Is it possible to keep old hypertable chunks on old nodes while writing new chunks to only newer nodes?

[eugen8]

In a usecase where the most recent data is queried the most (e.g. at most 1 month old) while old data still queried occasionally, is it possible to primarily keep the old data on those certain “archiving” nodes while the new chunks to be inserted only on other, newer nodes?
I was thinking this could be cleaner not to look at those old nodes if we look at recent data and also to not have to add more space to old nodes if the data keeps growing.

Also what happens if a particular node runs out of memory, will TimescaleDB automatically use nodes that have enough hard drive memory and stop writing new chunks to the full nodes?

[loveworldlovesky]

Hi,
dmitry from Timescale Team Tell me how to control
the data to be saved to the new node .
Please refer to

Distributed hypertables in multinode-timescaledb,How to control which node to save to

Hey @loveworldlovesky , We have two experimental functions which allow to do this: block_new_chunks and allow_new_chunks. First function prevents creating new chunks on a particular data node, but updates for existing chunks are still going. I believe those functions are not documented unfortunately, so I’ll share the link to the definition: https://github.com/timescale/timescaledb/blob/main/sql/ddl_experimental.sql#L15

There are two experimental functions that can solve.
( block_new_chunks and allow_new_chunks )