ReplicatedMergeTree and S3

[Pierre]

Hi,

I would like to configure a Clickhouse table with a ReplicatedMergeTree engine associated to a policy to transfer data from hot local storage to cold S3 remote storage after a few days.

Clickhouse is fine to do that. 

However, I need resilience for the hot storage by duplicating data ("ReplicatedMergeTree") but I do not need resilience for the cold storage since the S3 backend is doing that for me (with erasure coding). If I use a ReplicatedMergeTree, the data are going to be replicated in S3 as well (in addition to S3 erasure coding).

Is there a simple way to avoid duplicating data in S3 ?

Would any modifications in Clickhouse code be complex or not ?

Thanks for your help !

Regards,

Pierre

   

[Denis Zhuravlev]

ClickHouse release 21.4.1 2021-04-12

https://github.com/ClickHouse/ClickHouse/blob/master/CHANGELOG.md#experimental-feature

Zero-copy replication for ReplicatedMergeTree over S3 storage. #16240 (ianton-ru).