Hi all,

I recently sent a message starting a discussion on FIP-51: Support
Per-Partition Bucket Count. However, due to an issue with my email
configuration, that message may not have reached all subscribers on the
list. I apologize for the inconvenience.

I am resending the original message below for those who may have missed it.

---

I'd like to start a discussion on FIP-51: Support Per-Partition Bucket
Count.

Today, a Fluss table and all of its partitions share a single bucket count
fixed at table-creation time. Users who need more buckets for recent
partitions must either over-provision every historical partition from day
one, or accept that recent partitions are under-bucketed and become the
read/write bottleneck. Before this FIP, the only way to change the bucket
count was to create a new table and rewrite all data, including historical
partitions already tiered to the lake.

So, I'd like to propose FIP-51: Support Per-Partition Bucket Count [1] to
make bucket.num alterable so that new partitions are written at the new
bucket count while existing partitions are left untouched, with no data
movement.

Welcome your feedback and suggestions on this proposal. Looking forward to
a productive discussion!

[1]
https://cwiki.apache.org/confluence/spaces/FLUSS/pages/446071721/FIP-51+Support+Per-Partition+Bucket+Count

Best regards,
Kaixuan Duan

Reply via email to