The Heap™ – Consultancy ©®

Evan Carroll

12:03 AM

Kill the Winter, you have my blessing.

sp_BlitzErik

check out this eco terrorist

probably likes global warming

Joe Obbish

12:50 AM

NonParallelPlanReason="NoParallelPlansInDesktopOrExpressEdition"

oh my

hot2use

7:37 AM

Good morning, evening and day. Enjoy your lunch, breakfast or dinner.

@sp_BlitzErik hope you have a fantastic presentation

ypercubeᵀᴹ

Is SQL Server relevant in this context? Or put differently: How does the implementation of stored procedures (we call them sprocs) in SQL Server affect the future implementation of "stored procedures" in PostgreSQL? — hot2use 35 mins ago

@hot2use do you really call them "sprocs" when you talk about them?

hot2use

@ypercubeᵀᴹ in SQL Server? Yes. :-)

ypercubeᵀᴹ

I have a mild throw-up reaction whenever I read that word.

hot2use

How come?

ypercubeᵀᴹ

I just prefer the whole term.

hot2use

7:40 AM

:-)

A moderator might have to modify my comment then to "I" instead of "we".

ypercubeᵀᴹ

I've seen the use in dba.se posts often enough to know that it is used.

What I don't know is how common the use is in oral.

hot2use

Sorry for that. I thought it was commonly used by the majority of people.

ypercubeᵀᴹ

No need to apologize, really. My knowledge of English was mainly from books, Internet and movies, until recently. And the recent years working in English speaking environment was not related to SQL Server at all.

hot2use

Ok

Is this one worth a rollback and re-write?

dba.stackexchange.com/posts/194943/revisions

No capital letters at the beginning of sentences. Sizes formatted as code. Product names formatted as code.

ypercubeᵀᴹ

7:57 AM

Yes, the edit was rubbish.

hot2use

8:08 AM

Hope it's better now.

0

Q: Conditional MS SQL replication

I have an issue that I want to do conditional replication to one of my enterprise level client databases. This database is very large: approximately 300 GB MDF file and a 160 GB LDF file because of archiving and accountability. My primary DB has lots of accountability data and I want to transfe...

Tom V

@ypercubeᵀᴹ Me too, if I'm editing a post and I see sprocI usually change it to stored procedures

McNets

8:36 AM

Morning

Tom V

hackernoon.com/…

4

Marian

morning

Paul White

9:08 AM

morning

Royi Namir

9:25 AM

Hello. and good morning. I might be asking something which is more a infrastructure decision as opposed to sql server. currently in our company , when we upload a new version ( C# + SQL) - we brings many workers to check the production.
But I said to them why not duplicate/clone IIS + SQL server a month before version upload so QA can test a "cloned-going-to-be-production" environemnt ?
with C# there is no problem. I can just copy the code to the server. But the problem is that throughout this month - the real production's db got new data becuase of the actual company's work. So my questio…

Tom V

@RoyiNamir You can just test on the month old copy and then move only the code to production?

Royi Namir

@TomV Thanks for reply. while this month , there were many garbage records

and dont forget that schema has changed

becuase the new-going-to-be-production being tested ( let's call him child to be adult) has changes as ooposed to the adult

a new features was added to the child ( which is going to be adult next month)

(sp added ,triggers etc)

Tom V

10:00 AM

@RoyiNamir How do you manage schema changes? you just apply them to prod just like you did when you created your test environment,

Colin 't Hart

@RoyiNamir Well, yes, you need to make scripts to upgrade your database. Code is stateless, but anything that has state needs to be altered to conform to the new state the code expects.

Royi Namir

10:23 AM

There's no problem with schema scripts. what about the data ?
1) data for the whole month should be deleted ( becuase it's testings)
2) Delta of data from adult should be moved to child ( including right identity values etc)

@TomV there are many basic assumptions which makes it vulnerable at first.

A secured site should not allow any javascript to be able to read cookie

it can be done via the "httponly" flag

also today there are CSP ( content security policy) that can restrict running scripts which are not in your server

Not to mention that cookie is subject to CSRF attack ( which can be fixed by a hidden field). If you want I can explain it in more details

Colin 't Hart

Why are you doing it this way around? Why does QA need a month to test?

And you should never promote a test version to be production.

I always run the same scripts on production, or on a fresh copy and promote that.

