Rails Business
Brendan Buckingham and Ryan Frisch talk about developing with Ruby on Rails and how to leverage it to build a business.
Rails Business
Challenges of Performance Monitoring in Rails
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
In this episode, we dive into the complexities of performance monitoring in Rails applications. Our discussion revolves around popular tools like Honey Badger, New Relic, Skylight, and App Signal, and the challenges of effectively using them. We explore specific pain points such as queue times, slow endpoints, and memory usage, and contemplate whether rolling your own solution is viable. Listen in for insights, tips, and relatable struggles in maintaining and optimizing Rails apps, especially for smaller teams and solo developers.
00:31 Challenges with Performance Monitoring Tools
01:17 Specific Tools and Their Limitations
02:07 The Cost and Complexity of Monitoring
04:02 Seeking Solutions and Alternatives
09:13 Database Monitoring and Additional Tools
14:18 Considering Building Custom Solutions
16:23 Seeking Help from App Signal
17:01 Discussing Nate Berkopec's Speed Book
19:10 Challenges with Performance Monitoring
22:12 Complexity of Building Web Apps
25:00 Prioritizing Development Tasks
29:16 Concluding Thoughts
LINKS
Hey, how's it going?
Ryan:Good. How you doing today?
Brendan:Not too bad. Uh, we'll, we'll, we'll stick to the programming stuff. I won't get into the rest of my day, which has been an interesting one, but nothing bad. Um, yeah. Yeah. What's on your mind?
Ryan:Uh, well, uh, a bunch of stuff, but I was thinking, um. You know, maybe we talk a little bit about how to monitor, uh, performance.'cause it's something that I've been banging my head against, uh, lately to the point where I'm thinking now, like, how do I roll my own performance monitoring?'cause the, that I've tried aren't giving me what I need for the, you know, compared to the, the cost of continuing to run them.
Brendan:Sure. Yeah, no, I, I've definitely got a lot of thoughts on that as well. Be, before we dive in, can you specify what you mean by performance monitoring? Because to me. Like the first thing that comes to mind is the new relics and the skylights and the centuries and all those. But then there's also, you know, bug reporting, honey badger, rollbar, those sorts of things. So what, what's on your mind specifically?
Ryan:Well, I mean all of the above really. Um, we use Honeybadger for error reporting. Honeybadger. Um, you know, for that, right? Like it's, in my opinion, above and beyond the other tools in terms of the back trace and, and actually being helpful in solving errors. Um, at one point we had New Relic that was just for, at least for me, was just complete overkill. Couldn't figure out how to
Brendan:Yeah, too, too many dials and
Ryan:Too much configuration. And then, you know, it got to a point where,
Brendan:I.
Ryan:the price was just, it, it was becoming absurd for what I was using it for. So then I, I switched to, I think, what did we just say Skylight, uh,
Brendan:Yep.
Ryan:you know, now I'm using App signal. You know, the, the challenge with, with all of these is, is you, don't need it very often,
Brendan:Mm-hmm.
Ryan:it. It's obviously useful. I want to have performance monitoring, but when I have had a performance issue, and for me that usually is like, you know, uh, q time is, uh, you know, backing up, you know, dynos are, are maxed out
Brendan:Yep.
Ryan:apps crashing completely. trying to figure out, you know, what caused that performance thing has been a real pain.
Brendan:Mm-hmm.
Ryan:so far I have not found a tool, you know, that has. Help me answer that question, so then I, you know, it's like, all right, yeah, it's great that I can see when the queue time is high, and that's nice. But that's only nice to a certain point of, of cost, right? Like
Brendan:Yep.
Ryan:it's like, all right, well, you know, we use Heroku so that, you know, there's some, you know, you can sort of see response time and we use higher fire for scaling. So like that reports the queue time, so I can sort of see what the queue time is,
Brendan:Right.
Ryan:know, by proxy. Um, you know, and it's like, all right, well now I'm paying a lot of money just to see queue time, I'm also feel like I'm just, I, I, I feel like I have to be missing something,
Brendan:Mm-hmm.
Ryan:Everyone talks very highly of these tools, and, and this is to not say that the, the tools are bad. I think this is probably like user error, uh, to, to some extent. But I don't know how to get over the hump. And if I can't get over the hump, then it's like, all right, well what's the point in spending, you know, three, four or five, $600 a month, which is sometimes, you know, where we're at right now. And you know, I. I, I appreciate. We're probably on the small end of, of their, you know, user base in terms of the
Brendan:You. You're paying less than me, that's for sure.
