Hi,
any suggestion what can be this gap?
It’s one line of code which do Repo.preload(ta, preloads). I wrapped instrumentation about this line and you can see screenshot for AppSignal.
You can see telemetry from Ecto about sql queries(preloads). Almost at the end, there is huge gap in timeline for that part of instrumentation/function.
Locally numbers are small, but in production gap is about 500-1000ms when we load more entries from database.
Is it serialization or something like that? I would expect it to have it at the end, not before last 1/3 of queries. But still, we have some big columns in these schemas so we use load_in_query: false. So it should not happen.
