text stringlengths 0 1.49k |
|---|
**Johnny Boursiquot:** Bye! |
• Wally Quevedo introduction and background |
• Overview of NATS project and its use cases |
• Purpose behind NATS in Cloud Foundry control plane |
• Resilience of NATS and why it's hard to crash |
• Transition from Ruby to Go implementation of NATS |
• Reasons for choosing Go, including performance and concurrency benefits |
• NATS is a high-performance messaging system that can be used for pub-sub, request-response, and RPC patterns |
• NATS has a "fire-and-forget" model, where messages are not persisted if clients are offline when they're received |
• NATS Streaming provides persistence and message redelivery capabilities, similar to Apache Kafka |
• The cost of durability in NATS Streaming is lower performance compared to regular NATS |
• NATS is highly flexible and can be used as a transport layer for microservices with libraries like Go Micro |
• There are numerous client implementations for various programming languages, including JavaScript, C#, Python, and more |
• Apcera uses NATS heavily in their own infrastructure, particularly in the control plane and service discovery. |
• NATS as a messaging system for low-latency communications |
• Benefits of using NATS (simple deployment, lower collective overhead) |
• Alternatives to NATS and when it's a good choice to use it |
• Performance improvements with new Go releases and NATS |
• Wally Quevedo's upcoming talk at GopherCon on Writing Network Clients In Go |
• Apcera's community-oriented culture and involvement in the Go community |
• Issue with MacOS 10.12.4 update breaking cgo-enabled binaries in Go |
• Call for proposals for Golang UK conference |
• GopherCon workshops announced |
• Go ERD tool for generating Entity Relationship Diagrams |
• Vim-Go 1.12 released |
• Emacs vs Vim discussion, including Wally Quevedo's use of Emacs and Domink Honnef's Go-mode |
• Discussion of NATS project, its evolution, and Wally Quevedo's involvement |
• Mention of a blog post by Nate Finch on his experience with Canonical and 500,000 lines of Go code |
• The speakers discuss how Go and its ecosystem have evolved over the past four years |
• They reminisce about the early days of Go when vendoring was not a concern and there were no external packages |
• Brian Ketelsen talks about his daughter's slime-making hobby and compares it to Oobleck |
• The discussion turns to #FreeSoftwareFriday, where they promote open-source projects, including Brian's work on Go Micro for microservices |
• They also discuss the increasing adoption of gRPC in communication protocols, including its use in Etcd and Kubernetes |
• The speakers briefly touch on NATS' plaintext protocol and their editors' preferences (VS Code vs Vim) |
• Carlisia has issues with Vim plugins not functioning correctly |
• Brian suggests trying different Vim plugins to resolve the issue |
• Wally recommends a Go utility called GHR for releasing NATS artifacts |
• Erik gives a shoutout to Kubernetes maintainers for their work on recent releases |
• The group discusses KubeCon in Berlin and related projects |
**Erik St. Martin:** Welcome back, everybody, for another episode of GoTime. Today's episode is number 41, and our sponsors for today are Backtrace and The Ultimate Go Training Series. |
On today's show we have myself, Erik St. Martin, Brian Ketelsen is also here - say hello, Brian... |
**Brian Ketelsen:** I'm the assassin. |
**Erik St. Martin:** \[laughs\] And we also have Carlisia Pinto. Say hello, Carlisia. |
**Carlisia Thompson:** Hi, everybody! |
**Erik St. Martin:** And our special guest for today is an engineer with Apcera, working on [NATS](https://nats.io/) and also a speaker at GopherCon this year. Please welcome Wally Quevedo. |
**Wally Quevedo:** Hi everyone, thank you for having me. |
**Erik St. Martin:** So just to get started a little bit, for those who may not be familiar with you personally, we'll start there - who you are and the things you're working on, and maybe we'll talk a little bit about NATS and what it is and the use cases it solves. |
**Wally Quevedo:** Yeah, so I'm Waldemar, but most people call me Wally. I'm a software engineer at Apcera and I do development of the Apcera platform, which is a container orchestration solution. Also, I'm one of the core maintainers of some of the official NATS clients \[unintelligible 00:01:50.29\] original clients.... |
**Erik St. Martin:** So a lot of polyglot stuff, trying to maintain all the libraries for the different languages. |
**Wally Quevedo:** Yeah, and even though the NATS project started in Ruby, the canonical implementation these days now is the Go client. Go is a very important part of the project, because it's the one that we try to keep as a reference implementation... Basically, the spec that you have to implement for each one of th... |
That's where I started getting into the project, and it's been really cool seeing how the NATS project has been evolving, thanks to the Go community as well. |