Ryan:um, you know, for something that has not proven to help me solve, you know, the problem when it's like the most critical, when our app is down, I.
Brendan:Yep. Well, I, I've got a lot of thoughts and questions, but before we even dive into those, I just wanna say that I feel a lot of the same ways you just described there with these tools. And one thing that I'm sure you feel as well, like I've, I obviously have a bigger team than you, but even with a team my size, I still struggle to keep up with these tools and even know what to look at. That's the thing that you, you touched on there was like, these tools are great to, I guess, have all this information, but like it's really bad at showing you the things you need to, to really be focusing on
Ryan:Yeah. And here's the other thing I'll, I'll throw in there. It's like, all right, you know, you, you read through the docs and, and app signals, docs for what it's worth are actually really good. Same with Honeybadger, right? Like,
Brendan:yep.
Ryan:know, sort of world class organizations. They're right up there with Stripe in terms of how they present everything. Like never been, you know, a concern. And I, I love, I love those guys over there, but, um, you know, it's like. I feel like in order for it to be like completely useful, I need to like push all the logs right into
Brendan:Yeah.
Ryan:in order to then see trends or whatever, right? But that becomes very expensive. So you know, the solution is like, all right, well pick and choose what you wanna into that. And I'm like, well, if my main concern is like slow endpoints.
Brendan:Yep.
Ryan:How can I pick and choose, right? Like everything's gotta go in or
Brendan:You gotta know which ones are the slowest compared.
Ryan:so, you know, I'm sort of stuck there, you know, for what it's worth. And, you know, maybe we can get those guys on at some point to, to talk about this, uh, in more detail, but.
Brendan:Yeah, I think, I mean, that's a great idea because I know, like as you were just talking there, one thing that came to mind that I think those guys have, although I don't use this tool at Honeybadger, that is, is they've got their insights, which I've listened to their podcast a few times, and that's the main understanding I've got of that tool is from the, their podcast. But what I, what I at least understand, or at least I think from what I've heard. Is you can kind of ask it questions for lack of a better way of, of putting it. So it's a little, little different, but I, I, I don't use it, so I don't know for sure. I, I don't know if you've tried it or not, I guess, I know you said you use Honeybadger for air monitoring, but
Ryan:Yeah, I, I tried to turn on, shoot, what do they call it? Law. Law or in Insights,
Brendan:I think it's insights. Yeah.
Ryan:I, I turned it on and, you know, I blew through my quota in about 15 seconds. Um.
Brendan:Right.
Ryan:So I reached out to those guys and I was like, all right, how do I, you know, how do I tune this? And then I, you know, I came back to that question of, you know, they gave me a good answer, right? Like, this is what you do. Right? But then I
Brendan:Yep.
Ryan:all right, well, I, I don't know what to exclude or not exclude. And I'm sure if I went back to them, they would continue to help me. But, you know, I
Brendan:Yep.
Ryan:got busy and I was like, I can't do this right now. So like, I'll come back to it. But yeah, it. I didn't get very far. It was nice to see the logs, right? Like I actually don't even have, you know, I don't even like use like a, was it log Stash or any of those logging
Brendan:Oh yeah. Yeah. Or like CloudWatch or,
Ryan:to log in production, I'll throw a Honeybadger dot notify in there and just report it into the error tracking, right?'cause it's a little bit
Brendan:yep. Mm-hmm.
Ryan:or I'll tail the logs on Heroku, which is pretty much useless. Um.
Brendan:Yeah. Right.
Ryan:up logging when I need it. But,
Brendan:Mm-hmm.
Ryan:yeah, I think the idea is that you can ask it questions and I think, you know, their approach to it is, is certainly different, but I think most of the, or, so I'll, I'll say, performance monitoring tools more broadly, try to do that in different ways. Right? And it's sort of like each one has its own flavor of doing that. So it's like, you know, what works for you. Um. But it's like, and, and I didn't get into Honeybadger further enough, but with AppSignal, it's like, all right, they have like what they call Time Detective, so you can see, all right, I'm having an issue at this point in time, and I
Brendan:Mm-hmm.
