text
stringlengths
0
1.82k
**Peter Bourgon:** Yeah, exactly.
**Carlisia Thompson:** I don't think the stars are so meaningful. I think it shows interest... I was actually just looking at Go Kit's channel Gopher Slack and it's got over a thousand people.
**Erik St. Martin:** Wow.
**Peter Bourgon:** Yeah, that's super surprising to me, and I guess... Actually, if you go to the Gopher Slack list of channels by members, I think it's top five, or something.
**Carlisia Thompson:** Yeah.
**Erik St. Martin:** Well, that's crazy though... Because when you think about it in context, that's 10% of all the people in the Gopher Slack, total. For one project. That's good.
**Peter Bourgon:** I'm into it.
**Carlisia Thompson:** Yeah, and if you're taking into account that a lot of people are not even active, that might be more like... It's a much higher percentage of the active members.
**Erik St. Martin:** Right, and a lot of people don't hang out in all of the channels. Not being in there doesn't mean lack of participation in it too, so...
That's always hard, like how do you measure involvement in a community (in any community)? How many Go programmers are there? It's hard to even guess. You could be off by an order of magnitude. You try to look at projects, or conference attendees and things like that, and make an educated guess, but we'll probably all ...
**Peter Bourgon:** \[55:52\] Right... The great, dark mass of unspeaking programmers. I hear it's like, 80% is probably a good initial estimate. You only get interaction with about 20% of your users.
**Erik St. Martin:** Wow. I knew it was a big gap, but that's huge. So we're saying 80% of people who use any given technology typically aren't visible in public communities...?
**Peter Bourgon:** I've heard that statistic bandied about, but I can't put a source to it right now. But intuitively, that kind of makes sense to me.
**Erik St. Martin:** That's crazy. So I know that we have talked about Go Micro on this show too, and I'd love to get your opinion on some of the other microservices frameworks and how they related to each other and what you see as benefits and drawbacks. Do you feel that you address the problem differently, that you s...
**Peter Bourgon:** Yeah, I would say so. I would say my approach is pretty fundamentally different to the Go Micro approach. I know him by reputation, if not super well in person; we've had a lot of conversation in this space, obviously... His Go Micro project is, from where I sit, it's a lot more of a batteries-includ...
I guess it's a lot more opinionated in the sense that to boot up a service you're going to implement a particular interface with lifecycle management methods on it, and you're going to absolutely hook it up to a service discovery system and it's going to interact with a service discovery system, and you're absolutely g...
Go Kit, in contrast, is coming from the angle of "You already have in your organization an infrastructure, and you've already chosen what you're gonna use for service discovery." Or maybe you're not using anything for service discovery; maybe you're doing it completely manually. Similarly with your transport layer - ma...
**Erik St. Martin:** I think it does perfectly. That's kind of where my view has always been with... You know, Micro is kind of the "Here's your whole layout, here's how you do everything", and Go Kit is more of a framework that allows you to make decisions, but "Here's some paradigms and ways that are known to work we...
**Peter Bourgon:** \[59:56\] Exactly. And here's integrations with these common components, common service discovery systems, common instrumentation systems. Whenever I give talks about Go Kit I say, "Fundamentally, it's about leveling up software engineering, and leveling up the way you build microservices, like code ...
**Scott Mansfield:** That's quite a tall order.
**Peter Bourgon:** How do you mean?
**Scott Mansfield:** It takes plenty of time for an organization to adopt all those things. I guess being somewhat modular would allow them to do that, but having all of that all in one go is quite a lot to learn for people who don't have that deeply ingrained.
**Peter Bourgon:** "Having all of that" meaning all of these software engineering principles, or all these components?
**Scott Mansfield:** All the software engineering principles that you just named; that's quite a lot for people to absorb if they're not already doing a lot of those things.
**Peter Bourgon:** Oh, for sure. And for that reason, I view Go Kit as more than anything else a sort of educational enterprise where hopefully you can kind of start easy, start simple... There's one big hurdle at the beginning, but then you can see okay, once you get over that, where all the other pieces line up.
**Erik St. Martin:** I think that's kind of the fun part too, that it's almost like a recipe book. You come in here and you're like, "Okay, I think the first thing we need to do is solve service discovery", and then there's a list of things that are known to work well together. It's like, "Okay, let's take on the servi...