Tom V

10:40 AM

yes exactly

1) restore database from prod to test
2) run upgrade scripts on test
3) test
4) promote code to prod and run upgrade scripts on the database

Tom V

11:33 AM

@RoyiNamir forgot to ping you ^^

Colin 't Hart

12:38 PM

@RoyNamir It is also possible to use logical replication techniques and tools to keep a copy in sync with a master.

But it's pretty dangerous to have test data which you need to delete before going live.

Unless you have rock-solid multi-tenancy already working.

SAP can do that for example.

sp_BlitzErik

@hot2use thanks!

Tom V

1:03 PM

@Colin'tHart Do you have experience with that?

Dynamics AX could theoretically do that too with multi-tenancy but I don't think it's very safe

I assume they create a new "tennant" then for testing and remove it again?

Colin 't Hart

@TomV I have no experience, with that, no. I was always working in an Oracle team. The SAP team(s) had their own DBAs and they could do it. Yes, they had testing "tenants". No need to remove them, they can just continue to exist.

Most/all of their code promotions were within the SAP environment too.

Tom V

They probably test new parametrization that way (ie. Changing a parameter) but I would be surprised if they used that for code/application changes

Colin 't Hart

1:21 PM

I'm pretty sure they can deploy code that way too. At least the customisations written in the SAP procedural language(s).

Tom V

@Colin'tHart That's neat

Colin 't Hart

I forgot the term they use for promotion / deployment.

Tom V

@Colin'tHart multi-target application?

Colin 't Hart

@TomV Transport.

You make a "transport request" to move a version of code between environments.

See blogs.sap.com/2013/06/15/…

hot2use

@TomV SAP has a built-in transport mechanism

Tom V

1:27 PM

Looks neat

hot2use

Changes are first introduced in a Test instance (T) then transported over to a Quality Assurance instance (Q) and then when everything has fully been tested they (SAP DBAs & SAP Technology Experts) will transport everything over to Production (P).

Tom V

But that isn't a testing "tenant" that is a testing environment

hot2use

In the old company I worked for you would therefore have a P10, Q10, D10 and T10 environment for example. Multiple application / database tiers for each environment.

True.

Tom V

Then it works the same for us except it's powershell scripts instead of a gui

hot2use

Sometimes they would add a S(andbox) instance for security incidents.

Pretty expensive at times.

Tom V

1:36 PM

Yes, 4 or 5 environments isn't exceptional

We typically have DEV > TEST > UAT > BUILD > PROD, and sometimes a SANDBOX which is an exact copy of prod for debugging things we can't repro otherwise

For customers with a TB+ database that's quite a lot

Paul White

2:31 PM

Meh my phone could run those instances

5

gbn

2:47 PM

@PaulWhite My new Honor has 6GB RAM which was quite good 15 years ago in a DB server

Joe Obbish

@sp_BlitzErik if the MSFT van gets me you can have all of my unicorn points

sp_BlitzErik

@JoeObbish don't worry, i hear they're quite humane

most of their techniques are undocumented

Joe Obbish

@sp_BlitzErik I'll only let Joe Sack take me in

Paul White

You know Joe "the Sack" got that name from his interrogation techniques eh

sp_BlitzErik

(in bed)

Joe Obbish

3:02 PM

not family friendly

Paul White

Well that's not at all what I had in mind @sp_BlitzErik

sp_BlitzErik

deal with it

Paul White

sunglasses

Joe Obbish

@sp_BlitzErik I WILL

→ 1 message moved to Trash

Paul White

Oh hang on that was oranges in a sock I think. Perhaps sack is a typo.

Joe Obbish

3:07 PM

@PaulWhite well you could throw someone into a sack before throwing them into a van

or put a sack over their head

I thought that's where you were going with it

Paul White

← 1 message moved from Trash

@JoeObbish That works too

Tom V

@PaulWhite Maybe his sack looks like oranges in a sock

Paul White

@TomV eeek

Tom V

I figured that would stop anybody from taking it any further

Paul White

That should do it yeah

hot2use

3:10 PM

tiltling head to the left and hitting palm of right hand on right side of head

Nope, doesn't work.

sp_BlitzErik

where's joe the prude now?

