Really dig the candid, humble conversation and insights into the larger ecosystem. I mean, we all know Vercel and DHH are pushing their agenda and worldview from a business perspective and though nobody needs to call it out, conversations like these IMO make Fly more relatable on a developer level because it’s the definition of anti-marketing
The points surrounding abstraction complexity and fly’s primitives are well received. However, where I get stuck often is in how to work with fly.toml to properly use them. Would love a 30-minute in-depth video on how that file works and all the ways I can configure an app environment… and where it breaks. (Like trying to run an $envvar in a process string.) Keep the videos coming!
I'm sure I was recommended this because of earlier watching the video "lol don't do docker tho" and I quite enjoyed this "lol don't do the whole cloud tho" video
Just on the point about data egress. AWS does actually let you pull your data back out as Hard Disks (via AWS Snowball). It's not just one way transfers. Egress costs for AWS, Azure and Google is now free for the purposes of leaving them as vendors (but not free if you just want a subset of your data).
Snowball is terrible though - or at least was when we tried it 4-5 years ago. The IO on the hardware you get is somehow throttled, so copying data off of them takes weeks or months.
You don't have to use all the different services. You can just spin up some instances and run your workload on EC2 if you know what you're doing. You can install docker, mysql etc under Linux or windows... You don't need ECS/EKS or RDS if you can DIY.
Yeah, that’s something I‘ve seen lately, lots of dev will go into serverless because they don’t want to deal with setting up a VPS themselves. It’s probably a week to get a leg up on the basics, and maybe a week more (well… max), to understand how to maintain it basically. On the other hand, serverless scale auto-magically (only your bank account feels the pain of that one) when scaling out of an EC2 instance is ‘harder’. I guess I blame that on that « ship quickly » mentality where a dev/entrepreneur has to absolutely ship whatever the cost. Well… it shows one thing, Google and AWS taught us well not to ask too many questions about their methods ;-)
The rails people aren't recommending running your own database really. Most of us push people to use Crunchy or similar. It's more for stateless app/job servers.
Really dig the candid, humble conversation and insights into the larger ecosystem. I mean, we all know Vercel and DHH are pushing their agenda and worldview from a business perspective and though nobody needs to call it out, conversations like these IMO make Fly more relatable on a developer level because it’s the definition of anti-marketing
I love these types of conversations. Thanks for sharing.
Microphone holder
Very fancy
😊
The points surrounding abstraction complexity and fly’s primitives are well received.
However, where I get stuck often is in how to work with fly.toml to properly use them. Would love a 30-minute in-depth video on how that file works and all the ways I can configure an app environment… and where it breaks. (Like trying to run an $envvar in a process string.)
Keep the videos coming!
Totally agree with you there. I'm struggling to stand up an application which has both HTTP and grpc endpoints
I'm sure I was recommended this because of earlier watching the video "lol don't do docker tho" and I quite enjoyed this "lol don't do the whole cloud tho" video
This was a point DHH made in his RailsWorld keynote!
Super interesting, hope you do more vids like that, fun and informative !
Really loving these videos.
Just on the point about data egress.
AWS does actually let you pull your data back out as Hard Disks (via AWS Snowball). It's not just one way transfers.
Egress costs for AWS, Azure and Google is now free for the purposes of leaving them as vendors (but not free if you just want a subset of your data).
Snowball is terrible though - or at least was when we tried it 4-5 years ago. The IO on the hardware you get is somehow throttled, so copying data off of them takes weeks or months.
You don't have to use all the different services. You can just spin up some instances and run your workload on EC2 if you know what you're doing. You can install docker, mysql etc under Linux or windows... You don't need ECS/EKS or RDS if you can DIY.
Yeah, that’s something I‘ve seen lately, lots of dev will go into serverless because they don’t want to deal with setting up a VPS themselves. It’s probably a week to get a leg up on the basics, and maybe a week more (well… max), to understand how to maintain it basically. On the other hand, serverless scale auto-magically (only your bank account feels the pain of that one) when scaling out of an EC2 instance is ‘harder’. I guess I blame that on that « ship quickly » mentality where a dev/entrepreneur has to absolutely ship whatever the cost. Well… it shows one thing, Google and AWS taught us well not to ask too many questions about their methods ;-)
Just get an VPS with a smaller footprint
Also the video ends abruptly... Something missing?
Nope - we just didn't film an outro. This video was a little experiment, we'll try to have a proper ending next time :)
That was a valid reaction to someone hosting CockroachDB
The rails people aren't recommending running your own database really. Most of us push people to use Crunchy or similar. It's more for stateless app/job servers.
Is it just me or the volume is quite low?
yeah these videos are great but they need better audio mastering
It sounds good for me.
Just you
On-premises is better than any cloud
Noice
Locos
Mentira
Software sucks.