Data Tiering for Distributed Hypertables

Hi, is it possible to setup data tiering for a distributed hypertable? The regular move_chunk function that allows you to move chunks to another table space does not work for distributed hypertables. From the data node’s perspective these are regular hypertables so you could run the function on on each individual data node for each chunk you care about.

Besides being a bit of a pain to setup would there be any issues with this approach?
Is there a better approach, and if not are there plans to provide this kind of functionality for distributed hypertables in the future?

There may be plans in the future, but for now, efforts on move_data() work for distributed hypertables is focused on chunk re-distribution and not data-tiering.

In theory, however, you could do data-tiering at the data node level. The access node currently only keeps track of where each partition is (and their constraints) so that it can direct queries appropriately. Once it sends the query to the DN, most of the work is done there.

If you were going to try it, one option you could try is distributed_exec().

So, this isn’t an official signal of support, but in theory it should work for you. :smiley:


Thanks for the information. We’ll try to set this up directly on the data nodes. I wanted to make sure there was nothing obviously problematic with the approach. It would be nice to easily manage this kind of thing from the access node, but managing directly on the data nodes will work for us for now.

Just wanted to give a quick update. We followed the approach of setting up data tiering jobs for each table on each data node as you would for normal hypertables, and so far this has worked as expected.

1 Like