1
6 Comments

Track Technical Performance

Hey All,

I was curious, how are you tracking the technical performance of your applications post-launch? Are there any tools you use to understand and track reliability? How granular do you actually care about performance? Are there any insights into performance you wish you had?

Professionally, I am an SRE. This is mostly out of curiosity :)

  1. 1

    I thought I might chip in, as I just launched an App. When I was building it (with SpringBoot) I did not added any kind of metrics using Graphite. This is mainly due to the need of having an Graphite installation.

    Which makes me curious if there is something with Spring Boot that generates out of the box metrics for number of requests and status codes(2xx/4xx/5xx). Maybe I'll have to google this.

    Do you have any suggestion that you would make that is really just out of the box and hopefully free or cheaply available? as that was one of the criteria that I had in mind to have minimal deployment cost. while not having any paying customer.

    Now that I have launched an App, I am actively looking to add some Business + Performance metrics to see how the Backend is doing.

    1. 1

      I was curious about this exact thing - I was wondering if there was a solution indie hackers, and possibly small startups, gravitate to. I've been kicking around the idea of building a project for your use case. A solution that would keep a very limited number of metrics/logs for realtime investigations, but provide longterm metadata analysis on both. Most solutions cost a lot because of the amount of data generated/stored, with the burden of doing metadata analysis being on the dev/team. But I feel like it is not something most solo devs/small dev teams actually care about.

      Anyways, more of a rant than help for you. I did find a few opensource projects, including netdata, that would allow you to handle exporting logs and metrics: https://www.netdata.cloud/open-source/

  2. 1

    If there are no customers you don't care.
    I don't think many people care or should care early on unless the system specifically needs some sla level or has traction. It's easy to add with many services depending on the system. If it's catastrophic always down you would probably know without a system

    1. 1

      @hatkyinc For sure, if there are no customers you shouldn't care. Who do you need to make this thing reliable for? But I do find it interesting that we, or I should say most folks, would find writing tests necessary(maybe not initially, but certainly when people are actually using your product). But when it comes to how understanding how things are working after code is deployed, it's a second thought. It's hard to implement monitoring with current tooling, especially with the cost. Should we ever care to know more than "it's down"? Is it enough to wait for customer feedback that things are working poorly? Anyways, maybe the answer to those questions is yes. Just curious where indie hackers land on this subject.

      1. 1

        You should think about it in steps, you're probably focused on APM level, but

        First of all most downtimes would be deployments, and early you might not have even 1 deployment a day or a week

        If the system is down for multiple hours - multiple days either

        • A customer would let you know
        • You would stumble onto it test/showing
        • It would show up with interaction/traffic numbers that you would care more about

        There are cheap uptime options
        some of these say 5$/month systems that check like a 200 if the page is up and even if it has some expected text. (these expend offerings to world wide synthetic test, multi-page processes testing etc)

        APM level is really worth it when you consider revenue loss per hour/min greater than the time cost of the person needing to fix it for a couple of hours

        Over-engineering for most cases / premature optimisation

        Just some thoughts

        1. 1

          This is all great, totally makes sense. I agree that for most projects/indie hackers, this would be over-engineering. I am curious if there is a threshold where folks would be more curious about the performance beyond general checks, like text on a page with a 200 response code. But that probably is pushing more into small startups with more time and resources. Or possibly niche projects where early on, performance is a key selling point.

          Thank you for the feedback, this was helpful!

Trending on Indie Hackers
Here's how we got our first 200 users 30 comments Reaching $100k MRR Organically in 12 months 25 comments What you can learn from Marc Lou 20 comments Software Developers Can Build Beautiful Software 13 comments Worst Hire - my lessons 11 comments How to Secure #1 on Product Hunt: DO’s and DON'Ts / Experience from PitchBob – AI Pitch Deck Generator & Founders Co-Pilot 6 comments