**Peter Bourgon:** Yeah, totally. And if you're a two or three-man shop and you're starting fresh on brand new infrastructure, maybe Go Kit is a bit too much of an initial hurdle. Maybe there's too much there, and maybe you get a lot more productivity by starting with something like Go Micro where a lot of the decision...
**Erik St. Martin:** The other thing I wanna point out just looking through the list of all the things that are available in Go Kit (the list of components) is four years ago we were writing almost all of our Go code from scratch, and now nobody's implementing new service discovery mechanisms really, right? Those thing...
**Peter Bourgon:** I hope not.
**Erik St. Martin:** It's just crazy to think that stuff - that things that four years ago everybody had to build for themselves, we just don't even worry about it anymore. When you see somebody build it for themselves, you're like "But why?!"
**Carlisia Thompson:** Peter, give us some insights here as far as interoperation between Go Kit and other things, or self-written code. Can you use Go Kit logging together with Micro, or together with my own code, or just metrics? Is it possible?
**Peter Bourgon:** \[01:03:46.09\] Yeah, totally. In Go Kit there's a bunch of packages, and I would broadly say there's two types of packages - there's ones that you can use completely independent of anything else, that you can drop into your existing codebase with no other changes and just reap the benefits of that p...
Same thing with metrics, for example. The instrumentation package defines a set of common interfaces that are broadly applicable, and implementations that connect it to (I think the latest count was) 8 or 9 common metric and instrumentation systems, Prometheus being chief among them, in my view. So this type of stuff y...
There's other packages that sort of rely on your microservice being structured in a particular way, using the endpoint abstraction and the transport abstraction. This requires a bit more buy-in, that's kind of the hurdle I was talking about. But if you jump over that hurdle, if you buy into these abstractions that I've...
**Carlisia Thompson:** Yeah, I was trying to figure that out, where the boundaries were for each of the things.
**Peter Bourgon:** Yeah, and I could probably do a much better job of giving introductory, on-ramp style documentation on the website. Right now I've been so focused on the advanced use cases I've kind of let that atrophy a little bit... So I'll put that in my queue.
**Carlisia Thompson:** Cool.
**Erik St. Martin:** I think that we are basically out of time, but did anybody wanna talk about JBD's new tool before we roll this thing out? Because that's freaking cool, the gops tool. I wonder how it's supposed to be pronounced...
**Carlisia Thompson:** Yeah, I was wondering...
**Erik St. Martin:** I think it's gops.
**Carlisia Thompson:** I would say gops.
**Erik St. Martin:** It's this really cool debug tool for Go processes on your machine. You can run gops stack and pass it the PID of a Go process and you can actually see the current stack and you can get GC information and memory statistics from a running Go process.
**Peter Bourgon:** Yeah, it's something I haven't been able to use in anger yet, but I'm super excited about the potential here. Because we all know Go processes are really introspectable in theory, but I can't be the only one when looking at something in a staging environment, for example, and having to remember how t...
**Erik St. Martin:** Yeah, that's kind of the fun thing about having a runtime, is that all this stuff exists in there, you just kind of have to poke at it. I've only used this from poking and prodding things that are running just out of sheer curiosity and playing, but again, I look forward to trying that where I actu...
I think Scott disappeared on us. Where have you been at, Scott?
**Scott Mansfield:** \[01:07:54.23\] No, I'm still here. I was actually looking at the code while you were talking; it looks like you have to install an agent in the processes that you wanna introspect, so that it will open this UNIX socket in the temp directory, so that the gops program can actually go inspect things ...
**Peter Bourgon:** Oh, interesting.
**Erik St. Martin:** Yeah, we should definitely clarify that. It's not that you can take any running Go process and do that. It works similar to the way you can expose the runtime stats over HTTP, there is kind of a library for it... But this adds more functionality than just exporting the -- how's that package pronoun...
**Scott Mansfield:** Pronunciation just wasn't important at the time.
**Erik St. Martin:** It's like, I have two words, how do I make it one word? Here we go, expvar, I love hearing people's pronunciations of stuff. Peter, you use Kubernetes, do you call it KubeCuttle (Kubectl)?
**Peter Bourgon:** KubeCuttle, yeah, sure.