text stringlengths 0 2.35k |
|---|
Jerod Santo: I heard that... |
**Angelica Hill:** Thank you both. I could babble to you all day, and it gives me such happiness just to see your faces, and I cannot wait to see them at GopherCon in real life! |
**Erik St. Martin:** Me too. |
**Johnny Boursiquot:** Yay! |
**Angelica Hill:** Have a great rest of your Fridays, and a great weekend. |
• Instrumentation in Go code and its purpose |
• Definition of instrumentation as generating signals to monitor application behavior |
• Types of instrumentation: manual (adding statements) and automatic (eBPF solutions) |
• Benefits of instrumentation, including health monitoring and metrics for distributed systems |
• Examples of instrumentation use cases, such as tracking running tasks and tolerating some downtime |
• History of Prometheus and its origins in instrumenting code for metrics |
• The Go instrumentation library's surprise popularity |
• Importance of instrumentation in software development |
• Different layers of observability (development, operations, business) |
• Challenges of building a solid instrumentation library (efficiency, API scope, dependencies) |
• Risks of improper or excessive instrumentation (increased compute power costs) |
• Optimizing for performance should be done after an initial working version is created |
• Popular libraries and features need to consider the cost of their usage, especially in terms of metrics and cardinality |
• The Go runtime metrics package introduced new challenges, such as increased metric count and potential costs |
• Balancing feature additions with backward compatibility and user needs can be a complex task |
• Changing code in popular packages requires careful consideration of breaking existing users |
• Adding new features to an interface or library often results in cruft and weird coding patterns |
• Importance of readability and maintainability in code |
• Trade-offs between performance and readability |
• Stability of APIs and their impact on community |
• Instrumentation libraries and their complexity |
• Designing APIs with optimization in mind vs. functional correctness |
• Use cases for micro-benchmarks to optimize code |
• Challenges of using multiple instrumentation tools and frameworks |
• Guidance for developers choosing the right tool for observability |
• Importance of considering organizational frameworks and existing infrastructure when selecting an instrumentation library |
• Implementing metric instrumentation can be simplified using libraries that abstract common metrics. |
• Exemplars provide information to show example situations that triggered metric increments or observations. |
• Deciding which exemplars to include is a complex task, especially with sampling strategies like tail sampling. |
• Sampling rates should be adjusted based on system understanding and volume/cost considerations. |
• Tail sampling involves saving trace spans after the request has completed. |
• Importance of monitoring in SRE |
• Introducing SRE concepts at universities and early in careers |
• Criticism of the term "observability" and its overuse/misuse |
• Monitoring vs. observability: is one a subset of the other? |
• Use of marketing terms vs. technical accuracy in industry jargon |
• Discussion of the importance of embracing simplicity in programming |
• Unpopular opinion that Go is the best foundation for every software, including embedded systems and machine learning |
• Criticism of creating new languages from scratch when existing ones like Go can be refined |
• Debate over the trade-offs between garbage collection, memory ownership, and other language features |
**Mat Ryer:** Hello, and welcome to Go Time. I'm Mat Ryer, and today we're talking about instrumentation, and instrumenting your Go code, specifically. I'm joined today by Johnny Boursiquot. Hello, Johnny. |
**Johnny Boursiquot:** Hello there. |
**Mat Ryer:** How's it going? It's been a while. |
**Johnny Boursiquot:** Yeah, it's been a while. Yeah, we haven't been in a show together for a minute. |
**Mat Ryer:** Yeah. |
**Johnny Boursiquot:** How are you, mate? |
**Mat Ryer:** Yeah, not for a while. Well, you don't just have to tolerate me on your own, Johnny. Don't worry. We have some guests today to dig into this subject. We've got Björn Rabenstein. He's here, he's an engineer at Grafana Labs and long-time Prometheus contributor. Hello, Björn. Welcome to Go Time. |
**Björn Rabenstein:** Hello. I'm glad that I'm finally on this podcast, of which I'm a great fan. |
**Mat Ryer:** Oh, well, we're great fans of yours as well. And we're also great fans of Bartek Płotka, who's also joining us. Hello, Bartek. |
**Bartlomiej Płotka:** Hello. Nice to be here. |
**Mat Ryer:** Bartek is a engineer at Red Hat, a maintainer of many open source projects in Go. You've probably seen his name around GitHub here and there. Yeah, maintainer of Prometheus and Thanos, and you're authoring Efficient Go. Is that right, Bartek? |
**Bartlomiej Płotka:** That's correct. Hopefully, published this year. |
**Mat Ryer:** Yeah, that's exciting. How is it writing a book? Is it what you thought it was going to be? |
**Bartlomiej Płotka:** \[04:06\] Definitely not. Definitely not. Lots of pros and cons. Lots of learnings. I think it's worth it, but maybe just once a lifetime. \[laughs\] |
**Johnny Boursiquot:** One book will do. Yeah. |
**Mat Ryer:** That's how you feel now. Maybe you'll get the buzz. |
**Bartlomiej Płotka:** Exactly. |
**Mat Ryer:** Yeah, you'll get a bug for it. Well, we'll see. Okay. So maybe we could just start at the beginning. For people not familiar, what is instrumentation? What's it useful for? Bartek, maybe you could kick us off. |
**Bartlomiej Płotka:** Right. So how I see the instrumentation as, essentially, it's about generating the signals that will tell us later on how your application is behaving when you don't have visible access to how it's running. So maybe you are putting that in production, in your cloud, maybe on your friend's machine... |
Now, of course, this can go even more complex, because you can have either manual instrumentation, so you are directly adding a code statement to your Go application, to add a log line, to add a metric, to create a span, tracing span, or maybe there is a Golang run time logic that creates profiling. And probably we can... |
**Mat Ryer:** Great. And so what's this useful for then? So we find out that we can find out things going on inside-- or what's like a specific example of something that you might want to instrument and then report on? |
**Bartlomiej Płotka:** So feel free, Björn, to chime in, but there are plenty of things. And I think first and foremost is something that we can read in a popular SRE book, right? So essentially, the Google team, who created this SRE movement, site reliability engineering - we can read that monitoring is like a key com... |
**Mat Ryer:** And so that's like a real endpoint running on the server that just returns some "okay." And so if you can reach that, then you know the server is at least up. |
**Bartlomiej Płotka:** Yeah, that's the basic solution. That is not the best, because if you're not looking, then it might be down in those periods when you are not poking it for 200 \[unintelligible 00:07:41.18\] |
**Mat Ryer:** Yeah. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.