« first day (2570 days earlier)      last day (300 days later) » 
00:00 - 20:0020:00 - 00:00

8:07 PM
Microsoft made another minor baby step in its pursuit to not-suck.
 
sean did that
 
I was really looking forward to a 4 letter acronym named "cmdlet" that reinvented curl but was broken in fresh and exciting ways and always required the QUOTED_IDENTIFIER flag.
 
SET ANSI_NULLS ON;
SET ANSI_PADDING ON;
SET ANSI_WARNINGS ON;
SET ARITHABORT ON;
SET CONCAT_NULL_YIELDS_NULL ON;
SET QUOTED_IDENTIFIER ON;
 
Now let's create a competing and confusing amount of "official" clients each with their own default settings that modify that.
 
done
what's next?
 
8:14 PM
can we please simplify the in-memory troubleshooting memory pressure resolution path to cycle the service
 
done
what's next?
 
Providing an interface into C#'s geometry and geography types and calling that "Spatial Types for SQL Server." For fun, just say you comply with standards you haven't even looked at...
 
done
next
 
Create one repository for each piece of software, rather than understanding the technology, just market that you have multiple repositories because more is better!
 
i don't even know what that means
have you bought a mac yet?
 
8:20 PM
Then, call your stuff open source because it's written in Python. But just shim a bunch of proprietary .dlls. Python wrappers and binary blobs on github for everyone!
 
i hear photoshop is nice
 
"Microsoft user 2018"
 
did you photoshop that?
 
No I used Power BI. It can play movies too!
 
that's basically photoshop for it people
2
 
8:28 PM
EXACTLY
We agree! Now let's drink egg whites and inject each others asses with test.
 
wait wat?
 
that's what i was doing today anyway
but i was injecting my ass with egg whites
which would explain a lot
 
see? you have it all backwards
duhr
 
this is why my stomach is angry and i'm growing tail feathers
 
8:32 PM
could also be your distiller of choice, to be honest
 
did something happen to lagavulin?
 
I don't feel like reading back to see what this is about
I just had steak wellington and wine
 
@EvanCarroll curl is just an alias for invoke-webrequest
sooo
glad I could make your day
 
apparently not anymore
 
bhahahahhaha
 
8:38 PM
looks like curl is legit(-ish) curl
assuming you "install" it
 
lol
or I could use powershell which is already there and gives me pretty much the same thing
 
Last few comments seem like something that could ruin the eat well feel well feeling
 
but why do that when I could install open source software to do somethingI can already do...
 
@EvanCarroll it looks like Microsoft's GitHub is a collection of actual things as well as a dumping ground for personal projects
this thing looks interesting
 
Look at this awesome patch to add DAC to MSSQL-CLI github.com/dbcli/mssql-cli/pull/127/commits/…
Microsoft, they get it!
 
8:42 PM
Binary file not shown
My sarcasm detector is at 11
3
 
how's it going, @TomV?
 
You don't actually have to open up unless your git repository for the "client" has a 30 MB blob in it, github.com/dbcli/mssql-cli/blob/master/sqltoolsservice/…
 
@swasheck good actually
 
glad to hear it. you sound shocked
 
I engaged in a challenge
I promised I won't complain (about anything) for a month
4 days of pretending everything is going well so far
@swasheck not complaining kind of got me out of my natural habitat
 
8:56 PM
@TomV It's like you're in a relationship!
So, I have a plethora of notes written down on things to blog about... but I'm finding not everyone thinks they are as awesome as I do..
with that being said, what would you all like to see new posts about?
preferably the storage engine...
 
@SeanGallardy-Microsoft I won't complain
 
LOL
at least not for 26 more days ;)
 
For at least 3 and a half weeks
Yah
 
@SeanGallardy-Microsoft problem is your platform
 
@SeanGallardy-Microsoft I even pretend to be satisfied with Microsoft support
 
9:01 PM
we can't use T834 with CCIs but it seems to be the only fix for this bottleneck
 
That was just kidding
 
@JoeObbish obviously
 
well we'd had four versions without support for T834... hard not to be pessimistic about it
 
If that's all it takes to get you down...
 
seeing "this doesn't work in versions SQL Server 2012 to 2017" sticks out in the documentation
 
9:02 PM
@TomV nice! excellent thing to develop.
 
