[Video] Office Hours 2018/10/24 (With Transcriptions)

This week, Brent, Tara, and Richie discuss increasing a table’s varchar field from 1 to 2, predicting how long CHECKDB should take, SQL Server configuration options, reporting, reducing high waits, analysis services for SQL Azure, advice to first time PASS attendees, compatibility levels, cardinality estimator, PowerShell, and Tara’s hairdo.

Here’s the video on YouTube:

You can register to attendnextweek’s Office Hours , or subscribe to our podcast to listen on the go.

If you prefer to listen to the audio:

Podcast: Play in new window | Download

Enjoy the Podcast?

Don’t miss an episode, subscribe via iTunes , Stitcher or RSS .

Leave us a review in iTunes

Office Hours Webcast 2018-10-24 Should I use SSIS to manage AG backups?

Brent Ozar: Let’s see, Josh asks, “We have a suggestion for backup jobs to use an SSIS package around to move the jobs to whatever node is the primary in an Availability Group. Have you seen this setup and can it work well?”

Tara Kizer: I haven’t seen that for the Availability Groups that I’ve supported. We had two different methods at three companies that I had Availability Groups. But at the last one, we just had a server that would run the backups and it would just point to the listener name. So all of our backups, index maintenance, anything that needed to connect to a specific replica, we would just put the jobs on that server instead and it would be a SQL CMD and it would point to the listener name. And we did that for the backups because we wanted the backups to be on the primary replica. I don’t necessarily agree with offloading such a critical task.

Brent Ozar: And why not?

Tara Kizer: Well even on a synchronous replica, it’s not completely up to date. So that is just a good idea if your RPO goal is so low.

Brent Ozar: Between that advice too, and if you think about an SSIS package that would move stuff around, there’s going to be a delay there. Like, your backups won’t be up to date until the SSIS package moves a job around. And if that job process breaks, you’re not getting backups. I’d be like, I can’t unsubscribe fast enough from that plan. What I would do there instead is Ola Hallengren’s stuff. Ola Hallengren’s backup scripts will automatically run on every replica and just backup wherever you tell it to; like, if you want to prefer a secondary or if you want to prefer the primary. Configuration’s not super simple on Always On Availability Groups, but it works.

I need to go from VARCHAR(1) to VARCHAR(2)…

Brent Ozar: Pablo says, “Hello, friends…” Hello Pablo, “I have an 800GB table that needs to increase the size of a varchar field from one to two. What way would you recommend doing that to reduce impact?” Isn’t that an instant change?

Tara Kizer: Yeah, I was going to say, I don’t know that this is going to take a while. Run that on a test box.

Brent Ozar: Yeah, I’m pretty sure that that’s going to be instant.

Tara Kizer: Yeah, and if it isn’t, I would probably just move the column to another table. I would cheat because your table’s just so big. Remove the column I realize that there’s a huge impact there, but…

Brent Ozar: Somebody had a really good post just recently about a switching cups way to do it. What you do is add a new column with the new data type that you want, but with a different name. You put in a trigger so that whenever an update happens, it sets the new column over to the old value’s name. Then you go through and roll through in batches of like 4000 rows at a time, gradually updating the new column. Then you switch column names.

Tara Kizer: Oh, I like it. [crosstalk] I mean, it’s 800GB. I mean, we don’t even know the row count, because maybe it’s 800GB because of data, you know, the size of a row.

Brent Ozar: Yeah, let’s try it and see. And I’m just going to grab, out of the Stack Overflow post table, I’m just going to grab one of the columns post-only because oh, you know what I’ll do, find out if votes has a varchar in it. No, it doesn’t. Posts has a varchar and posts has a decent number of rows in this. I’m going to change title from nvarchar 250 to nvarchar251. Alter column, and then I’m going to freeze because I get a spinning beach ball. Oh, that’s excellent, c-c-c-c-column. It’s like some kind of pop song.

What did I say I was going to do? Title nvarchar 251, and let’s see if that happens instantly. So the next thing I’m going to do is switch into the right database, then I’m going to run it. There you go. Yeah, that’s a metadata change. It should happen instantly, regardless of the number of rows in the table. Love it when we can give people good news instead of bad news.

Richie Rump: It doesn’t happen often.

Brent Ozar: It does not happen often.

How long will CHECKDB take?

Brent Ozar: So next up, Brian says Brian, this is a great question, “Is there a good way to predict how long CHECKDB should take? I’ve tried to space out my various jobs, but sometimes they step on each other’s toes with my jobs running into business hours on the next morning.” Any ideas, Tara, Richie?

Tara Kizer: I mean, you know, CHECKDB, I don’t even like to run that on a production server. I offload that task to another box which, obviously, you have to license that other box. But I like a box that maybe doesn’t have a lot of hardware, but it can churn through just running CHECKDB in all the databases. Eventually it completes, and maybe it’s doing the work for more than one server, but it might have lesser cores for licensing reasons, because you do need to license it.

Richie Rump: I mean, do you need to license that?

Tara Kizer: You do.

Richie Rump: Wouldn’t the development work?

Tara Kizer: No, because you’re offloading a production task. That’s the key there.

Richie Rump: Semantics.

Brent Ozar: If you just happened to be restoring it every night into your development environment and you wanted to make sure that your development environment was okay, that’s totally still not legal, but there we go.

Brent Ozar: Marci says, “Hey, did you get my question?” Yes.

What should my setup defaults be?

Brent Ozar: Jason says, “When you helped us configure our SQL Server 2012 database a few years ago, you recommended that we start with MAXDOP 8, cost threshold at 50, and eight tempdb data files and then tweak from there. We’re not switching to 2016. Do you recommend we start those same settings?”

Tara Kizer: Yeah.

Brent Ozar: Assuming you’re going to the same size or bigger of a box. Heaven forbid you’re going to a two core box or something like that.

Tara Kizer: I don’t know if this is in our set of checklists. I don’t think it is. But on 2014 and