I was curious if I can use new Lambda MicroVMs as self-hosted GitHub Runners. On paper, they are super nice:
It's cheaper: GHA-hosted is $0.005 / min (2 vCPU), MicroVMs ~$0.0042 / min, and no minimum 60-second commitment as with GHA-hosted.
It can run longer: GHA-hosted max 6 hours, MicroVMs max 8 hours
It starts in a few seconds, compared to whichever other serverless solution built on top of ECS
It scales to 0, or rather, it only runs when jobs are running
They are VMs, so you can still run containers/docker/whatever else inside;
I got a bit too invested, and ended up building this Terraform module. You only need to create GitHub App manually, the rest is just a single "terraform apply" and your MicroVM Runners are ready to go. I've switched come of projects at my company to use, works great, same or better performance as GHA-provided runners. Natural limitation is that MicroVMs are only arm64, and in general they don't have much flexibility around the "hardware" setup - but hey, for most cases, it should work great, and it's just 1 webhook + GHA JIT Runners + 1 MicroVM Run per Job.
Amazing what one can vibe-code these days. I am not an iOS dev, but I am pretty technical, focused mostly in cloud infrastructure and architecture. I like playing video games, and this new Forza Horizon 6 is fantastic. While going through the menus, I saw that it can send telemetry data to any IP over UDP, so I was curious how far I can get with creating an app that would serve as a server and visualize this data. Hours of having fun tweaking it (good thing is that to test this app, I need to play Forza), and 3 weeks of waiting for Apple approval, the app is out and is free to use with no ads. Hardest part was to build a race log, detection whether that car is on a particular race/track in the game, and attributing race attempst to these tracks - but I managed to do it!
Hey, HN! So my wife is writing her thesis and she needed to do 12 hour-long interviews and then transcribe them to text. I've built this small product for her and it saved her a ton of time. It uses around a dozen AWS services, most importantly AWS Transribe and AWS ECS Fargate, and is built on Ruby on Rails, with Stimulus.js and Bootstrap. I've shared complete stack on StackShare: https://stackshare.io/transcripto/transcripto. I tried to keep it simple and make it as affordable as possible.
I agree that the final “thing” that runs containers and provides isolation is Linux kernel, it’s also pointed out in the article :-) Thanks for the feedback! Bocker looks like a fun little project indeed
In this case it's worse: there was _no_ complaint on _our_ business. They just started blocking millions of IPs only to ban Telegram. And we are not Telegram. :)
I didn't mean calling tools themselves "junk", but rather dozens of auto-generated files which don't give me any profits at this time. I am pretty sure all these things make a lot of sense, and we are getting to the point where trying out WebPack might make sense for us.
I am very careful with picking mentors and I generally just trust them to do their job. :-) It might not scale well long term, but that's the problem I don't need to solve right now. Right now if some lead tries to go directly to mentor, mentor always 1. Reports it to me; 2. Tells this lead to pay through mkdev.me.
It's simply much more convenient to mentor through mkdev.me than to do it on ones own.
I made a platform where people can hire a mentor to learn programming (ruby/rails, frontend, devops, big data). I've focused on mentorship with a per-week payment, so not a 1hour tutor or 15min "solve concrete task" service, but a real, long-term mentor dedicated to sharing all the knowledge with students.
Initial customers base grew from the free ebook I wrote and various blog posts available on the website. Lots of success are due to email newsletter, which I'm trying to keep useful and rarely send any ads there.
I'm taking 20% from each payment and that's be growing quiet well so far, especially after entering english-language market. Website is https://mkdev.me/en
It's cheaper: GHA-hosted is $0.005 / min (2 vCPU), MicroVMs ~$0.0042 / min, and no minimum 60-second commitment as with GHA-hosted.
It can run longer: GHA-hosted max 6 hours, MicroVMs max 8 hours
It starts in a few seconds, compared to whichever other serverless solution built on top of ECS
It scales to 0, or rather, it only runs when jobs are running
They are VMs, so you can still run containers/docker/whatever else inside;
I got a bit too invested, and ended up building this Terraform module. You only need to create GitHub App manually, the rest is just a single "terraform apply" and your MicroVM Runners are ready to go. I've switched come of projects at my company to use, works great, same or better performance as GHA-provided runners. Natural limitation is that MicroVMs are only arm64, and in general they don't have much flexibility around the "hardware" setup - but hey, for most cases, it should work great, and it's just 1 webhook + GHA JIT Runners + 1 MicroVM Run per Job.