It’s time for the first live Office Hours in my new home studio in San Diego! I took your highest-upvoted questions from https://pollgab.com/room/brento/ and answered ’em:

Here’s what we covered:

• 00:00 Welcome
• 01:30 Accidental DBA: Hi Brent, do you have a recommended checklist for trubleshooting occasional slow INSERT performance?
• 03:19 Mark: How do companies deal with ever growing databases and writing to very large tables? Have you heard of ways to splitting your data, allowing smaller ‘active’ data on fast storage, while older data reside on cheaper storage?
• 05:27 Kristófer Ólaffson: My TempB files fill the drive they are on about once a month. Currently sitting at 700GB combined. I have around 3TB of data on this server. I’ve been increasing the size of the drive TempDB sits on, but are there some tips in writing sql to not fill the TempDB as much?
• 07:28 Rick: Hi Brent, I have a table in a client database with ~1million rows and a statistic that hasn’t been updated since 2018 with a modification counter of over 30.5 million. Auto-Stats update is enabled. Why would this statistic be getting updated?
• 09:15 Dinis: Hi Brent, if I want to have maximum security in encryption, I need to have a combination of TDE, Encrypted Connections and Always Encrypted (for some sensitive columns). Is that correct?
• 11:00 John James: My team wants to start collecting index data and use your DEATH method to send out suggestions to teams. Our company restarts all servers weekly. Does it make sense to collect the data before the servers restarts and aggregate it?
• 12:33 Mahesh: Which is the best storage configuration, HDD (inserted locally into server) create one large virtual disk and make partition like data, log, back or create a separate virtual disk for data and like wise of log and back. What do you recommend. In my case we are using RAID 5.
• 14:30 Tim Lara: Hi Brent, if I have a bunch of “identical” servers, but some of them perform very poorly on certain queries, while others run the same queries just fine, how can I tell what’s different between the “good” and the “bad” servers?
• 16:08 Frank: Hi Brent, I have two high availability groups in a 2-node-sql-cluster. In case of a failover sometimes only one HA-group moved to the other node. What do I have to do so that BOTH groups always fail over together to the same node?
• 17:46 Tom: What are the common roles of responsibility you have seen between programmers and DBAs? I am an accidental and only DBA in a dev shop of 24 and our roles have not been defined so I find it difficult to balance between what I think I should do and what I really should do.
• 20:34 WillemHenk: Is switching from PostgreSql to SQL Server a big step? I find some articles regarding differences, but I was curious about your experience.
• 23:50 Mr. Griffith: I understand you frown on in-place upgrades to SQL Server. Something about “juggling crown jewels”. Does the same rule apply to in-place upgrades for Windows Server?
• 25:21 Jaden: My company using the SQL Server and they want to save the database into JSON. Why they do it and what changes to me?
• 26:35 DBA Girl: Hi Brent! Our Company creates indexes for all foreign keys on a table. For some tables, though, this can create 20 or more indexes. How do you balance the recommendation of indexing those columns with the performance degradation of having too many indexes on a table?
• 29:19 Bandar Almutairi: I came across a weird situation where queries against one table is way slower when ran from the readable secondary. It’s fixed after rebuilding the primary index on primary. My question is why isn’t it as slow on primary? Execution plans looked identical.
• 32:42 Mark: We have some huge table. After huge inserts into the table the query plan takes index scan rather than seek(before it took). updatestat might fix that table but breaks other query. I can’t change query because a tool handles that. Do you have any idea why that happens. Thanks!
• 34:26 LittleBobbyTable5: Hi Brent, When optimising queries using HINTS and OPTIONS I am hitting Bitmap Create part that are really slowwww. What voodoo optimisation technique should I employ to avoid this slow Bitmap creation. Could you show us a quick example please.
• 36:40 Farshid: Hi Brent, Have you ever had a customer which you tried to convince them to switch from Oracle to SQL Server?
• 39:00 Mikkel (Denmark): Some times the QueryText in the BlitzCache result is truncated. And it is not the full text in the QueryPlan. Is it possible to get the server to save the full query text?
• 40:00 Jay: Is the CDC (change data capture) a viable strategy to publish “event” data to external systems instead of publishing from application code?
• 41:17 Oleg: if you have to choose to tune query, one where SQL overestimates and another underestimates number of rows, which one would you prefer to tune and which to leave as is?