Reading from a continuously growing zarr: any relevant caching ?

[mdespriee]

Hello,

I'm exploring the use of xarray+zarr to store raster data that is arriving every 5mn.
I'm storing that data by continuously appending to a zarr store in s3.

On the read side, I have set up some caching in the app (some lru_cache here and there). But there are still a lot of round-trips to s3, and I wonder if some of the fsspec buffering options are relevant ?

Is there a way to cache locally some of the files, while being aware of the changes in .zmetadata and changing dimensions ?

[mdespriee]

I'm also discovering the existence of chunk_store option and the LRUStoreCache.
A LRU cache on the chunk_store, and "normal" store for metadata ?