Joe Obbish

@sp_BlitzErik you act like you live in a city with no community standards

Paul White

He's hit the sack

sp_BlitzErik

i don't follow

ConcernedOfTunbridgeWells

@TomV Meh. 14. Uphill both ways etc.

You can never have too many environments.

sp_BlitzErik

3:20 PM

@JoeObbish i'm doing the links this week and including your trace flags post

Joe Obbish

@sp_BlitzErik Brent on vacation?

sp_BlitzErik

downside: you're gonna have to actually figure out what they do

brent is busy

too busy for links

Joe Obbish

can I just ask Paul?

sp_BlitzErik

ideally

Paul White

Some of those I've already explained elsewhere

Not going through them line by line

Joe Obbish

3:22 PM

were any of them new?

Paul White

couple or three, but none where you've said what they do

oh that's not true - the optimizer statistics element one was new

sp_BlitzErik

"i have discovered a number"

2

Joe Obbish

I got one!

well if you give me the numbers I can try to figure those out first

sp_BlitzErik

12345

Joe Obbish

@sp_BlitzErik that's not even a valid trace flag

Paul White

3:28 PM

2368, 2418, 8678

all three I had listed with ??? as the note against them

I have lots of those :)

Joe Obbish

":)"

sp_BlitzErik

what does the rest of your list look like?

Joe Obbish

subtle

sp_BlitzErik

is there a trace flag that disables trivial plans?

Paul White

9341, 9384, 9390 were new

@sp_BlitzErik 8757 (sqllang!COptContext::FTryTrivialPlan)

Joe Obbish

3:33 PM

@sp_BlitzErik yeah

sp_BlitzErik

globally?

Paul White

@sp_BlitzErik Sure, if you set it globally

sp_BlitzErik

sweet

i hate those things

always monkeying my day

Joe Obbish

trivial plans also make it difficult to find trace flags

Paul White

I like trivial plans (old columnstore issues aside)

Joe Obbish

3:35 PM

@PaulWhite I kind of regret not doing a better job with documentation, but if they're out there then maybe others can look into them and discover something of value

sp_BlitzErik

have you ever seen this?

github.com/AaronMorelli/SQLServerTraceFlags/tree/master/…

Paul White

Yes that's one of the better lists

Some of the ones you have listed as unknown are quite easy to figure out.

sp_BlitzErik

i hate the way it's organized but yeah

Joe Obbish

Doesn't Konstantin have the best list?

@PaulWhite I imagine so...

sp_BlitzErik

dunno if it's the best

maybe you should make the best list

Paul White

3:39 PM

You could start with 7356, 8750, 9164, 9165

Joe Obbish

@sp_BlitzErik it seems to have the most and he still updates it

it'll be up to 585 soon

@PaulWhite I appreciate the hint

@sp_BlitzErik I wonder if Paul named someone in his will to get his trace flag list

Paul White

@JoeObbish Link please

Joe Obbish

@PaulWhite github.com/ktaranov/sqlserver-kit/blob/master/…

Paul White

@JoeObbish Thanks I didn't recognise the name

Joe Obbish

I should have said it was the one published last year on SQL Server Central

sp_BlitzErik

3:51 PM

ooh i contributed to that one

Joe Obbish

@sp_BlitzErik naturally

Paul White

4:05 PM

I'm surprised 8750 is missing from that list

sp_BlitzErik

go on

Joe Obbish

looks like that's missing from all the lists

Paul White

I'm sure I've mentioned that one several times. It skips stage 0 (transaction processing) of query optimization.

sp_BlitzErik

if you've mentioned it, google isn't aware of it

Joe Obbish

not mentioned in chat either

sp_BlitzErik

4:16 PM

Joe Obbish

maybe we should start translating TF numbers to hex so MSFT can't find us

sp_BlitzErik

just write them backwards

Jack Douglas

@sp_BlitzErik your quotes in the wrong place aren't they?

2

sp_BlitzErik

never

Jack Douglas

that downloads a zip file which I'm not brave enough to open

sp_BlitzErik

4:22 PM

you must have used bing

Jack Douglas

you jest

sp_BlitzErik

ah, duckpile

Paul White

@JackDouglas Well done, sir.

