Hi. Yesterday i added a NON-cL index on entityID and it seems to do much better, see attached plan.

So it seems we need to cover the bases with a second...

Hello, and thanks for the answers! I realized that i never specified exactly how the data was "read".

Statisticstype is always part of the "where", because it makes no sense to...

Sorry for the red herrings, I just whipped up an "anonymized" version of one of our internal queries for demo purposes 🙂

The brackets are the thing of interest here, since...

Sounds strange. Are you running queries from SSMS?

Can you do a screenshot of following from your system:

select compatibility_level, @@VERSION, *
from sys.databases
where name = DB_NAME()
go
select *
from openjson ('{"test"...

You could probably add some more details, like what's exactly failing etc.

According to https://stackoverflow.com/questions/24118508/how-can-i-pass-a-body-from-stored-procedure-to-a-rest-service/38185814 it should be possible so you probably messed up the call

Run a trace to capture queryplans etc against the unit:ed database and see why it's sometimes slow and sometimes not. You can script the traces as part of your unit...

Follow-up for someone interested in this 🙂

The issue was due to dirty reads. In some rare occurances, outorder.status was updated in the middle of that insert, leading two values to...

A bit silly to reply to own post, but a better solution to this was to use the MERGE clause, it has access to both source and target and inserted...

Well, outorder-table is PK:ed on outorder, so it cannot have any dupes ever.
So the problem "must" be inside the derived table, but how can it be dupes if i...

As a followup. I added a DISTINCT into the #t-insert and the problem went away.
Will try to create a test case to see if the problem can be...

It's unlikely since it's a job, but yeah, it's possible. But anyway, it's inserting into a temp table so overlapping shouldn't matter i guess?

You probably fixed it already but if not, i'd probably go for the "delayed" compilation by hardcoding the dynamic sql depending on variant_property. Worked on in 2016 machine, but has...

If you're running a default trace and global #t-table has a PK, you might get the info from the default trace data if it hasn't rolled over yet
[code...