Ryan:it and it'll show me, all right, here are errors that occurred during that period. Here are alerts that are occurring during that period. But it still doesn't get me to like, all right, here's the slow endpoint that is
Brendan:Yeah. Right. This is what's causing it.
Ryan:this time.
Brendan:Yep.
Ryan:right. So it, it feels like a needle in a haystack. And, and pretty much every time we've had a performance problem, the issue has solely been solved by tuning threads, workers number of Dynos,
Brendan:Mm-hmm.
Ryan:I mean, there's always room for improvement in terms of response time, right?
Brendan:Yep.
Ryan:I know where the, the pain points are in our app that need to be addressed, but those don't ever come up in the, you know, in the analysis, right? It, like, it gets
Brendan:Yep.
Ryan:out, and if it does, it gets drowned out by the, you know, the being reported by Heroku, you know, for, you know, memory or whatever it may
Brendan:Right.
Ryan:um. Just
Brendan:Yeah. That.
Ryan:really hard problem for me right now is like I, I don't have the visibility that I wish I had.
Brendan:Mm-hmm. Yeah, and I can definitely relate to that, like I said earlier, and one other layer that I'll just throw in there, or, or tool I guess, but it's, I also look at it as like another layer of these types of tools that we've had for about a year or so. Um, Andy Atkinson actually was the one who kind of put, put me onto this is PG Analyze and that. Is another layer of, you know, just looking at the database. It's not looking at application code or endpoints or any of the stuff you've kind of been focused on there. And sometimes that's the problem. Sometimes that is the bottleneck. And I've found, you know, we use New Relic and New Relic doesn't like it has some database stuff, but it doesn't have exactly what you need. And so it's nice to have both of those.'cause that kind of helps dig in. Sometimes you find the needle in the haystack like you, you mentioned, but it doesn't always guarantee you're gonna find that. Um,
Ryan:yeah, so like AppSignal has, uh, you know, does both database and, know, response and q time, right? So like you, you
Brendan:okay.
Ryan:section you can go into and it's like, all right, here're slow actions or your slow external requests. And that's great 'cause I can, you know, they give me just a, a rank order list of slow
Brendan:Right? Yep.
Ryan:Right? But the problem is none of that gets, at least in my eye, gets tied back to the current issue. If I'm investigating something, it's like, yeah, I know that endpoint's slow, but it gets hit like once every,
Brendan:Yeah. Mm-hmm.
Ryan:there could be a thing that's like, all right, well this responds in 115 milliseconds, but it gets hit so frequently. It's like, know, it gets ranked as the number one problem. I'm like, yeah, I, I could shave 15 milliseconds off of that, but it's
Brendan:Yeah. But what's it gonna do?
Ryan:right? Like,
Brendan:Yep.
Ryan:the problem, that we're, we're dealing with here, right? Like, because, you know, we
Brendan:Mm-hmm.
Ryan:we have auto scaling enabled where, you know, we, we can, can handle pretty much any burst. So there's something else
Brendan:Right.
Ryan:going on. Um. That,
Brendan:Yeah.
Ryan:it just is like, all right, well, when I'm in a critical situation, I don't have time to go fix every, you know, slow endpoint until I figure out which one it is. I need to know like, what, is the thing that's causing this right now?
Brendan:Yeah, I don't have a great answer for that, but I can definitely, I, I've been there and even with a bigger team and I can, you know, delegate like you look at, you look here and I'll look here, that kind of thing. Like, it, it isn't always an easy answer, but I, I still feel like I'm missing something with these tools a lot of times. And it, it feels like too much noise in a lot of ways. Right. And yeah, like seeing the things, to your point earlier, like finding the things that are the most important. But you also mentioned this earlier, like if you only send over certain. You know, certain endpoints or maybe you try and like, you know, keep the cost down or whatever you're trying to do or, or reduce the noise. You're not sending everything, but how do you know what to send at that point? Because you might not be sending the most important thing if you are, you know, uh, re reducing or, or just kinda like weeding through those things and, and trying to reduce what you're sending over.
Ryan:Yeah, I, I guess what I really need is just a little bit more guidance of how to tune a Rails app.'cause really the things that I currently look at now are that, that I want to be able to see are queue time, right?
Brendan:Yep.
Ryan:wanna know if there's a queue and if there is, I wanna know what is causing it, which is, I think, a, a bit challenging.
Brendan:Yeah.