The zip contains demos from that session where 8750 is used.

sp_BlitzErik

a+ would unzip again

Joe Obbish

looks like we got a new preferred trace flag search engine

sp_BlitzErik

4:30 PM

seriously

Paul White

Not the most exciting flag in the world, but there it is

sp_BlitzErik

i'm thinking about changing the way blitzcache flags trivial plans to only flip the public bit when there's an index scan

trivial plan with a seek is whatever

it's already only collecting them when there's a parameter list

Paul White

why does it alert for a trivial plan?

(not a trick question)

Jack Douglas

4:53 PM

"I’d like to hear SO’s rationale behind removing the “Oracle Database” option for the database question. It’s still the top-ranked database in the world, even if it’s not the top database with questions on SO."

sp_BlitzErik

@PaulWhite because we've run into a plan being trivialized when it shouldn't have been

like EF app running select (every single column) from table where (non pk/cx) = @p0

trivial plan but huge missing index when opt is full

queries like that run thousands of times a day

and no missing index req since it's tricial

etc

Joe Obbish

5:17 PM

> be me

bluefeet

So as of this morning, Oracle is ON THE SURVEY. There was an error on our part and we apologize. It was absolutely not purposeful or malicious, especially since Oracle is a client of ours. I am going to be handling the analysis of this question carefully (looking at the responses from the 1st day separately from the rest) to be as accurate as possible. — Julia Silge 1 min ago

Joe Obbish

> get 27 seconds of PAGELATCH_UP waits per second

> efw

Paul White

5:29 PM

@sp_BlitzErik Oh I see so it's mostly for the missing index recommendations?

Jack Douglas

@bluefeet yippee, thanks :)

how about SQLite now ;)

sp_BlitzErik

6:15 PM

@PaulWhite the non-pauls of the world need these things

four non-pauls

Joe Obbish

it's weird to see system processes above session id 50

Evan Carroll

jesus christ this survey is long

I need a beer

the ethics questions and ai questions are boring af.

that survey should be somewhat opt-in on segments, those are complex topics they're not easily reduced to questions and those questions were super poorly worded.

Joe Obbish

7:10 PM

@SeanGallardy-Microsoft @sp_BlitzErik I don't know if you guys remember, but I got that detailed information on PATCHLATCH_UP waits finally

was there something that you wanted to see in particular?

the symptom is that we see around 25 seconds of PAGELATCH_UP wait type per second when running our workload

sp_BlitzErik

what info did you collect?

hot2use

@JackDouglas I wonder if she's relate to Lou Ferrigno (The original Hulk)

Joe Obbish

@sp_BlitzErik got the pages that the latch wait were on

sp_BlitzErik

i'm excited!

what were they?

hot2use

@JoeObbish how so?

Joe Obbish

7:20 PM

total_wait_duration_ms Page Type
645052 Stage PFS
128863 Stage PFS
110998 Stage PFS
102138 Stage GAM
75818 Stage PFS
56747 Stage PFS
53829 Stage GAM
51253 Stage PFS
48395 Stage PFS
48302 Tempdb GAM

for the top ten

that's sampled data of course

@hot2use normally all system processes have a session id <= 50 in sql server

hot2use

@JoeObbish Yes, I know, but no guarantee. That's why the sys.dm_exec_sessions has a column is_user_process (1=user; 0=system) nowadays.

Joe Obbish

@hot2use well it's my first time seeing it, in any case

do you see it often?

hot2use

I meant more: How did you achieve that?

Me? No.

Joe Obbish

oh I see

the server has 96 schedulers. I assume that is related?

hot2use

Hmmm. Yes, could be a good reason.

Forrest

7:29 PM

How did you get page types? An XE?

Joe Obbish

@Forrest DBCC PAGE :(

Forrest

Heh, so a lot of legwork then. How did you get the specific pages that latches were on?

Joe Obbish

well I parsed out the page and generated the sql, so it wasn't too bad

sys.dm_os_waiting_tasks

what's the extended event for this? latch_acquire_time?

Forrest

Haven't run the latch for this myself, so I don't know

Philᵀᴹ

What a mess doublepulsar.com/…

Joe Obbish

7:33 PM

I just can't believe that we had over ten minutes of wait for a single PFS page

also typing out "PFS" instead of "FPS" is incredibly challenging for me

sp_BlitzErik

wow

Evan Carroll

looollll

you see the asm code the kernel patch generates?

