful responses (is SASL missing?)","addr":"localhost:9092","broker":"seed_0","err":"EOF"}

[emaxerrno]

good idea on returning the error code.