Hey I recently noticed that the most expensive jobs of our BigQuery dbt project is the

dbt_run_results

MERGE statement. Anybody else sharing similar observations? Any recommended workarounds? We already have it running in prod and CI only I guess using the more expensive MERGE statement over more efficient delete+insert in order to ensure that the operation happens in a single run and nothing gets lost I found github.com/elementary-data/elementary/issues/1959 online but it was closed

Are you running dbt core 1.x? On fusion, I believe elementary does use an INSERT statement … but also, per my above issue, it hangs indefinitely, so I wouldn’t suggest this workaround 😅

Hi @Tim Hiebenthal! We looked into this issue and it does seem we can easily change this model and a few others to be append-only, which should improve performance significantly. Thanks for surfacing. We'll merge a fix shortly and it will be in the upcoming version. As a workaround, I can suggest simply skipping this model in your runs for now:

dbt run --exclude dbt_run_results

This is actually fine because run results are inserted via the

on-run-end

hook anyway. The model mainly needs to run on version updates in case there's a schema change to this model.

Thanks! will try that out

@Tim Hiebenthal @Alex I believe this should be solved in v0.26.0 that was released last week. In addition to the fix mentioned above there was a fix related to the Fusion hang.

Based on my prelim testing, this does appear to be fixed!

Great to hear @Alex Stevens ! I'll close it.