« first day (4194 days earlier)      last day (51 days later) » 
06:00 - 15:0015:00 - 22:00

6:14 AM
6:54 AM
7:57 AM
an aircraft skeleton, from the long extinct Aircraftsaurus Rex
I didn't know that if you zoom out far enough on Google maps, it shows you daytime/nightime
1 hour later…
9:07 AM
A chairde - Morning all!
1 hour later…
10:25 AM
@Zikato Google maps or earth? I just tried it with maps in a browser and couldn't zoom out far enough
@PaulWhite Google Maps. I've clicked on the link to Aircraftsaurus Rex from @ypercubeᵀᴹ
Still doesn't work for me from that link. Oh well nvm
Google Earth does show light/dark, but it's not accurate to the time of viewing
you need to click on the bottom left (Layers or whatever) -> More and check "Global view"
to see the round earth - if you are believer ;)
but I still don;t see night/day either
10:34 AM
that's pretty well hidden in the more... menu
also to get accurate day/night, I have to zoom one more step than the full globe
suddenly changes from a standard angle to actual day/night
my view
ah true!
looks pretty flat tbh
turtles not shown
Perhaps they are racing Achilles
heh, didn't realize they also have maps of several planets and moons as well
1 hour later…
12:07 PM
the assumption that removing indexes will dramatically improve write activity is so weird to me
for most oltp-type singleton modifications the impact is in the microseconds
i wonder if this is the same spinning disk mentality that leads to people freaking out about fr*gmentation or p#ge spl!ts
12:28 PM
it's true that in a sensible system with sensible resources, a sensible number of secondary indexes isn't a problem
moving away from the sensible to the dumb or extreme can produce problems
unfortunately, people and clouds excel at both
also, people often over-emphasise the value of covering indexes over a point lookup
I guess I'm saying it depends
but most of the time, a handful or half-a-dozen secondary indexes don't indicate a problem in themselves
i don't usually see much of an impact until you get to the 10-15 index mark
you must work with sensible systems with sensible resources
I think it depends in the table design, the index design, and the write frequency requirements.
10-15 isn't super-unusual, but I'd wonder if all of them were strictly adding value
I've seen tables where one index is one too many.
12:41 PM
@PaulWhite they aren't; plenty of unused and merge-able indexes around, but they were able to keep adding on to that point without cleaning up because there was nothing noticeable happening
And of course, tables with an index per column with 30 or 40 columns is just asking for an issue.
@ErikDarling For sure you can get away with it, up to a point
If you're doing low-volume 'OLTP' and/or your log, memory, disk can keep up with the required I/O
yeah, i just can't imagine a transactional table's existence being greatly improved by going from (likely) 1 pk/cx and 3 nc to 1 nc.
yes you'd need to be very close to the edge for that to be important
i do sometimes wonder if merging includes to make generally wider indexes would lead to more deadlock likelihood
12:45 PM
deadlocks are a meme
as are includes
but seriously, anaemic cloud I/O can be a factor
both for log and main db
i would like to start a class action suit against the cloud et al for that
log governor
@HannahVernon 30 or 40? you're not even trying
but yeah indexes added to support catch-all queries are the worst
though then we're talking a hybrid workload at best, not OLTP
If a table has over 250 columns, no amount of indexes will provide a good coverage
one nonclustered columnstore index
all you'll ever need
12:51 PM
I dunno, I rather solve the bad design than attempt to hide the problem
the debate reminds me somewhat of the argument that heaps are 3-4x better than clustered indexes because lookups are direct rather than requiring a b-tree traversal
@Zikato lol normalization
good luck making 10 new tables and moving the data over and getting the app to use them
Let's normalize normalization
well you can spend a day carefully designing the right indexes and constructing routines to build optimal, and safe, dynamic SQL
too much work
12:53 PM
or you could add a NCCI
I'd rather not encourage that behaviour
if you have that kind of time on your hands and people will let you do it, god bless
horse, water, but no drinking
redesigning schema in a 10+ year old app ain't happening most of the time
12:56 PM
usually because the tooling sucks
Speaking as someone who has implemented quite beautifully optimal designs to support kitchen-sink queries in the past, I wouldn't bother 90% of the time these days
with which part?
I'd just add an NCCI and send the invoice
Isn't consultants' job just to fight the fire and move on?
