Skip to content

Insert at most 1,000 rows at once in MR diff background migration

What does this MR do?

Prevent statement timeouts from happening by limiting individual INSERT statements to 1,000 rows for commits, and 100 rows for diff files.

Are there points in the code the reviewer needs to double check?

These limits are pretty arbitrary - should they be lower?

Why was this MR needed?

We are getting a lot of dead tuples in production.

Does this MR meet the acceptance criteria?

What are the relevant issue numbers?

Closes #36631 (closed) and #37505 (closed).

Edited by yorickpeterse-staging

Merge request reports