What happens?

I am testing new duckdb function of larger than memory aggregations.
My query finishes for 2B rows in 1m30s, but never completes for 3B rows.

Query execution progress reaches 49% while doing heavy IO operations in few minutes, and then I see that duckdb process constantly uses 20 cores, and doesn't produce any IO and progress never changes, I waited for 12h for completion.

To Reproduce

.open t
.timer on

SET memory_limit='16GB';
SET threads TO 30;

create table t as
select
  cast(v.range % 300000000 as varchar) v
from range(3000000000) v;

create table t2 as select v, count(*) from t group by 1;

OS:

Debian Linux

DuckDB Version:

0.9.1

DuckDB Client:

CLI

Full Name:

Riku Iki

Affiliation:

nextintelligent.com

Have you tried this on the latest main branch?

I have not tested with any build

Have you tried the steps to reproduce? Do they include all relevant data and configuration? Does the issue you report still appear there?

• Yes, I have