Adam Kotwasinski
1 min readFeb 21, 2020

--

Great article!
One more minor limitation — you can also have “dangling references” in S3 if you use compaction, as S3 does not know when the record gets removed.

The trivial way to solve that would be to somehow track “externally-persisted” records and periodically re-scan the Kafka broker log or fork the server to register the compacted id-s and then run a cleanup action in S3.

--

--