2 days ago, by Zikato
If you're a startup, you don't have the money to pay experts. When you are successful, you focus on expansion and not tech debt. It's gotta take a near-death experience to make it a priority
12:57 PM
oh yeah
Z giving me whiplash
@Zikato not when I do it, can't speak for everyone
the beautiful design needs re-evaluating whenever someone wants to search on a new field; the NCCI just continues to work
@Zikato yes; this week i turned on rcsi and used query store to force serial plans for 3 queries rather than try to fix a bunch of bad design choices and dynamic orm queries
most of the time, it's client-driven. many want the quick and dirty fix
12:59 PM
gotta stop the bleeding
Can't argue with that
i'm usually working with a few execs/managers and a bunch of tech staff watching expecting a miracle immediately
well that's what you get for charging $1k/hr
ha ha, that's just what i pay you
oh right
> expecting miracles on the cheap
1:01 PM
white privilege©
I get that from a consultant's perspective but from employee's perspective wouldn't you want to improve the design because you're gonna stick around for longer?
Or do you just change jobs frequently enough to outrun the tech debt?
@Zikato i hope to never find out
@Zikato IME most employees capable of that level of work leave to become consultants so they have more freedom and pay
in theory, anyway
working for people sucks massively
the best you can do is keep the experiences short
1:03 PM
most of my FTE work was supporting 3rd party apps where i couldn't redesign anything anyway
And then those experienced consultants just implement NCCI everywhere and move on
when it's the optimal solution, yes
just getting batch mode is enough half the time anyway
I had a few edge cases when batch mode went crazy, but generally yeah
1:04 PM
like yesterday it got a query from 50 seconds of cpu time and 12 seconds duration to 1.4 seconds of cpu time and 400ms duration
it's funny because the initial expectations were that CS storage would provide most of the benefit. in practice, most comes from RS batch mode ops on modern hardware
i do like cs storage in many scenarios
it's a bonus
but it many cases batch mode RS scan/seek is close enough
for true 'OLAP' CS is important
but most workloads are hybrid to some extent
witness all the demos people had to rewrite when RS batch mode serial queries came into being
i'm sorry you must be talking about REAL TIME OPERATIONAL ANALYTICS
1:08 PM
it's a nice meme, but it's surprisingly tough to find a scenario where NCCI isn't truly optimal for that scenario
not just because it's quick and easy
@ErikDarling you broke my meme-o-meter
should we demand hybrid indexes
you could argue filtered NCCIs are such
they don't support point lookups so well
I should probably experiment with them more. I'm not sure about the maintenance and tables with not enough rows
If batch mode is the main benefit, there's still that trick with the filtered NCCI or dummy LEFT JOIN
i just hit them with a weekly reorg with the compress all row groups option
1:14 PM
@ErikDarling not many kitchen-sink reports look up a single row?
that's generally good enough as long as you don't have massive daily deletes
@PaulWhite that's a trick question
but it's not like you can't have a couple of RS b-tree indexes as well if you want
@ErikDarling best kind
Thank you both for the tips. Even though I'm more on the idealistic/purist side, I'm still open-minded
I'm an idealistic fascist dictator and I still prefer NCCI
Erik tends to favour clustered columnstore on hekaton
The turning point for me was an instance where the correct 'purist' design was slower and larger than the NCCI and couldn't cope with future changes
Plus I really hate dynamic SQL
Why not both? Good design and NCCI? Yeah, I know…
1:20 PM
this guy
The Heap™ – Consultancy ©®
Where dreams are broken
Well a decent design is important of course
But don't make me tap the sign
changing a bad design to a good design is often a months to years long process that no one wants to take on
I still blame tooling
and there's no guarantee it'll be better aside from making you feel superior for doing things right
hitting large tables with unpredictable search and retrieval patterns is still a generally bad fit for row store indexes, even with paging in place
1:22 PM
It's funny because non-db devs are often big fans of an API
then refuse to code against one for the database
Everything's so tightly-coupled it's insane
In principle, the db design should be entirely an internal concern
code-first web devs being the worst
Low cohesion, tight coupling
Putting the F in EF
@ErikDarling one might say especially with paging
People do love solving the same problem over and over again
@PaulWhite Which API do you mean?
@Zikato Stored procedures and views for the most part
The external developer's contract with the database
@PaulWhite people love having the same problem solved for them over and over again
1:29 PM
Right, wanted to be sure.
Rather than direct table access
@ErikDarling and will pay invoices for the privilege.
