Press "Enter" to skip to content

Category: Internals

Identifying an Object Name from a Wait Resource

Published 2025-03-18 by Kevin Feasel

Haripriya Naidu wants to know what object this is:

You run a query to check for locking or blocking in SSMS and find a wait resource in the format (8:1:3610). To identify the object name, you would typically run multiple queries, first find database name, turn on trace flag 3604, then find object id from DBCC PAGE and then find object name from sys.objects.
However, with a new function “sys.dm_db_page_info” introduced in SQL Server 2019, you no longer need to go through these steps. Instead, you can run a single query to get the object name directly.

Read on to see how it all works. This is definitely a lot easier than in the olden days.

Leave a Comment

Reviewing Azure SQL DB Internals for Hints at SQL Server 2025

Published 2025-03-14 by Kevin Feasel

Brent Ozar does some digging:

Microsoft staff used to say that Azure SQL DB is “the next version” of the boxed product, meaning that Microsoft tested, tweaked, and proved new features in the cloud while they could still rapidly iterate over code, shipping updates to make the product more reliable and scalable. In practice, that’s not entirely true: some things never leave the cloud, and some things actually ship to SQL Server long before they’re available up in Azure SQL DB.

Today, we’re going to go through documented & undocumented system objects that are available up in the cloud today (March 2025), but keep in mind that some of these may be cloud-only. I’m just dumping out the list of Azure SQL DB objects that aren’t present in SQL Server 2022 CU17.

Read on for those sets, with the caveat that not all of this may necessarily be in SQL Server 2025.

Leave a Comment

Lock Escalation in SQL Server

Published 2025-02-20 by Kevin Feasel

Stephen Planck talks lock escalation:

Lock escalation is a built-in mechanism in SQL Server designed to optimize how the database engine manages locks. By reducing thousands of fine-grained locks (such as row or page locks) to a single table-level lock, SQL Server aims to lower overhead in the lock manager. In practice, lock escalation saves system resources but can also introduce challenges in high-concurrency environments. This post explores why lock escalation happens, how it works under the hood, the problems it can cause, and strategies to prevent or mitigate unwanted escalations.

Read on for the full story.

Comments closed

Checking out the mssqlsystemresource Database

Published 2025-02-17 by Kevin Feasel

Stephen Planck goes into single-user mode:

Have you ever wondered where SQL Server actually stores the definitions for system objects like built-in stored procedures, system views, and functions? Enter the mssqlsystemresource database—often just called the “Resource” database. While this database remains hidden from everyday use, it plays a pivotal role in the internal workings of SQL Server. In this post, we will explore what the Resource database is, why it’s necessary, and what every DBA should know about it.

Read on for a primer on what’s in this database and why we typically shouldn’t mess with it.

Comments closed

Drawbacks of Version Store in SQL Server

Published 2025-01-28 by Kevin Feasel

Haripriya Naidu points out that nothing comes for free:

With every feature comes both advantages and disadvantages and it is important to be aware of both. In regards to version store, we had discussed how to understand version store better and how version store is enabled.

Today, we will talk about the trade-offs and implications that come along when version store is enabled.

Click through for the downsides. With storage prices and speeds where they are, the tradeoffs of using the version store for RCSI or Snapshot isolation are almost always worth it. But it’s still good to know what those trade-offs are and how you can end up in a degenerate state.

Comments closed

The Contents of a Database Page in SQL Server

Published 2025-01-23 by Kevin Feasel

Simon Frazer takes a peek:

In SQL Server, data is stored in tables. Behind the scenes, however, these tables are divided into 8-kilobyte (8 KB) units called pages. Each page is 8,192 bytes in size, and this is a fixed value that cannot be changed.

Out of the 8,192 bytes, the first 96 bytes are reserved for the page header, which stores metadata about the page. This leaves 8,096 bytes for storing rows of data. However, a single row can only use up to 8,060 bytes, as 36 bytes are reserved for additional metadata, such as the slot array, versioning tags, and forwarded record pointers.

Read on for more information about the fundamental unit of storage in SQL Server.

Comments closed

SQL Server 2019 and Row Mode Memory Grant Feedback

Published 2025-01-23 by Kevin Feasel

Yvonne Vanslageren explains a solid feature in SQL Server 2019:

When SQL Server runs a query, it needs memory for operations like sorting and joining data. It also relies on memory during query compilation to hold intermediate plans while the Query Optimizer finds the best execution strategy. In parallel processing scenarios, the memory requirement grows even further.

SQL Server manages this by pre-allocating memory for each query through the SQL Server Operating System (SQLOS). This process ensures that no single query can monopolize the server’s memory.

Read on to learn more about memory grants, problems you can run into with memory grants, and one way SQL Server 2019 has improved to reduce the risk of bad memory grant estimates.

Comments closed

An Overview on Spinlocks in SQL Server

Published 2025-01-20 by Kevin Feasel

Stephen Planck talks spinlocks:

High concurrency can expose subtle performance bottlenecks in SQL Server, particularly those stemming from spinlocks and latch contention. Both mechanisms exist to synchronize access to shared data structures, yet they operate differently and require distinct troubleshooting approaches. By recognizing how they work and knowing what causes them to overload a system, DBAs can reduce CPU spikes, timeouts, and overall application slowdowns.

Read on to more about spinlocks and latch contention. My experiential bias is that spinlocks are the actual problem approximately 5% of the number of times that DBAs believe spinlocks are the actual problem.

Comments closed

Understanding the SQL Server Version Store

Published 2025-01-20 by Kevin Feasel

Haripriya Naidu digs into tempdb:

This article discusses when version store starts generating rows, when cleanup takes place, and how TempDB space gets occupied due to this activity.

I’ve hardly used PerfMon much in the past, nor was I a fan of it, just like many of you. But, when I used it a few times to determine how many data files needed to be added to TempDB while resolving contention, it helped me make a clear decision on the number of files needed. Since then, I’ve become a fan of PerfMon.
Although I don’t use it daily, I don’t hesitate to turn to it when I think it might help me understand things better.

Coming back to our topic of version store in this article, I resorted to PerfMon counters because visually watching the lines go up and down when an update is performed on a table helped me better understand how the version generation and cleanup works.

Click through for a demonstration of the process, including relevant PerfMon counters and the type of information you can get directly from SQL Server.

Comments closed

Reading Data from Deleted Columns

Published 2025-01-07 by Kevin Feasel

Michael J. Swart checks the typewriter ribbon:

It’s hard to destroy data. Even when a column is dropped, the data is still physically there on the data page. We can use the undocumented/unsupported command DBCC PAGE to look at it.

This is tied in with how we can drop a column in SQL Server and have it not take a very long time: because when we drop the column, we’re just modifying the metadata associated with the table and telling SQL Server to ignore this bit here. Do read the whole thing, and also check out a fun comment from Paul White.

Comments closed