**Brian Ketelsen:** Let's start talking about the purpose behind NATS. I understand NATS came out of Cloud Foundry and it was the control plane for all of the components inside Cloud Foundry... What purpose does it serve inside Cloud Foundry? What's unique about NATS that makes it so good for that control plane? |
**Wally Quevedo:** Yeah, it was a piece that would allow you to do the internal communication throughout the platform. I think with Cloud Foundry it did really well because it was simple and resilient enough for the type of communication patterns that you would have inside of that type of architecture. We ended up havi... |
\[04:02\] I really like the simplicity of doing operations with NATS. It just fits really well for that type of usage. They're used for fire-and-forget, request-response, basic communications for starting applications and heartbeats for fault-tolerance, for example. Does it make sense? |
**Brian Ketelsen:** It does make sense. Now, one of the things that I hear very frequently is that NATS is pretty bulletproof, and I just saw a tweet (maybe yesterday), a person was talking about their infrastructure and how they had to update a bunch of things, and when they were messing around with the servers, they ... |
What is it about NATS that makes it so resilient? Because that's a very common message I hear - it's hard to kill NATS. |
**Wally Quevedo:** Yeah, I like that with NATS a lot. As a person that was operating on these large clusters, having the uptime, having everything available was critical. I like that confidence that it was going to be basically around and available for the components to be able to communicate. I guess it is because of ... |
If you have a client that is trying to do some harm, the server will basically take out the connection, and other clients would still be allowed to communicate and basically not be impacted by these other parts of the system. So yeah, it is true... |
At the beginning - with the original Ruby implementation - they didn't have clustering from day one, but it was still possible, for example, to do basic failover scenarios with NATS, and the recovery was pretty fast still. So yeah, I guess it's well suited for that kind of a use case. It really does stay up for long pe... |
It's very difficult to crash. I've managed to crash it a number of times, but it was by making these weird clients trying to attack it, basically. Those are all fixed now, but I like finding these weird corner cases with the buffers from a NATS server where you could make it crash... But they have all been fixed in mas... |
**Carlisia Thompson:** You mentioned the original implementation in Ruby... |
**Wally Quevedo:** Yeah. |
**Carlisia Thompson:** I'd love to go there... Were you there when that happened? Were you already at Apcera? |
**Wally Quevedo:** No, I was living in Tokyo for around five years, and that's where the Cloud Foundry was originally implemented. The company I was at, they were looking for basically like a platform as a service kind of system, pretty much like what Heroku worked for, but for an internal for the company... Fortunatel... |
**Carlisia Thompson:** Because what I wanted to ask is how much you know about the transition, the motivations behind transitioning from Ruby to Go... Were there benchmark tests? Did they consider other languages? What was it that made Go the choice to transition into? What kind of problems were they having that they f... |
**Wally Quevedo:** \[08:04\] It's worth mentioning that the original server was also within an event machine. A good reference for this is [the talk from GopherCon from Derek](https://www.youtube.com/watch?v=ylRKac5kSOk). I think it was at the first GopherCon, in 2014. |
**Brian Ketelsen:** It was, yeah. |
**Erik St. Martin:** Yeah, he spoke at the first one. |
**Wally Quevedo:** Yeah. So actually this year it's going to be my first GopherCon, I'm really looking forward to it. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.