In conversations I had, Developers don't consider it an API. And SPs can have weak contract if designed badly - changing the number of result sets
Many things can be done badly
chat interfaces for example
some of them will double bonk you when someone edits a message you got tagged in
Duplicate voting
1:31 PM
I see Google renamed their chat product again
@ErikDarling sounds annoying
i only saw they were nixing hangouts
It's not Hangouts anymore?
Apparently, it's a different product.
They discontinued the Hangouts and introduced Google Chat
Not even Google wants to do refactoring
Always was
1:34 PM
building a new product was probably easier than fixing poor design choices in the old one
I feel disenchanted today
Looking forward to Gmail attempting to force me to use it
@Zikato welcome to the modern world
@Zikato if you want to do things right, build a new product
@PaulWhite Shouldn't that come with alcohol?
well mine does
1:36 PM
@PaulWhite don't talk about me like i'm not here
@ErikDarling I was at 3 companies that tried that instead of the refactoring. It's always called something like Phoenix or Lazarus. They hire double the people - the old guard to support the old platform and new ones to build the state-of-the-art.
Huh. I just realised I don't have a bottle open
and after 5 years it's not even at 20% parity
the absolute state of the art
I heard it has cloud and microservices!
1:38 PM
sometimes good is the enemy of mediocre
I hate the mediocre
Tolerating it requires significant exchanges of money
mediocrestore indexes
mediocrestore enhanced heap - meh for short
I blame capitalism
Huge mistake
We should hire new people to work on a new system while the old guard supports the capitalism
1:40 PM
ETA 6-8 weeks
someone's taking their jacinda pills
Never met the woman
well i never met dr. pfizer
I had to google that. I thought it was some herb
but I hear she's a rabid capitalist
@Zikato jury's out on that
1:43 PM
Speaking of capitalism, have you played the Suzerain game?
the only game paul plays is windbg
I spend all my time trying to accumulate dollars
4 mins ago, by Paul White
I blame capitalism
I do wonder if it is possible to win dbg
1:45 PM
@PaulWhite i hear seo is good for that
double ping!
> double bonk
@ErikDarling is SEO the ultimate meme
seems like an absolute racket to me
I pity the poor soul that rejoins the room and sees 100+ messages
1:47 PM
I imagine one could waste a lot of money 'investing' in SEO optimisation
i used to laugh at the bou folks for adding "in sql server" or spelling things out to the max but then i ended up going through 900+ posts on my site to add the same seo memes to headlines and post bodies etc
An editor tried to SEO one of my articles once
Your content started in paragraph 5?
It's a game where people pretend to have insight into search engine algorithms
They'll be selling NFTs of crypto apes next
@Zikato It was worse than that. They tried adding stupid and irrelevant hyperlinks everywhere and all sorts. It was stupid
Why do you need insight into the search engine algorithms? Just vomit every possible keyword into your text
1:50 PM
Focussing on stupid metrics like raw click numbers. Even if every click pisses off the clickee so much they vow never to use your product
To be fair, I only have about 20 visits a day
Ah but who are those 20 people
you have a lot of family members eh
I don't know, I try to respect their privacy
I bet they're all top-flight influencers
1:51 PM
It's probably all me, I don't know if I've set up a filter
Movers & shakers
small businesses with critical sql server issues
and vendors
ready and willing to pay any invoice to solve their problems
well, one is definitely Kevin Feasel, because he features me on CuratedSQL
1:53 PM
what a nice lad
Erik pays protection money to the Brent Memeletter
mama, I paid it
ha ha
there are some weeks i can't believe that i don't make it in there when i see the other links
it's gotta be balanced. It can't be all Erik Darling show
1:55 PM
Well it could be, for a price
Capitalism strikes again
grabs steering wheel
Testing markdown
get in the limo we're doing capitalism
@Zikato Now do strikethrough
1:57 PM
~challenge accepted~
a set of eyebrows appears
I just wanted to decorate that text
thought as much
Strike one
that's much better than most people manage
we've seen some terrible flapping about in the past
<stike>oh no</strike>
1:59 PM
I dedicated time and effort to finding out. And I'll probably never use it again.
yeah <stike> doesn't work at all
imagine what you just forgot to make room for that info
flapping continues
@PaulWhite probably my kids' birthday
2:01 PM
I finally know what that means
@Zikato they were all born on the same day?
well I don't know now, do I?
