« first day (514 days earlier)      last day (175 days later) » 

12:02 AM
it's for you, not me
it has your favorite actress
 
Oh dear lord
What a strikeout
 
@ErikDarling I don't get this comment
 
12:28 AM
The lead actress
Zero bueno
 
12:40 AM
Jun 9 at 14:08, by Erik Darling
irl jasmine is a megababe
 
12:56 AM
Kristen Stewart tho 🙄
 
1:12 AM
So contrary.
 
I'm utterly helpful
 
@ErikDarling did you let jafar ruin your enjoyment of jasmine?
 
1:44 AM
@JoeObbish I'm one hop ahead of the hump
 
@ErikDarling I got a query plan I don't understand
why is there a spool?
the target table has a clustered index on an IDENTITY column
 
2:10 AM
@JoeObbish I'll look in a minute. Mobile.
 
2:33 AM
dumb question: would SQL Server ever build a plan like that in order to minimize lock time on the source tables? Is that even built into costing?
 
@Forrest you think it's an odd plan too?
 
I can't figure out the reason for the spool, yeah
can you remove it with a USE PLAN hint?
 
the row size is a bit odd
 
HP is only when the modified table is also a source, yes?
 
@Forrest or it can't guarantee that it isn't a source, yeah
 
2:38 AM
compute scalars adding MAX types to row?
 
oh
nevermind
there's a scalar UDF in the query
I didn't notice because the query text is so long and poorly formatted
so my guess is that is the cause of HP
 
in other words, UDF = black box to optimizer, therefore HP?
 
yes
 
cool
 
I've seen unnecessary HP when using a MS-TVF as a source
I don't want to look more because this really pisses me off
also the query filled the transaction log somehow the first time
turned on ADR just for it
 
3:07 AM
hey did anyone notice there's a udf in the plan
> <QueryPlan NonParallelPlanReason="CouldNotGenerateValidParallelPlan"
well i'm off to bed
well done gang
 
3:53 AM
50
Q: Is there any benefit to SCHEMABINDING a function beyond Halloween Protection?

Paul WhiteIt is well-known that SCHEMABINDING a function can avoid an unnecessary spool in update plans: If you are using simple T-SQL UDFs that do not touch any tables (i.e. do not access data), make sure you specify the SCHEMABINDING option during creation of the UDFs. This will make the UDFs schema-...

@Forrest no
 
4:09 AM
"well-known"
 
4:20 AM
@ErikDarling here's a realistic function for your next demo:
CREATE FUNCTION dbo.ComputeNum(@i int)

RETURNS int

BEGIN

  RETURN @i * 2 + 50

END
hth
 
 
3 hours later…
7:02 AM
@JoeObbish Ah well it was written in the context of an ongoing HP conversation. The context was lots of people referencing that msdn blog post, so to that extent it is "well known" or at least "often cited". That's not the same as saying that anyone that doesn't know that is stupid or ill-informed, or new.
 
 
3 hours later…
9:51 AM
Your attrition rate is getting really bad when it comes to veteran users whom you've apparently made it your mission to alienate. That's a lot worse, because we have rather a lot to do with making this site what it is and in driving new users to it. You might want to take a step back and reconsider before you engage the nuclear weapons. Meta has always been a platform for people to share their opinions in a constructive way, and although you might disagree with it, that answer was neither rude nor inappropriate. Wrong answers can be downvoted, not deleted. That's always been the rule, Tim. — Cody Gray ♦ 9 hours ago
 
10:41 AM
@JoeObbish that looks incredibly agile
@PaulWhite I bet that comment gets deleted
 
I see 50+ comments. I guess meta has different policy than dba regarding comments
 
10:57 AM
23 hours ago, by Paul White
@George.Palacios Yes. Meta is for discussion.
 
 
3 hours later…
1:33 PM
mod justice was swift
 
let me know if you get another reply and I'll have a word
 
i'm not that worried about it, hahaha
 
1:52 PM
trying out some different configs for seq key
fewer cores, more threads, parameterized proc
for some reason things are "better" on the SSD bank than the nvme
which is weird but whatever
there's probably some dumb thing with the way my cpu talks to it
that glenn berry would know about
let's see how 400 threads goes
 
@ErikDarling better as in elapsed time?
 
2:09 PM
@PaulWhite back in 2016 people just knew more things
 
well they didn't have to learn R or java
 
today hasn't gone well
I've been awake for 20 minutes
 
go back to bed
 
@ErikDarling I'm not a sloth
I see tempdb.dbo.parallel_redo. Wonder how long that's been there
anyway, stuff failed and I don't understand why at all
while trying to figure out I seem to have run into a BMOR regression
 
> Hi Erik, thanks heaps for shopping with us!
what polite people
@JoeObbish how bad?
 
2:22 PM
@ErikDarling 30 minutes vs 8 minutes or so
estimating here
it very much does not help that there's something wrong with I/O
I'm getting ~600 MB/s
why
 
What is BMOR?
 
batch mode on row store
 
ahhh
 
SQL Server 2019
 
I am irrationally pissed off about this
that's a GROUP BY ORDER BY with COUNT(*)
only one distinct value
why shouldn't that query be able to take full advantage of query parallelism?
I'm seeing about one CPU second used per second with MAXDOP 8
do I totally misunderstand partial aggregation?
 
sigh
it's a partitioning problem
all of the rows are in one partition
 
is that luck or intent?
alright, with a 6 core vm and 400 threads, there's a ~10 second difference for seq key
 
well that's nice
 
yes
gonna rec something quickly
 
now all we need is some clue when to enable it for an index
 
2:43 PM
heh
 
and some info about what it actually does
 
@PaulWhite have you ever listened to pink turns blue?
 
why do they always seem to do stuff backwards these days
@ErikDarling rings a bell?
 
came on this morning
forgot about them kinda
good stuff
 
sorry I'm a bit distracted this morning by NZ playing Australia
 
2:46 PM
no worries
which sport?
 
did you notice how partitioning wasn't mentioned in the OP btw
 
ruggy, footy, cricky?
 
cricket world cup
 
@ErikDarling it's idiocy
@PaulWhite sounds like you got tricked 😎 OP-style 😎
 
@JoeObbish on whose part?
 
2:56 PM
not mine
 
 
2 hours later…
4:54 PM
my anime won't load :(
 
3 hours ago, by Erik Darling
go back to bed
 
multiple room owners telling me to leave
my feelings
 
seems like good advice
day going from bad to worse
 
found some new bugs in the app code too
which I suppose is better than not finding them
but it's Saturday
 
All code is unrealistic
 
 
5 hours later…
9:54 PM
What's the appropriate amount to charge for bad advice?
 
10k a day
 
I thought he already did that
"Hello Crunchyroll subscriber,

In compliance with tax laws, Crunchyroll will be responsible for charging state and local sales tax for purchases made by customers after August 1, 2019"
REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
 
Joe turned into Sean so slowly I didn't even notice
2
 
I'm moving to Texas where animu is still UNTAXED
FREE ANIMU
 
Don't forget to shoot your pistols at tornados and still your own hooch
 
10:22 PM
@ErikDarling What's the rate on good advice?
 
10:39 PM
@JoshDarnell depends on how much of it you want
 
10:53 PM
@JoshDarnell if you have to ask you can't afford it
 

« first day (514 days earlier)      last day (175 days later) »