I have a query with join. When Trino has 10 workers it works. When 5 - it fails. As I understand the situation (Trino call this DistributedHashJoon) - it splits entire table into 10 pieces, so smaller piece lead to less resource usage. But what if to split join into 10 piece/tasks but run these all of on 5 workers?
Looks like the same situation, just tasks/splits will "live" a bit longer in coordination queue.
Any ideas how to spli hash join into more pieces?
•
It seems like you want an actual solution - as opposed to general advice. In which case you should delete this "advice" question and ask a regular question.
As an aside you don't appear to have accepted any answers from your recent questions, nor commented as to why the provided answers don't meet your needs. This is unusual from someone of your reputation.
And maybe you have some spelling mistakes in your question e.g.
DistributedHashJoon -> DistributedHashJoin?
spli hash join -> split hash join?