Ryan:I wanna know what the slow actions are and if there's any n plus one queries that I've, I've missed, right. That could
Brendan:Mm-hmm.
Ryan:relating to that. And lately I've been, uh, looking at memory footprint, um, you know, to try to get, I really wanna stay on a two x dyno at Heroku or lower,
Brendan:Yeah.
Ryan:don't want to get into the perf dynos 'cause you can't scale, uh, vertically as, as much like you max out at 10 and you know. Most of the time that's okay, but sometimes it's not. And I want
Brendan:Yep.
Ryan:to go higher.
Brendan:Right.
Ryan:you know, e everything else beyond that, I, you know, I don't have a need for right now. It doesn't mean that it isn't useful, but, you know, if, if someone could tell, like it feels like I need to report every controller action at,
Brendan:Yeah.
Ryan:the very least. And, and once I do that, I'm into the. You know, very quickly into the a thousand dollars or more a month plan just based on our volume. And you know, that is a lot of money. Um. When it's a tool that I don't feel like I can leverage when I need it, right? Like
Brendan:Right.
Ryan:if, if on the other hand, right when I needed it, I could go in and it would immediately pinpoint the problem and point me in the right direction, I would
Brendan:Mm-hmm.
Ryan:a thousand bucks a month for that.
Brendan:Yep.
Ryan:Right now I just get frustrated.
Brendan:Yeah, that makes sense. And, and just going back to where we started. What I'd like to follow up on now after we've kind of gone into the, the details here a little bit. Your pain, you know, you mentioned you're getting close to wanting to rule your own,
Ryan:Yeah.
Brendan:and if, if we compare what you just said a minute ago, which is you're, I know you're not spending thousands of dollars yet, but like your time to build something like this would be way more than that, at least initially. And so that's where one place I'm thinking maybe we start is what, what do you think? Or maybe like, where is your head? Like what? What do you think it is that you would build differently? And maybe you don't have a good answer, but I'm just, I wanna explore that a little bit because knowing your own app, maybe there's something there that you're thinking about that these tools aren't even doing
Ryan:No, the, the truth of the matter is. I enjoy the thought process. I'm never gonna build this thing. This is like way beyond my capability right now, both in terms of, I think, skillset and in available time.
Brendan:Right.
Ryan:However, you know, like everyone else right now, I threw it into chat, GPT and said, how do I measure these, you know, five things. Uh, I, I think adding the actual instrumentation to calculate and record queue time and n plus ones. The memory, I have no idea. But like those, those, like those slow endpoint things, like, I don't think the instrumentation would be that difficult.
Brendan:Yeah,
Ryan:the problem is that
Brendan:I.
Ryan:you, I am gonna be pushing an enormous amount of data through a pipeline and now I'm, I'm gonna have database things to deal with, like how, and then,
Brendan:Right.
Ryan:to do. And that's I think where. You know, you know, the lack of my understanding of how to do this would really, you know, make it a non-starter. But in theory, right, like at the very least, if I could report Q Time and say, all right, here's my q. Time over time. That would at least me something I, I don't know. What, um,
Brendan:Mm-hmm.
Ryan:would it be useful to be determined? My guess is on its own, probably not, and it would probably come with a, a maintenance burden. So I think my.
Brendan:Right.
Ryan:here, you know, if anyone from AppSignal is listening and they want to use me as a case study, you know, as part of this podcast or otherwise, you know, reach out, right? I'm, I'm open to learning, um, trying to figure it out. But right now I'm a little bit lost and, you know, maybe if you can, uh, get me Unlost and you can come on the podcast and help everyone else that maybe feels similarly so.
Brendan:Yeah, that, that makes sense. And I kind of figured your answer would be, yeah, I don't have time to build it, but I, I still think it's a good exercise to at least think it through a little bit. Right. Like what would, what would be the most important things, which you mentioned queue time especially. Um, and, and that makes my head also go to another question I was thinking about earlier when you were talking is, have you read, um, Nate Berkopec's speed book, um,
Ryan:Yes,
Brendan:Guide to Rail's performance or whatever it is?
Ryan:I have read that a couple
Brendan:Because E everything you said Okay. A couple times because everything you're saying is like leading me back to that book. But I don't have the answers off the top of my head of like what he said specifically. I just know that that's kind of the, the, the gold standard anyways for a Rails app.
Ryan:I, I read that a while ago at this point. Like,
Brendan:Yep.