It's hilariously cute too. Without the ability to disable branch prediction, they overwhelm it with a series of short jumps

sp_BlitzErik

is slack messed up for anyone else?

bluefeet

yup, it's down

Lamak

@bluefeet oh, tank jeebus, I thought they had blocked it at my job

Evan Carroll

7:44 PM

If you're having issues connecting to Slack: we're working on getting things back to normal with top priority. Thanks kindly for your patience. 🙏

Tweeted by SlackHQ on January 9, 2018 at 7:32 PM

sp_BlitzErik

thanks twitter man

Evan Carroll

NP.

I'm reading this book, Database Internals 7th E.

I think 95% of the terms used are unique to this book

sp_BlitzErik

never heard of it

sounds goofy

Joe Obbish

@sp_BlitzErik So are you impressed?

We have an extreme workload (TM)

Zane

yo

sp_BlitzErik

7:58 PM

@JoeObbish i'm always impressed by how bad your demo queries are :)

Evan Carroll

ahahhaha

Joe Obbish

uh huh

you should tell me about how increasing the number of files isn't necessary for user databases

sp_BlitzErik

did adding more fix the problem?

Forrest

Hmm. A PFS only covers about 64MB, right? So why would specific PFS pages have so much contention for DW loading?

sp_BlitzErik

columnstore

?

Joe Obbish

8:02 PM

@sp_BlitzErik not sure, now we have 96 cores

no CCIs on the staging database

sp_BlitzErik

all heaps?

Joe Obbish

yes

sp_BlitzErik

cool, lemme know if adding files helps

Joe Obbish

right, cause it's so easy to add files

sp_BlitzErik

that's jr dba talk

Joe Obbish

8:03 PM

I don't get why you're so philosophically opposed to this

sp_BlitzErik

i'm not

Joe Obbish

just need hard evidence?

sp_BlitzErik

i'd be fine with soft serve

Joe Obbish

it helped when we were on 24 cores

we increased files from 4 to 24

PAGELATCH_UP waits went down

sp_BlitzErik

linearly, or was there a diminished roi?

Joe Obbish

8:05 PM

did it all in one go

sp_BlitzErik

ah

Joe Obbish

we have TBs of data in the database

Zane

How many drive letters do you have?

Joe Obbish

@Zane using mount points (I think that's the term), but my understanding is drive letters matter for IO

Zane

I'm late here but curious. In my experience I have found little benefit from splitting up a data file unless I'm spreading it across drives. Whether on SAN or not.

Joe Obbish

8:10 PM

right now the theory is this:

"Common causes of PAGELATCH_XX contention are:

Allocation bitmap contention in tempdb (PAGELATCH_UP for multiple threads trying to change the same bitmap), and under extreme loads, in user databases"

sp_BlitzErik

with terabytes of data you should have oodles of pfs pages

interesting

Joe Obbish

@sp_BlitzErik it's based on total database size, right?

sp_BlitzErik

yeah

if your database starts empty, you may way to consider pre-sizing it

with ifi on that should be fairly painless

Joe Obbish

autogrowth is off

sp_BlitzErik

that doesn't sound very professional

Joe Obbish

8:22 PM

@sp_BlitzErik didn't you just say to presize it?

Lamak

@JoeObbish I assumed it was about the initial size?

sp_BlitzErik

i didn't say to box it like a veal

Joe Obbish

database doesn't start empty

what's the disadvantage in growing it to full size?

maybe I should make one filegroup per numa node

that doesn't sound annoying at all

sp_BlitzErik

never is

Joe Obbish

then I can force queries to go to certain numa nodes

and to write to only their filegroup

then the PFS pages stay in local memory

sp_BlitzErik

8:29 PM

a fair plan

Zane

8:46 PM

It's really nice having a SQL Sentry Licence that I can't use because our windows team blocked all permissions for monitoring tools.

sp_BlitzErik

i bet they use solarwinds to monitor that

you should kill them

Joe Obbish

don't solarwinds fry electronics?

I don't really get the name

Zane

@sp_BlitzErik I believe we have an internal team called Frustration Werks and they tend to implement most of these decisions.

