Does reorder_chunk benefits with setting filterfactor on table?

I use Red Hat 8.9, PostgreSQL 15.6 and TimescaleDB v2.14.2 on self-hosted on-prem.

Data in my hypertable are inserted completely random, because there are inserted from different devices. I see that reordering chunks with “device+timestamp” index order may significantly improve select statements that are based on devices inside timer interval.

Questions:

1. Should I change “filter factor” on table to lower then 100% (default)? For example:
   ALTER TABLE table_name SET (FILLFACTOR = 70);
2. Does TimescaleDB uses this kind of reordering data inside chunks that parameter FILLFACTOR would be beneficial at all?
3. Is there some starting point recommendation to begin with? Like start with 90 and then test and lower the value by 10, do next test etc?

2 posts - 2 participants

Read full topic