Ryan:remember all the context, um, or contents. I definitely remember liking what I read
Brendan:Mm-hmm.
Ryan:and you know, I, I actually have some of, uh, the stuff on my book printed out and put on this bulletin board above my desk of like, you know, what should be the average response time or average load times. And I
Brendan:Yep.
Ryan:him and his, you know, tweets mostly these days, uh, fairly closely, uh, right.
Brendan:Yep.
Ryan:the sort of to go to on this topic.
Brendan:Right.
Ryan:but at the end of the day, when I get into it, I still have the same problems. And I, and I've, you know, I, I've worked with Andy Atkinson in the past too, and he's given me some really great stuff with, database tuning and, and things of, of that nature. But it's still like, it's not my area of expertise, right?
Brendan:Yeah. Right.
Ryan:I'm a, you know, I'm a self-taught developer who likes to build things. I think I'm a, a competent developer, but you know, when it gets into like these scale problems, it, it's, it, it's new, right? I mean,
Brendan:Yep.
Ryan:and you know, I'm not scared to say that it's something that I have to figure out. Uh, and I feel the same way about the performance stuff. I want to be the performance like. I wanna have the perf mindset. I don't wanna over optimize, but I definitely want to be able to get in and, and make improvements when necessary and understand where the bottlenecks are. Um, you know, just the same way I want to be, you know, uh, writing tests every day and get better at that and make that part of my thing. Like I think it is important. And here's the other piece that we haven't even touched on, which is, a lot of these tools do, uh, server side. Uh,
Brendan:Yep.
Ryan:monitor performance monitoring. I know, I know New Relic does the, uh, the client side now as well. At the time that I was on them, they didn't, they rolled that out after. Right. And I know there's a few other folks that do client side. It's like, all right, well right now everything is queue time. Like that's the metric I care most about at the moment. But it's like, all
Brendan:Yep.
Ryan:is the like, uh. What is the browser load time for my user? Like I'd love to like have better insights into that and make that a thing that I measure against, right. Because that would have like real, you know, I think impact, right? Um, you know, to our user base. Um,
Brendan:Yeah, that's a really good point. Just that there's so many layers to all of this, and that's another layer we didn't even touch on for sure. Um, and I struggle with that too, like we've gotta a React app on top of our Rails app, or integrated with, I guess it's all, it's all one monolith together kind of thing for us, but.
Ryan:In line, react for the win.
Brendan:Yeah. Yep, exactly. And that it is, you know, we use New Relic and it is a struggle to kind of know, like I'm, I'm still looking at the end points more often. I don't, maybe I don't have something turned on. Honestly, I don't even know. I, I, it's been so long since we set that up. Um, but you know, that. All of those questions and you know, even if you run, um, is it Lighthouse I think, or Google runs through Lighthouse, all those things that you get out of those performance tests and like the first Contentful paint, and I'm probably saying this all wrong, but all of those loading things are also very important and, and yeah, it's like a whole nother skill to get those right as well. And it depends too, if you're like running React versus, you know, stimulus or sure they're all similar, but what were you gonna ask?
Ryan:Well, I, I, I, I was gonna say, I think if I remember correctly, and I could be dead wrong on this, but in, for Nate, for the speed optimization handbook from Nate, it was like average load time, I think was a proxy for first content, full paint, right? So
Brendan:I think you're right. Yeah. We,
Ryan:fast
Brendan:we read that recently.
Ryan:seconds. Average is less than six. You know, slow is greater than six. So,
Brendan:Yep.
Ryan:I think that was first Contentful paint because full load, that would be, I, I think that would be, well, it's probably not that difficult. Uh, it is probably less difficult now with HTTP/2 and,
Brendan:Yeah.
Ryan:all the new rails, fun stuff. But, Not that I wanna go down a different topic, but we're from an upgrade perspective, like we're not quite ready to use all the new JavaScript and CSS bundling
Brendan:Right.
Ryan:So we're, you know, I'm stuck on a, you know, a single, uh. Concatenated file through the asset pipeline and that that
Brendan:Yep.
Ryan:is bigger than it should be. Right. So that slows you down. And, and, and you know, this spills over into testing too, right? Like, I've been debugging a thing where like my test week got super slow all of a sudden.
Brendan:Yeah.
