SQL 2017 CU11 highlights:
- Assertion error when you add a database (look, folks, I can’t make this stuff up. Software is truly hard. I mean, I could make this stuff up, but it wouldn’t be anywhere near this funny.)
- Add spill information to plans for batch mode operators – showing how many pages were written to TempDB. This also means you *don’t* get that data before CU11.
- Lightweight query profiling hint enabled
- Log truncation may not work when you enable Query Store – the link currently 404s, but I’m sure that’s a temporary thing, like eventual consistency, but different
- Access violation when page compression is enabled
- Dump file when you run CHECKDB on a database with clustered columnstore indexes – also 404s right now. Look, it’s hard to publish web pages in the year 2018. You try getting your content past the marketing department, support, translations, etc. It’s like playing a video game and trying to beat all the different bosses. Sometimes, you’re gonna have to spend a few lives to get there.
- Instance shuts down when auditing is enabled – hey, look, it’s secure at that point, right?
SQL Server 2016 SP1 CU11
Only has 8 fixes, so we’ll list ’em all:
- Engine – Dynamic Data Masking doesn’t mask data when read by cursors
- Engine – memory leak in Windows Management Instrumentation (WMI)
- Engine – Database encryption keys longer than 3,456 bits now supported
- Analysis Services – circular dependencies on fields with the same name
- Analysis Services – Thai characters export incorrectly to a PDF
- Analysis Services – unhandled exceptions on MDX queries after XMLA queries
- Always On – backups fail on async secondaries
- Always On – access violations and unhandled exceptions with direct seeding
About that last Always On direct seeding one – the KB article has an amusing workaround:
Wait – to fix the issue, you should go back to CU9? Remember, we’re looking at a bug that is fixed in CU11, per the below screenshot:
That evidently means the bug was introduced in CU10. Just guessing here (a popular hobby these days given the sparse details in KB articles), but this automatic direct seeding improvement in CU10 might have been the culprit.
SQL 2016 SP2 CU3
Has the stuff in SP1 CU11, plus:
- Out of memory error even when there are free pages – okay, well, yeah, that’s not good
- Assertion error when you query sys.dm_exec_query_statistics_xml – I’m gonna be honest with you here, and this might be the port wine talking, but I don’t even remember that DMV
- Transaction delays on an AG primary – I’m genuinely delighted when only one AG bug is fixed in a cumulative update, frankly
- Workloads with frequent, short transactions may consume more CPU
- Error when you use in-memory transactions with server auditing – look, if you want things to go fast, you shouldn’t be auditing them
- Slow performance when you use NULL filters on a partitioning key – okay, seriously, why the hell would you partition on something nullable? I’m starting to think these bugs are your own fault, people.
- Problems with side by side instances on different service packs – #1: you shouldn’t be running multiple instances on the same server. That’s called a clown car, and it ain’t a good time. #2: different patch levels on the same box? Are you kidding me? You are lucky Microsoft didn’t just laugh and hang up on you when you made this support call.
- DROP_ASYMMETRIC_KEY causes a crash if it’s rolled back from a trigger – this combination of words is an abomination.
I love you, dear reader, but those of you who filed those support tickets, we need to talk at some point. You’re doing some very ill-advised behavior. Thank goodness Microsoft loves you enough to support your crazy activities.