@ErikDarling That explanation doesn't sound convincing. They're still trying to blame the upgrade. I don't know exactly what's going on without trying to reproduce it but it sounds fishy

I agree. It’s all rather nonsensical sounding, but perhaps my reading comprehension is clouded by their original assertion.

Clouds ruin everything

They do

I've been trying to work up some interest in EF's latest self-inflicted problem but no luck so far

i don't think it would matter either way

Everything's Fucked...

it's a snake eating a large apple

@JoshDarnell That's our go-to pattern for Synapse.

@Zikato You're on a roll!

For those fighting with Teams...

> Advanced Support for Microsoft Teams 1 Day (Jean Opperman) [Microsoft Premier Unified Workshops]

Many Uses of SQL CASE Expression in a JOIN Clause

Should join the EF team

oh no

@ErikDarling so their CASE isn't valid?

One might say

A chairde - Morning all!

Again dammit

😮

it's like a boat slowly filling up with water

> but I get bunch of errors : Msg 214, Level 16, State 201, Procedure xp_cmdshell, Line 1 Procedure expects parameter 'command_string' of type 'varchar' [...] I couldn't solve this problem.

i wonder if the poppins could loan them a varchar variable

Just one more reason to avoid (max) types

I've managed to screengrab @HannahVernon Wordle statistics

🤣🤣🤣

LOL

Nice job!

malformed 1

you mean a wizard hat

lamak can't triforce, etc.

If a WorkTable falls in a query plan, and there's no I/O statistics around it, does it make a sound?

brentozar.com/archive/2023/05/…

Nice post, Erik

Are you saying you have a problem with identical environments

Because we've heard that one

I'm specifically avoiding saying "identical environments". But pretty damn close

@Zikato any long running transaction on the primary?

@Zikato a billion and 3 things

Last one I troubleshot was due to XE's being on one node, and only one node

but you'll need to gather some data

@SeanGallardy That narrows it down

Which XE? Anything HADR related?

@Zikato no, this person had an XE to capture statement completed on that node which was adding a ton of overhead. Disabling the XE stopped all their issues.

You have a redo queue, so you'll want to find the redo threads and then capture waits and external waits for those threads

Start there

Unlike Erik, I have reasonable rates :P

i think our rates are the same in chat

setting the meta for sure

@SeanGallardy LATCH_EX on RECOVERY_MANAGER

The docs page for that latch says, "Ask Sean"

@Forrest That's where all the time is spent?

Majority, yeah

what's the callstack?

and how much time, out of total, is spent there?

Looks like it's just busy, both of those are typical stacks

is this secondary readable?

yeah, running 60% CPU with activity, though the redo waits have almost no signal time

another readable secondary at similar CPU doesn't have the slow redo

both are primaries for other databases

What's the actual redo amount/sec?

What's the wait for the latch?

Seems like it's just running through redo

look at forry with the call stacks PRONTO

what a great supportee

is there a reliable counter for actual redo rate?

it's just running through redo, there's nothing out of the ordinary with the stacks

so if it's waiting on an ex latch, most likely other parts are going slow

I would suggest setting it to single threaded redo

check the throughput

That'll remove the dependency on other parts of the system getting enough time

no difference, maybe a little slower single-threaded

does the wait type change?

oooh, good point, not sure we've checked that

> The customer wants to use a third party tool which doesn’t handle BigInt values [...] customer would like Microsoft to provide support for not using BigInt to get around 3rd part tool issue.

What does that even mean

who knows, man

small numbers lead to big request, Sean's rage overflows

@SeanGallardy sounds like customer should use INT then and Microsoft already supports this?

I was really hoping that was quoted from a Post, but then realized this is your life, lol.

@J.D. Apparently there's a system table used where it has a bigint in it, but the 3rd party tool doesn't support bigint, so they want the table changed to use int so that the 3rd party tool works.

@J.D. Yes, my sad, sad, sad life.

When you haven't written any code, but your build output says: 275 Warning(s)

@SeanGallardy ah gotcha. Customer doesn't realize CONVERT(INT, SystemTableColumn) is a thing too I guess, assuming the system table doesn't actually exceed 2,147,483,647 lol.

I don't really care enough to investigate, they need to contact 3rd party to update their code or not use it.

Indeed.

Wait

Which system table

lol why....

Erik, did you write bad code

Never

Just ask Paul

Or more importantly ask @Zikato or @JoshDarnell

classic Windows XP screenshots still in use in 2022 lol.

yah. DPA is so.much.fun.

we enabled Forced Encryption on our instances and DPA won't connect at all now.

I love how they call it the "Success Center"

Probably for the best

undoubtedly

It’s as useless as monitoring tools get

seems like it. I have some SQL Agent jobs that are designed to have the first step report failure if the job is running on a non-primary AG instance - so step 2 etc doesn't try to run against a read-only database. The job itself reports success, after step 1 completes, however DPA reports the job has failed.

and that's just one tiny issue.

I've witnessed that on a few of the monitoring items

puke on ags

Now there are contained AGs... oh boy

everything is going to break

I’m sure the Azure portal will be just fine

on another matter entirely, our Data Analytics team has a product that uses CDC and fn_dump_dblog to read the log files in order to gather recent data changes for propagation into the data warehouse. They swear up and down that enabling CDC for all the tables won't be a problem because it doesn't actually copy all the changes, just keeps track of LSNs.

sounds like attunity

I can't wait to see how that affects log usage in our Distributed AG

@SeanGallardy its not. I asked the vendor for a copy of their software so I could see how it works, and they initially set up a conference call with me but they've since completely ghosted me.

👻

Well that’s not true about CDC but it’s still better than CT

Wonder how their product handles schema changes...