Ryan:and I'm trying to figure out, it's like, all right, well the only thing that's really changed in any meaningful way, is that we've upgraded rails over that time period. So it's like, all right, did something change with. We went from, like, CI would run in, I think it was six minutes, which you know, isn't great, but it's not horrible. Right? It's like,
Brendan:Oh, I've heard I've, I've heard of and seen way worse, so that's pretty good.
Ryan:Um, but it went from six to like 12
Brendan:Yep.
Ryan:overnight with no new tests. Like maybe like one or two, but like nothing, like crazy.
Brendan:Just the rails upgrade. Yeah.
Ryan:I'm like, all right,
Brendan:Yep.
Ryan:changed.
Brendan:Mm-hmm.
Ryan:And, and then I, you know, I started looking into it and someone I was working with pairing with said that, all right, well, you don't have to. I run a lot of system tests, browser tests. Right. Because
Brendan:Yep.
Ryan:I'm still learning to test. We have a legacy code base. I'm trying to figure out how to add tests to it,
Brendan:Right.
Ryan:solution per, per Jason sweat is to add browser tests to start and then try to backfill you're
Brendan:Yep.
Ryan:refactor stuff. So like, all right. I'm taking that advice. I thought you had to put JS true, like as a requirement for every
Brendan:Oh.
Ryan:test.
Brendan:Mm-hmm.
Ryan:Apparently that's not true. So anyone else that, uh, did not know that, you know, you can remove that, as long as JavaScript isn't actually required to make the test, uh, run,
Brendan:Mm-hmm.
Ryan:leave that off and that that is shaving some time off. Uh,
Brendan:I was gonna say that that was like a config change you didn't, didn't know about. Yeah.
Ryan:so,
Brendan:Yep.
Ryan:the, the point of that is like, in terms of monitoring and performance, I was like, I've got some slow loading pages. It
Brendan:Yep.
Ryan:while for the full dom to load, but I have no idea why which pages. Right. I'm just gonna have to go through and, you know, sort of, uh, profile each test and say, all right, does removing js, first of all, can I remove js true? Like, is JavaScript actually necessary here to, you know, for the
Brendan:Yep.
Ryan:of the test? And if not, let's remove it and did it make a improvement or not? Um, that's, that's a painful process too, right? There's
Brendan:Right.
Ryan:easy answer there.
Brendan:Mm-hmm. Yeah, I mean, it, it's starting to feel like the theme of this episode is the complexity of, of building web apps in rails or probably anything really.
Ryan:Yeah.
Brendan:Um, and, and kind of like how, just how do you just monitor all this stuff? How do you know the things you should be working on? Is like another thing that I'm thinking of here, right? It's like you've got new features people are gonna want. Obviously bugs and performance issues and like, we don't have to get into all the details. We've talked about this stuff before, but it's just another reminder that there's a lot to do and when you're, you know, pretty much a one man shop like you are, I know you got consultants here and there, but for the most part you're a very small team and like how do you prioritize which ones to work on? It's probably just generally the most important thing, the thing people are complaining about the most, right? I mean.
Ryan:Yeah, I mean, I, uh, yes. Right. I mean, there, there's various layers of choosing the next thing. And at any given time, you know, my current philosophy, which has actually shifted a little bit since we had this conversation last, is like, all right,
Brendan:Hmm.
Ryan:one. Priority project, whether like, it doesn't matter if that's a feature request or a refactor or a little bit of
Brendan:Yeah,
Ryan:You know, one smaller thing that I'm trying to move forward and then, you know, you know, continued focus on
Brendan:I.
Ryan:You know, anything that's gonna impact the user base and making sure that those are handled quickly. Um, yeah, I, I don't, it's interesting right over the last. I mean, it's probably been 15 months now. We've been working to upgrade our apps and get them, you know, to the latest and greatest, uh, versions of everything. You know, we're now on rail seven two. We haven't gone to eight yet, but we probably will this year. You know, we're on the latest Ruby three four. We started, uh, upgrading Elasticsearch. That's what I'm in the middle of now. Right. And then there's like so many gems, right? And
Brendan:Yeah.
Ryan:you know, we identify these performance problems, right? And it's like, all right, I know this endpoint is a problem when it gets to a certain size of, you know, either records or query or whatever it may be. And it's like, all right, I need to refactor that. So it's just like, all right, well, you know, I try to think through what the problem is and be like, all right. How do I break this down? Right? Which is another topic we, we thought about talking about, right Again, right. Breaking down big changes into small chunks and we don't have to dive deep on it, but it's the, when I'm doing these sort of for performance or you know, to make other changes easier, they tend to. They tend to be required because they were poorly designed in the first place and they, their tendrils sort of spread across multiple places in the app.