but you did use the singular
who put the bonk in the bonk bah bonk bah bonk
so whichever day it was, it was the same
2:03 PM
I gotta be honest, my English not very good
it's better than most in this room
no apostrophe abuse
apostrophe abuse is rampant in here
wait he hasn't experienced sean
* fewer
@ErikDarling no one has, despite his outlandish claims
2:04 PM
apostrophes are one of the harder ones for me. Missing articles is another one. Doubled letters and commas in a sentence
also, a blogger whose rss feed i follow has been posting about "CTE's" a lot recently and i cringe every single time
CTE's syntax?
there are some complex rules around the plurals of initialisms, but that's not a tricky one
my kid's birthday or my kids' birthdays, for the record
I'm an idealistic fascist dictator grammar nazi and I still prefer NCCI
@PaulWhite Thank you
2:08 PM
i always have to validate how to apostrophe a word that already ends in s
I could have gone for triple boink
Automatic kick for a triple
@ErikDarling both options are acceptable in most cases?
Physical or out of the room?
Like formatting parentheses
@Zikato depends on what rank grammar nazi you are
2:10 PM
@Zikato Physical is an option, but for that I have to pay Erik's invoice
@ErikDarling It was my job at one point. But for my native language
Czech please
true apostrophuhrers have different standards
V jednu chvíli to byla i moje práce, ale pouze pro mou rodnou řeč
sounds right
2:12 PM
that's how i would have said it
> ‎At one point, it was my job, but only for my native language.‎
Results of in-line translate
I should just use that instead of trying
might be funny for a while to pretend not to understand anything Tomáš Zíka says
it occurs to me that many people in this room speak more than one language
2:16 PM
some speak fewer than one
not counting Canadian
ESL gang unite
English as a Second Language - ESL
2:19 PM
nope sorry can't make it out
translate not working
no worries, it's probably my accent
oh wait, paul prefers his aliases differently
a second language = english
looks backwards
you're just upside down
stand on your head and drink wine
2:21 PM
it was not syntactically correct. I was missing brackets
the level of hemisphereism is too damned high
you started it with the whole "wrong season" meme
benefits of blue text, I need not be consistent
anyway when does Winter Bash start
seems late this year
i wonder if todd cares about his domain and full name being posted online
Typical StackOverflow experience. First, they complain you don't give them enough info, now it's too much info…
2:29 PM
i wonder when one box support for tik tok videos will come to chat
Maybe it's obfuscated so well that's not even his real domain and name
i was just listening to a podcast about how government hackers will find linkedin profiles of db/sys admins who work places they want to get into and then find forum posts by them asking for help with stuff to recon their infrastructure
they'll even answer questions for them in ways that make their domains unsafe
anyway it was amusing
i wish todd all the online safety in the world
meaning you'll delete the picture in chat as well?
too late for that
we're all being spied on through our coffee machines anyway
2:33 PM
@ErikDarling where's that from?
Q: Cannot Backup to Network Share using Ola Hallengren's Scripts

LingerSQL Version: Microsoft SQL Server 2016(SP3)(KB5003279) - 13.0.6300.2 (X64) OS Version: Windows Server 2012 R2 Standard I am trying to setup the backups and database maintenance jobs using Ola Hallengren's sripts. I have the following job setup: sqlcmd -E -S $(ESCAPE_SQUOTE(SRVR)) -d master -Q " ...

serious question: can those images be deleted from the stack hosting?
only usually bother for extreme cases though
people posting porn and whatnot
2:37 PM
And if you're feeling generous, and do the obfuscation yourself instead of the OP. Can you edit the pictures?
you can edit the question to replace the image url
there's a lot of text info in the question
no indication Linger cares
we do redact passwords and stuff
on occasion
it's a PITA though
and moderating is unpaid job, right?
partly b/c no one does it often enough to remember exactly how
utterly unpaid
another failure of capitalism
keeps SO viable as a company tho
I think that's success for capitalism tbh.
2:41 PM
from a certain point of view
well, I gotta bounce today. Thanks for the chat!
did anyone understand a single word Zikato said?
dunno, didn't really read
maybe if chat had a twitter skin i'd read it
2:46 PM
If any of the information in the question being public is of concern to you, please edit it out — Paul White ♦ 5 mins ago
what a mod
sends invoice
to todd?
to everyone
community service
lemme know how many paddocks i owe you
2:53 PM
06:00 - 15:0015:00 - 22:00

« first day (4194 days earlier)      last day (51 days later) »