Feature request: Different replication_mode within one cluster #550

New issue

Open

opened 2023-04-22 09:51:31 +00:00 by yuka · 0 comments

yuka commented

2023-04-22 09:51:31 +00:00

Contributor

I have three servers, each is its own zone:
yaya 250G
yoyo 5T
yuyu 5T

I have some amount of data (media files, ~1T) that I want to store but replication mode "2" is fine.
I have some amount of data that is being actively worked on (matrix media, nextcloud primary storage, ~100G) and this data should have replication mode "3" because when one node goes offline it should still be writable.

Currently my workaround for this is that I have two completely separate garage clusters, with different S3 API and RPC ports. One has replication_mode = "3" and stores its data on yaya, yoyo and yuyu. The other has replication_mode = "2" and stores its data on yoyo and yuyu only.

It would be wonderful if I didn't have to maintain different clusters for this. In the sense of re-using the key-pairs, and connections, and not having to run two processes.

Either the replication_mode could be configurable per bucket, or even better there could be multiple "pools" with individual replication mode and layouts.

I have three servers, each is its own zone: yaya 250G yoyo 5T yuyu 5T I have some amount of data (media files, ~1T) that I want to store but replication mode "2" is fine. I have some amount of data that is being actively worked on (matrix media, nextcloud primary storage, ~100G) and this data should have replication mode "3" because when one node goes offline it should still be writable. Currently my workaround for this is that I have two completely separate garage clusters, with different S3 API and RPC ports. One has replication_mode = "3" and stores its data on yaya, yoyo and yuyu. The other has replication_mode = "2" and stores its data on yoyo and yuyu only. It would be wonderful if I didn't have to maintain different clusters for this. In the sense of re-using the key-pairs, and connections, and not having to run two processes. Either the replication_mode could be configurable per bucket, or even better there could be multiple "pools" with individual replication mode and layouts.