Brendan:Yep.
Ryan:So fixing it is difficult because it's like, how do you fix it in one place, but not the others?
Brendan:Mm-hmm.
Ryan:Right? Like, if I fix it one place, then I gotta do it in the 10 other places that it occurs and all of a sudden, you know, something that I wanted to be like a quick, easy thing. It's like. And then. And then. And then,
Brendan:Right.
Ryan:right. And to me that's technical debt, right? Like.
Brendan:Mm-hmm.
Ryan:So for whatever reason, that's where we ended up. Right. And
Brendan:Yep.
Ryan:cases, that was probably my choice at some point. And, you know, my choice was probably good at the time, but it, you
Brendan:Yeah.
Ryan:it, it just, it, it lost its way, somewhere along the way.
Brendan:Mm-hmm.
Ryan:and now it's like, all right, I gotta clean this up. So it's like, I, I just feel like I'm in a constant mode of trying to get things where I want them to be. Um, and someone on my team reminded me. Either this week or last week. It's like, all right, well we actually have a really awesome app, Like our users love it. So that
Brendan:Yeah.
Ryan:right? Like I am the most critical person of the app
Brendan:Hmm.
Ryan:that exists, right? So that gave me a little bit of comfort. But
Brendan:Yeah.
Ryan:know, it's general desire is just to always make it better,
Brendan:Right.
Ryan:not for making it better's sake, but because I enjoy the process, right?
Brendan:Mm-hmm.
Ryan:um. of that is deciding what is important to work on and what is not. And I can tell you with a hundred percent certainty, uh, a lot of people don't agree with the things I choose on a day to day basis. You know, whether it's customers or my business partner or you know, internally the team, they're like, why are you working on that? I'm like, trust me.'cause like, I gotta get this done so I can get this other thing done that you wanted to, you
Brendan:Right, right.
Ryan:this other customer wants. It's like there's a domino effect. But
Brendan:Yep.
Ryan:anyway.
Brendan:Yeah, I mean, I, I struggle with that same exact thing, which could be a whole topic on its own, but just to kind of put a pin in that specific thing there, there's always that struggle in development, which is how do you convey to someone who's not a development that's puts on your team? That you need to do all this extra stuff, whether it's tech, debt, refactoring, you know, writing tests, whatever, so that you can actually accomplish the goals and the, the features or whatever that they're looking for. Um, and yeah, that's a, a constant thing I'm dealing with as well. And, no, no good answers there, but. Yeah. I, I, I will just add one other thing that you kind of made me think of. Um, I was just listening to yesterday. It's the the New On Rails podcast, and I will butcher, uh, Nadia's last name, but she's the, the person who runs StoryWorth or not StoryWorth, uh, story graph. I'll edit that out. Anyways, um. And, and she's, she's basically a solo developer as well. And so she had a lot of these same things that she was talking about, um, with Robby Russell, who's the host there, you know, they, they were talking a lot of the same things. So in some ways it's just kind of comforting just to know that we're not the only ones struggling with this. Rails gives us so much power to be able to do. A lot of this stuff, which is awesome, but there's always gonna be things to do. There's always gonna be new things to learn, figure out, keep up with. So it's an ongoing thing. But, uh, you know, I, I think it's, the more we talk about it, the more we can kind of come up with guests to talk about and just, you know, learn as we go is, is probably a good thing. So.
Ryan:So if anyone, uh, has anything to add, uh, hit us up. Uh, and come on and, and talk to us about it.'cause you know, I think it's an interesting topic.
Brendan:Yeah, definitely. All right, cool. Should we wrap it there?
Ryan:Yeah, let's do it.
Brendan:All right. See.
People on this episode
Podcasts we love
Check out these other fine podcasts recommended by us, not an algorithm.
Code with Jason
Jason Swett
Standing in the Fire
Very Good Software
Remote Ruby
Chris Oliver, Andrew Mason, David Hill
IndieRails
Jess Brown & Jeremy Smith
Startups For the Rest of Us
Rob Walling
Vibing with AI Code
Brendan Buckingham