Eddie dressler

@not_unambitious Loved it. Well done.

Thank you thank you thank you #KCDC2024 for showing up and packing the room! 👏🏻🙌🏻 Now, go use that white male systemic advantage for good!

Loving my first two sessions. One familiar, one new - both teaching me something. @TheCodeTraveler #kcdc2024

Is sqlsaturday still a thing? Im looking to get connected to kansas city’s (or restart it). halp. @sqlandy #sqlsaturday #SQLServer

@quorralyne im going to try to make you KC.NET user group meeting on 7/26! let’s see if my work schedule behaves.

thanks @ShaneWelgama for your locking session. presentation was great, you may keep a glass of water with a straw nearby for dry throat/mouth.

who knows the details of msdb.dbo.syssessions? im desperately tring to stop a zombie agent job. #sqlhelp

@SluggoDoug i like Elvis Costello’s “you turned to me” for the weeping angels episode as well as “the angels want to wear my red shoes”

Thank you for your vid entitled “BCP in errors” from 3 years ago! @midnightdba. Kept what little bit of sanity i have left.

@Hugo_Kornelis Ill see if i can get a good repro to make it real. Thanks for your help

@Sqled Yeah, sorry. Seems I misread your post at first. Nonclustered index with or without INCLUDE of clustered index key should make no difference. Not sure how the optimizer picks the index to use. You mention it runs faster. By how much? Is it always faster or only sometimes?

Assume: col1 has clustered index defined on it via a PK constraint. I have a weird case where col2 with a NCI on it performs worse than when col2 has a NCI and INCLUDES col1. I thought the include was kind of implicit due to the NCI built on the CI. details appreciated! #sqlhelp

@Hugo_Kornelis Hugo TYVM. The confusing part in my case is that all NCIs have a pointer to the CI (its baked into NCIs). So why, if the NCI knows to which row in the CI it points did an INCLUDE of the column (that IS that pointer) speed up the query? It shouldbhave made NO difference

@Sqled There are some tiny differences between the storage structure of clustered vs nonclustered indexes. If I recall correctly it's a few bytes per row on intermediate pages. 1/2

@Hugo_Kornelis I dont hugo. But its pretty simple. Pk on col1, then run Select col1, col2 from table1 where col2 = 0. Build an index1 on col2. And the select will use the index. build index2 on col2 INCLUDE col1 and Rerun the select. You will see rhe ExePlan choses the index2...why?

@Sqled Do you have a repro script?

For anyone who needs a reminder.... #sqlhelp.

@mrdenny @jdanton @erinstellato Ha! Never would I ignore you @mrdenny Thanks for the chiming in!

@Sqled @jdanton @erinstellato then ignore me, and follow Erin's advise. :) She'll totally lead you in the correct direction.

Can i easly set up an ee session to capture DML? All i want to do is capture and store before and after values of DML. Yes i know i can use triggers, yes i know i can use CDC. Im pouring over all of my options and figured id ask the expert(s)... #sqlhelp @erinstellato

@jdanton @erinstellato @mrdenny In rereading my post i realize i refered to two different phenominoms. I wish to capture data values not sql statements. Thanks Joey and Denny for the followup.

@erinstellato @mrdenny @Sqled Temporal would be the easiest way to implement this. That shows what changed/auditing shows you made a change. #sqlhelp

@erinstellato @jdanton @mrdenny Erin you are a rocker! Thanks for directly answering my question! And thanks to this community for chiming in. After a full day of researching i too found no way for exented events to capture a before and after value of an update!

@jdanton @mrdenny @Sqled I disagree there. I think there's a lot more to consider when implementing temporal in terms of how you could/might design that historical table. And actually, I was incorrect in that only CDC would give the before and after value.

@pinaldave Thanks for another great performance demo. This time in MySQL!

@PaulRandal Thanks for your responses Paul! I think im closer to solving my issue.

I have idle blockers running amok in prod. ill get one “random” session go idle and that will start a nasty blocking chain. Does anyone have good systemic ways to root cause why the spid/transaction went idle in the first place!? Thanks all! #sqlhelp.

@PaulRandal Dpa defines an idle blocker as one that WAS doing work but is not currently. Often times from an mis managed trans. So i have a spid and a time point and i know its blocking locker. Another road i could go down is “how do you find the last wait a spid had?” Thanks!