beryllw opened a new issue, #627:
URL: https://github.com/apache/flink-agents/issues/627

   ### Search before asking
   
   - [x] I searched in the 
[issues](https://github.com/apache/flink-agents/issues) and found nothing 
similar.
   
   ### Description
   
   ## Summary
   
   Currently only Kafka is supported as the external action state store backend 
for durable execution. This proposal adds Apache Fluss as an alternative 
backend.
   
   ## Motivation
   
   The Kafka-based implementation stores action states in an append-only log 
and rebuilds an in-memory HashMap on recovery by sequentially consuming the 
entire topic. This leads to:
   
   - **O(N) recovery cost** proportional to the number of stored action states
   - **Memory pressure** from caching all action states in a HashMap
   - **No native point lookup** — relies entirely on the in-memory cache
   
   Fluss provides primary key tables with native KV point lookup (backed by 
RocksDB), which eliminates these issues:
   
   - **O(1) point lookup** via `Lookuper.lookup()` — no in-memory cache needed
   - **No-op recovery** — `rebuildState()` is unnecessary since data is 
persisted in the KV table
   - **In-place upsert** via `UpsertWriter` instead of append-only writes
   
   ## Scope
   
   - Add `BackendType.FLUSS` and `FlussActionStateStore` implementation
   - Add Fluss connection and authentication config options
   - Add integration tests against an embedded Fluss cluster
   - Update user documentation (configuration and deployment guides)
   
   ### Are you willing to submit a PR?
   
   - [x] I'm willing to submit a PR!


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to