Nice piece of code. We just manually added first step to all HADR-sensitive jobs to check whether the used database is primary (sys.fn_hadr_is_primary_replica) and to raise error if not. The...

The error seemed to be caused by large SQL command via linked server (another SQL server selected a big chunk of data from our SQL server). Once this was eliminated,...

Yes, amount of first and last customer purchase would require a different solution. Sorry for my rant then.

Or you could do it totally old school by MIN(OrderDate), MAX(OrderDate) and GROUP BY CusomerID. And it would be faster, too.

Is there any reason to overcomplicate things?

I concur, that was an interesting question.
Even though I answered it wrong 🙂

I don't think this would fly if I was the one holding the interview. The avatar isn't going to write the code, the person is. Also. with the avatar being...

This is pretty clever and will work even on overlapping intervals pretty fast - at least until someone inserts interval (minint, maxint) 🙂

Statistics updated with fullscan, no change - still estimated 1 rows, actual 87k. I don't know why sql server chooses this plan, really, But, since the intervals do not overlap,...

Auto update for statistics is not disabled; they are even manually updated in a nightly maintenance job, however, the query does not want to use the PK index. I guess...

Ok, I am sorry; here are the execution plans. The first one is without any hints, the second is with FORCESCAN on GeoIpBlock. FORCE ORDER hint produces the same plan,...

We have just:

- production database

- developer database, copied from production database once a month

Backups are managed by our sysadmin and are stored god knows where. 6 developers on the...

Highly available database is a must for us (a large eshop). We used failover clustering but are going to move to AlwaysOn in a few weeks.

Ok, I got it wrong after reading the books online specifically on that subject. I agree, with single log file, both answers are correct.

Your instinct was...

I am aware of that article and even acknowledge that the optimalization in QN is good - it's just that the old way was more useful for a caching engine.

Ours...