@swasheck it's relaxing
Forcing yourself to either not react or at least react without sounding irritated somehow works
 
@SeanGallardy-Microsoft columnstore memory management
 
9:21 PM
@swasheck I'm out for tonight. Hope you are progressing on the adoption track if that is still something you are aiming for
 
9:37 PM
@JoeObbish "management"
 
@sp_BlitzErik I know right
like this if you cry every time:
 
at least it's not a heap!
 
what
the entire problem is that it's a CCI
 
you seem to have worse luck with those
maybe if you provided less context that would be more clear
 
9:52 PM
I got a set of 40 nearly identical queries that should have the same plan
some of them do row mode hash joins instead of batch mode
probably related to plan timeouts?
even worse, the repartition streams is likely to put all of the rows on one thread
 
16 or 17?
 
would memory pressure change batch/row mode choice?
guessing trivial plans aren't the issue
 
memory pressure can affect query plan exploration
I think it can terminate early
sql_handle understands the mechanics pretty well
the funny thing is that only one join is affected
and it's the worst possible join to be affected
it's the join that happens first and that has the worst row distribution by hash
 
no reference to column store, but you may find similar behavior
 
10:00 PM
do you know if query store has it?
if not I'll try to find the right XE
 
has what exactly?
 
query plan timeout reasons
or memory used during compilation
it does
 
SELECT *
FROM sys.query_store_query AS qsq
 
sys.query_store_query has last_compile_memory_kb among with a few others
 
you could also check for forced grants
SELECT  deqrs.forced_grant_count
FROM    sys.dm_exec_query_resource_semaphores AS deqrs;
 
10:10 PM
query store hates me :(
 
have you considered blitzing it?
 
database name should always be first imo
 
open a github issues
 
but that would be a breaking change
 
for who?
 
10:14 PM
everyone?
also, it hasn't finished after a minute
I'll just use XE
those have never let anyone down
 
check messages for status
use @skipxml if you don't care about analysis
 
currently on Gathering highest memory use plans
I guess we're giving it a work out
also there appears to be a wonderful extended event for this
 
la
dee
da
 
10:32 PM
@sp_BlitzErik I don't like it
it reminds me of all of the cool 2017 features I don't have
 
i suppose you could use blitzcache and sort by memory grant or avg memory grant
that will remind you of nothing futuristic
 
to be honest I wrote my own blitz-like procedure for this testing
please don't sue me
 
i don't care
what does it do?
 
check electronic mail
 
i'm so far in the future man
 
10:50 PM
from where I am standing, you're all cro-magnon
 
you wear flip flops
scram
 
That puts me on the same tier with Jesus, Moses, and Muhammad; and, just one sheet of plastic away from Socrates and a health code violation.
 
no one on reddit believes in those people
 
11:09 PM
I get a bunch of traditional optimizer timeouts
but no memory related ones, according to XE
hm
 
time for the trace flag
 
I'm already using FORCE ORDER
 
what's not happening that you want to happen?
 
Do you mean in general? or with respect to what?
 
clearly something has gotten out of control and you've resorted to the query tuning version of buying sweatpants
 
11:21 PM
oh man, if only you knew
sql server was picking join orders that were basically guaranteed to spill
I didn't like that
wasn't a CE issue
 
just enjoys spilling then
 
I have a question that I am not sure is on topic for the main site
I need to store the values of when a row was last updated and am unable to do the simple fix of adding a datetime column.
Right now my idea would be to use a trigger to store the primary keys of any updated rows in a separate table,
Does anyone have a better idea?
 
dude i don't even know what platform you're using
 
@sp_BlitzErik MariaDB
 
not sure, not my thing
not too many people in here seem to be users either, you're better off on the main site
 
11:34 PM
@sp_BlitzErik Thanks
 
yeah, sorry i couldn't just answer!
 
00:00 - 20:0020:00 - 00:00

« first day (2570 days earlier)      last day (300 days later) »