The Heap™ – Consultancy ©®

12:29 AM

Huge earthquake hits Turkey and Syria, leaving more than 3,800 dead and thousands injured

Really jumped from the 1000 mentioned the other day

From what I can tell that will get a lot worse once they start clearing the rubble

swasheck

1:14 AM

there was a second quake (not an aftershock) that hit in a different location was what i heard

3 hours later…

4:43 AM

Yep, it's a complete disaster

4:57 AM

Over 4,000 dead in Türkiye including almost 1500 in Syria is the latest I saw

3 hours later…

John K. N.

8:09 AM

Morning

8:32 AM

Morning

2 hours later…

A: Why does sys.fn_xe_file_target_read_file require an explicit cast on datetime2 column?

10:50 AM

Erland Sommarskog asked me about this in May 2022. It's a bug exposed when a predicate is pushed down into the internal streaming table-valued function. The return type is indeed datetime2(7) but the source is a Windows FILETIME. The predicate is pushed down below the internal conversion, so the ...

I mean they gotta be aware of the bug if the documentation example has the workaround

11:05 AM

Yes, they are aware. I know this for a fact.

It's a really stupid bug with an easy fix.

I'm wondering if the predicate pushdown would speed up filtering of the files or not

They could also return the event data as XML. I usually copy it into temp table for conversion

I expected them to simply disallow pushdown for that column as a short-term reaction.

But yes, pushdown can be very important for performance.

Not only because you end up doing less work but because the repeated activity is more local, and therefore cache-friendly.

I don't know if there is a better way to query only increments. I haven't played with the concrete names and offsets that much

Probably not, the whole thing is pretty basic.

It's an embarrassing bug too. Imagine not testing this.

Q: What's an appropriate edit to an old post containing links to old documentation?

My expectations were low and I was still disappointed

3 hours later…

mustaccio

2:30 PM

See this edit for an example: a four years old question contains a link to the then-current documentation. The world has moved on since, and someone changes the old link to point to the new documentation instead. Is this appropriate? Would it be better to leave the old link alone (provided it's s...

mustaccio

2:46 PM

Hey, I was the first to post it here!

3:20 PM

mustaccio: faster than robots

@DanGuzman Yeah same — Paul White ♦ 5 hours ago

What's the "DL" Dan is referring to?

I assume the MVP distribution list.

@PaulWhite did that go away?

I find some of the ways people use logic on this site quite odd. A person asks a question on the internet. Rando answers question. Person says they don't believe them because another Rando said something else. So they both believe and don't believe? I don't get it.

imagine not knowing anything and having to parse conflicting information about the subject

3:26 PM

But why believe one internet person over another when you don't do anything on your own?

does not compute.

maybe they just want to know why the other person is wrong?

i dunno

me either, that's why I am asking random internet people :D

If you need more wrong people, just ask ChatGPT!

if i didn't know shit and had conflicting details where the choice would impact something important i'd probably ask too

@CadeRoux It's so bad. It's also super biased.

3:30 PM

Everyone starts from some basis for evaluating truth, and the goal needs to be to make that stronger and stronger so people can critical synthesize new information properly. My worry is not the ChatGPT produces wrong answers just like people, but that people are not thinking about the costs of evaluating its output just like you would any work product coming in from a first-time employee or whatever.

i had someone leave a completely wrong comment on a blog post yesterday

So you can safely accept what's good and discard what isn't. (And also decide to stop using it if the cost of using its output is excessive because it causes liability or usability or cleanup effort or whatever).

3:58 PM

@CadeRoux Well most of the people I work with ask the internet anyway and many of the kids just go on the web to ask it to do their homework... so really whomever controls that controls the "truth"

It's from something on the secondary replica. How you described transactions with AGs is not at all how they work. — Sean Gallardy 3 hours ago

lol

Yeah I'm done with that question

Basically sessions are broken, sql doesn't work, everything is a bug

I wonder if you end up having to edit ChatGPT instead of your own first draft, do you really save time and effort over all? There is the need to read and process someone else's work and critically analyze it. That's still a significant cognitive load.

@SeanGallardy Obviously. Almost goes without saying.

As evidenced by the state of predicate pushdown on sys.fn_xe_file_target_read_file.

4:14 PM

I'd fix that for you guys but I don't mess with the QE code

Are you still allowed to touch SQL Server code? I thought you switched teams out of utter disrespect for all of us.

Yeah I still make changes, fixed the dm_server_services DMV in 2019 CU19 that should roll out soon

It's also fixed in 2022 RTM

I tried to get some new items for AGs in, but it was kind of a larger feature :(

but yeah, I'm 100% cloud now... don't do it, it sucks

4:33 PM

@SeanGallardy Nice.

4:44 PM

@ErikDarling Where is it?

erikdarlingdata.com/…

that requirements list is uge

whose fault is that

5:17 PM

@ErikDarling lol wow.

yeah so like

imagine being wrong

posting the wrong

on a video that is right

so like you can't be bothered to watch the video or read the documentation

or even test it yourself

you're just sitting around being wrong

forever

@PaulWhite you might like this: youtube.com/watch?v=vN6gmfuJ02s

@SeanGallardy FWIW, I think the latest version of sp_WhoIsActive does show sleeping sessions now. But Erik probably would know better about that.

buddy.

@J.D. You have to add the option IIRC

EXEC sp_WhoIsActive @show_sleeping_spids = 1; has been around for as long as the proc has existed

5:32 PM

Idk unless I'm on an older version. But using it right now parameterless seems to show me sleeping things.

Ah yes @show_sleeping_spids = 1 is the default in the version I'm on

Per the comments, it's only if they have open transactions too. Probably fair lol.

--Controls how sleeping SPIDs are handled, based on the idea of levels of interest
--0 does not pull any sleeping SPIDs
--1 pulls only those sleeping SPIDs that also have an open transaction
--2 pulls all sleeping SPIDs

Dejavu

Reading documentation is hard

It's got letters and stuff

Truth

big words

5:39 PM

I prefer small numbers

TWSS

5:52 PM

Speaking of WIA, is the wait_info column still updated / valid for a process that's being rolled back or in that case is it something like the last wait before the process was rolled back?

6:03 PM

i don't know, i don't roll things back

ever forward

Worddd

Q: Need some help with SQL randomly freezing

SQL will freeze for about 5 to 15 seconds. There are a couple of errors I see but the following is the most frequent. Any suggestions? SSPI handshake failed with error code 0x8009030c, state 14 while establishing a connection with integrated security; the connection has been closed. Reason: Accep...

sql-server

i don't think this is easily solvable as q&a

6:28 PM

Based on the latest information from OP, agreed.

6:38 PM

well if someone wants to say it nicely go for it

i'm gonna stick a 🧊 vote on it

Hannah Vernon

7:07 PM

I'd guess blocking off the top of my head. The vendor probably has never heard of optimistic concurrency, indexing, or load testing at Internet scale, lol.

7:36 PM

@HannahVernon probs, with a side of just poorly written code I'm sure.

3 hours later…

10:16 PM

So I've been looking for a good way to automatically correlate the User who made the change to the history in a Temporal Table's history table. I found adding a computed column to the source table using the SUSER_NAME() function does persist the correct User to the history table of whoever executed the DML statement.

Seems too easy, why is this a terrible idea?

billinkc

If you add a column to the table, does it still work? I think you had to add to history and then the base table if memory serves