Too much logging for AWSKafkaAvroSerializer (part two)

[snemarch]

Hi, we're getting a "Schema Version Id is null. Trying to register the schema" log for every message serialized with the AWSKafkaAvroSerializer.

It seems that this was changed from info to debug in #9 , but changed back to info as part of #49 (commit a7017fb). Re-changing the log level to debug would remove the spam, but I'm contemplating if there could be some underlying issues?

I haven't fully grokked the code, but in AWSKafkaAvroSerializer#serialize the null-check is against the final schemaVersionId... is there any feasible way to get this configured/initialized to be non-null?

My hunch is that either there's a bit of an architectural issue, or there's some configuration issue we've totally missed – it doesn't seem right that the lookup fails for each and every message. Fortunately the caching in GlueSchemaRegistrySerializationFacade seems to be working, though! 😅

[blacktooth]

Re-directing the discussion to #64

[blacktooth]

Thanks for reporting this! Looks like the code got reverted in this file. We will investigate and fix it.

[snemarch]

Thanks for reporting this! Looks like the code got reverted in this file. We will investigate and fix it.

Note that I think there's some underlying issue (possibly with our setup, we might very well have missed some configuration!), which will be masked out by re-changing the log level from info to debug.

Some more context, in case it helps: we're running a CDC pipeline: Postgres -> Debezium -> Kafka -> ElasticSearch, orchestrated with Kafka Connect. We're using AWS MSK and Glue Registry, while Connect is running in our own Kubernetes cluster.