Sean Gallardy - Microsoft

Hola amigos

sp_BlitzErik

sup

Sean Gallardy - Microsoft

8:54 PM

@JoeObbish Aww man, I should have given you the magic formula.. pfs pages are every 8088 pages and gam intervals are every 511232 pages

sp_BlitzErik

joe won't run who is active

too much overhead

he uses dmvs

Joe Obbish

@sp_BlitzErik it takes like 30 seconds to run

Sean Gallardy - Microsoft

Kids these days

sp_BlitzErik

@JoeObbish sounds like a vm issue

Joe Obbish

@SeanGallardy-Microsoft thanks. I knew there was a formula, just didn't look for it

I suppose that would have been easier

Sean Gallardy - Microsoft

8:57 PM

... but don't tell @EvanCarroll!

Why does he call himself even that when his name is Peter?

Joe Obbish

I'm actually trying row compression like a crazy person

Lamak

@SeanGallardy-Microsoft do you speak spanish better than @billinkc?

Sean Gallardy - Microsoft

Nothing wrong with that, I approve!

@Lamak No

there was my answer, in Spanish.

Zane

Sentry = WMI is blocked blocked. Perfmon = forbidden. Server Log on = Emergency only. Powershell = remotely blocked. Any suggestions for how to monitor a box when all of these have been blocked? lol

Lamak

@SeanGallardy-Microsoft so, it was really a yes

Sean Gallardy - Microsoft

8:59 PM

unless you want to talk in sentences about the little girl that's running....

or the women that are reading

or the men that are cooking

Zane

@JoeObbish whaaaaaaaaaaat?!

Joe Obbish

eXtreme workload

Sean Gallardy - Microsoft

@Zane If no one can monitor it... how do you know it's running ;)

sp_BlitzErik

maybe you need xtremio

Sean Gallardy - Microsoft

I suggest SQL Injection

Zane

9:00 PM

@JoeObbish like Doritos extreme or?

Joe Obbish

@sp_BlitzErik it's PAGELATCH_UP, not PAGEIOLATCH_UP

Lamak

@SeanGallardy-Microsoft ah, that's what you guys mean with a quantum database?

Joe Obbish

but it was great

Zane

@SeanGallardy-Microsoft just wait for users to throw a fit I suppose.

Joe Obbish

after reading Paul Randal tell me it that it wasn't an IO problem many times

I caught myself asking someone if we could monitor the SAN

shameful

Lamak

9:02 PM

you should never trust a Paul

Evan Carroll

@SeanGallardy-Microsoft keep it up, I'm going to send everything I got to Wikileaks!

Sean Gallardy - Microsoft

@EvanCarroll NNNNNNNNNNOOOOOOOOOOOOOOOOOOOOO!

sp_BlitzErik

@JoeObbish wasn't your big plan to use buffer pool extensions?

Sean Gallardy - Microsoft

@Zane Sounds like the normal reactive process that companies use.

Joe Obbish

@sp_BlitzErik yes

Evan Carroll

9:04 PM

The Microsoft Files: How to Install PostgreSQL on Windows using SQL Server.

sp_BlitzErik

there you go.

Evan Carroll

sp_EvanBlitz

Joe Obbish

they won't give me access to the server USB ports though

Sean Gallardy - Microsoft

@EvanCarroll Queue scary music

Zane

@SeanGallardy-Microsoft Yeah the company is a certain massive bank that got caught stealing from customers and is currently reacting by cracking down on IT people instead of executives.

sp_BlitzErik

9:07 PM

@JoeObbish it sounds like they don't trust you for some reason

Joe Obbish

@sp_BlitzErik Mr. sql_handle just had an idea to delete from the heaps without tablock

sp_BlitzErik

if the locks escalate to table locks anyway it won't matter

you'd have to batch delete in tiny chunks

bluefeet

9:26 PM

@JackDouglas should be there as well

Sean Gallardy - Microsoft

@Zane So.... par for the course?

@EvanCarroll If you write it up... I'll post it ;)

Evan Carroll

Sounds like a plan. So I distribute a cab, and a dll and create an extended stored proc that self-extracts it and runs it. SELECT * FROM pg_system_upgrade();

Joe Obbish

9:44 PM

@sp_BlitzErik these are staging tables that typically get truncated daily

the idea is that you'd do deletes instead so that you don't need to allocate the pages for the next day

sp_BlitzErik

if you just do delete from dbo.myheap there's a good chance you'll get a table level lock anyway and pages will get deallocated

you'd have to delete in batches that stay at page or row locks

Joe Obbish

10:16 PM

oh

that would be pretty much the worst thing

sp_BlitzErik