Press "Enter" to skip to content

Day: February 7, 2025

Handling Errors in Apache Flink Apps

Alexis Tekin and Jeremy Ber handle an error:

Data streaming applications continuously process incoming data, much like a never-ending query against a database. Unlike traditional database queries where you request data one time and receive a single response, streaming data applications constantly receive new data in real time. This introduces some complexity, particularly around error handling. This post discusses the strategies for handling errors in Apache Flink applications. However, the general principles discussed here apply to stream processing applications at large.

Read on to see how this all works when you’re hosting a Flink application. This directly relates to Flink applications that live in AWS, though very little in the article is AWS-specific.

Leave a Comment

Understanding Error Severity Levels 19 through 25 in SQL Server

Steve Stedman handles an error:

SQL Server uses error severity levels to indicate the nature and seriousness of an error. These severity levels range from 0 to 25, with each level indicating a different type of issue, from informational messages to catastrophic failures. In this post, we’ll focus on severity levels 19 through 25, which are typically the most critical errors that can occur in SQL Server.

Fortunately, given how devastating they are, severity level 21 and higher are extremely rare in my experience.

Leave a Comment

The Power of an Error Message

Joey D’Antoni shares some thoughts on error messages:

I’ve been working in technology for entirely too long—but one of the first, and most important lessons I learned was try to identify the correct error message, in the right log file, and then it can be generally trivial to resolve that error message. The challenge to forever and ever always has been “finding the right error message”, as what typically happens is that once there is one failure in a computer program, there is a series of cascading errors, which makes it harder to identify the right error message that you need to copy and paste into google/stack overflow/chatgpt/copilot.

Read on for more. I think it’s easy for us to rag on bad error messages, though even opaque error codes can still be reasonable if there’s a comprehensive guide. This was a classic problem with blue screens: oh, I have an error code 0x2378f83f. So what does 0x2378f83f mean? Dunno, nobody else on the internet has apparently ever had that code. And Joey also brings up my favorite error message to rag on: “String or binary data would be truncated.” Sadly, although they did finally add column name and message text to the error, they never fixed the most galling mistake in that error message: the passive voice.

Leave a Comment

Kusto Query Performance in Microsoft Fabric

Dennes Torres checks some stats:

We already discovered how to investigate Kusto query history. Let’s discover how to analyse query performance considering the information on this history.

The query history returns 3 fields we can use to make a more detailed analysis of the queries: CachedStatisticsScannedExtentsStatistics and ResultsetStatistics.

Disclaimer: There are low to no documentation about this content. In this way, the content below may not be 100% precise but will give you good guidance.

Click through to learn more about these three.

Leave a Comment

Visualizing a SQL Server Kubernetes Statefulset

Andrew Pruski builds a diagram:

The other day I came across an interesting repo on github, KubeDiagrams.

What this repo does is generate Kubernetes architecture diagrams from Kubernetes manifest files…nice!

Deploying applications to Kubernetes can get complicated fast…especially with stateful applications such as SQL Server.

So having the ability to easily generate diagrams is really helpful…because we all should be documenting everything, right?

Click through for instructions and a couple of gotchas Andrew ran into along the way.

Leave a Comment