[Video] Office Hours Speed Round


In ten minutes, let’s cover as many questions as possible from https://pollgab.com/room/brento:

Here’s what we raced through:

  • 00:00 Introductions
  • 00:17 dbacat: We’ve been fighting an AG issue for a while. Sometimes during failover, one or a few databases go into initializing/recovery state on the new secondary. It’s random, never consistent. Only fix is to drop DB from AG and add back. MS support was not helpful. Have you seen this?
  • 00:38 Gaspard: Is performing RESTORE WITH VERIFY_ONLY enough to test backup health? If not, what could go wrong between this and a full database restore?
  • 01:22 RoJo: It used to be common to have a dedicated IO channel for App, Data, Logs. Now that SAN is common can everything go to the same channel or would it be better to keep separate? Would it ever be better to keep SAN as one channel and others on say local RAID?
  • 01:57 MyFianchetto: LPIM. Is it a must have after setting MAX SERVER MEMORY?
  • 02:50 Steve McGarrett: Do you have any sizing guidelines / tips when migrating an onprem SQL 2019 Enterprise instance to Azure SQL?
  • 03:37 Dave: Hi Brent, we have an on-prem sql server used as a DWH, my company is starting to move some of the workload to AWS, what is your take on Amazon Aurora as an alternative (we can change the application to support it)?
  • 04:34 Anatoli: How do you know if your SQL server physical RAM is going bad?
  • 04:55 Mirza: Hey Brent, you had mentioned that PLE metric is useless. However, I had a talk with someone from Microsoft and that person says that PLE is a good way to measure memory pressure. How do I convince him to stop paying attention to PLE? Thanks in advance.
  • 05:34 Gerardo: sp_BlitzFirst is showing several hours of CMemThread waits on our 64 core SQL 2019 Enterprise CU16 instance. Trace flag 8048 did not help. How do we use sp_blitzcache to find the top queries causing this wait?
  • 07:00 Hiroto Katagiri: What is your opinion of Azure Synapse and Snowflake cloud data ware houses?
  • 07:50 Danny Reagan: What are your pros and cons for using SQL tables to implement queuing?
  • 08:25 Levi: Is SQL page level restore a practical solution for dealing with data corruption? Have you seen it successfully used?
  • 09:17 chandwich: Will we ever get a, “Watch Brent upgrade SQL Server”?
  • 09:59 Bruno Luis: We use PARSENAME as a sneaky way to split strings in 2017 (eg IP addresses). However, we can’t create an index with it because it’s non-deterministic? But… WHY? Please, WHY?
Previous Post
Who’s Hiring in the Database Community? July 2022 Edition
Next Post
Here’s What I’m Teaching at the 2022 PASS Summit.

Leave a Reply

Your email address will not be published. Required fields are marked *

Fill out this field
Fill out this field
Please enter a valid email address.