311–320 of 357 posts
I'd be interested to know if this is integrated and rebranded nanobox [0]? Or something completely from scratch? [0]: https://nanobox.io/
Earlier quoted context omitted.
Can HN really deliver enough traffic to a static site to cost a significant amount? I've had mildly popular posts on HN for my Netlify blog (John Carmack tweeted about it!) and not had to pay for bandwidth.
No. I don't think so. The concern for me is a lack of hard limit on spending on GCP, Azure, and AWS. If I screw up and allocate a bunch of resources unintentionally, I'm left holding the bill. That's a terrible setup for PaaS because all programming involves mistakes eventually, especially for new users learning the system. Granted, there are likely limits on accounts, but those are to protect the services from fraud…
I run a small side business, and these unlimited cloud plans are just a no go. A medium to large company could totally absorb a 5 figures bill, but that would be a dead sentence to my side project. Also, considering the variable costs of bandwidth of AWS, Azure or Cloudflare, one competitor could simple rent an OVH server and incur insane costs to my business while only spending 1/10 of the money.
Right now, I'm using Heroku (with a limited number of dynos and a single PgSQL database) together with BunnyCDN (which allow me to pay for prepaid usage). If I ever get DDoS'ed, my app will most probably be inaccessible or at least significantly slower, while I'll receive an email alert, from which I can decide myself to allocate more resources.
Earlier quoted context omitted.
I’m waiting for digital ocean to have something like google cloud run. Google cloud run is essentially here is a docker image that listens on the $PORT env variable. Spin it up when you get requests. It will handle X queries per second (you can set limit). If more than X, scale it up to this many replicas. I pay about 10 cents for my site. Zero maintenance. I push code to GitHub, GitHub builds an image, pushes to GCR…
What sort of cold start time do you get with that out of interest?
Earlier quoted context omitted.
What sort of cold start time do you get with that out of interest?
Not sure about exact numbers on the cold start but we avoid it altogether with a keep-warm request every minute.
I tried this out today and figured out how to run a custom Python application there. You only need two files in a GitHub repo for that: a requirements.txt listing your dependencies and a Procfile that starts a process listening on port $PORT and IP 0.0.0.0. My Procfile looks like this: web: datasette . -h 0.0.0.0 -p $PORT Full notes here: https://til.simonwillison.net/til/til/digitalocean_datasette...
This is built on top of Heroku runtime, more info on Procfile here: https://devcenter.heroku.com/articles/procfile Also, note that you shouldn't use SQLite, at least with heroku, because apps should be stateless https://12factor.net/processes
First, you don't actually know that using SQLite would make the OP's app stateful—if datasette is set up in immutable mode, then the .db files are no more indicative of a stateful process than a static CSV or a JSON or YAML config.
Second, not every app needs to be a 12 factor app, and you're not in a position to understand the trade-offs OP is dealing with. "Best practices" rarely are best in every circumstance, and often conflict.
Earlier quoted context omitted.
I'm not one of your customers but when I have looked at pay-as-you-go services or autoscaling services before I basically don't even consider any that don't allow me to cap the costs per month or similar. So you could perhaps also consider that if it is only marketed when you sign up and not a clearly defined feature some people (like me) will just never sign up at all.
That's actually why we made a big deal out of it in our launch post. We thought maybe it would attract underserved customers. I am surprised that it had no effect. My guess is that people who are interested in controlling costs are also not getting enough value from auto scaling services. So we're already not attracting those folks, and offering a cost control feature isn't enough to get us over the hump.
I tried this out today and figured out how to run a custom Python application there. You only need two files in a GitHub repo for that: a requirements.txt listing your dependencies and a Procfile that starts a process listening on port $PORT and IP 0.0.0.0. My Procfile looks like this: web: datasette . -h 0.0.0.0 -p $PORT Full notes here: https://til.simonwillison.net/til/til/digitalocean_datasette...
This is built on top of Heroku runtime, more info on Procfile here: https://devcenter.heroku.com/articles/procfile Also, note that you shouldn't use SQLite, at least with heroku, because apps should be stateless https://12factor.net/processes
Earlier quoted context omitted.
Fly.io is more exciting for me because it's edge app servers and it terminates web sockets. I've been envisioning a live-view like framework, or heck even a regular old rest api, where the client connects to the fly.io server with a single request then the app server makes a request to N number of back end processing servers using an efficient protocol like grpc.
It is cool, but it does not scale to zero, so if you have N services you will end up paying for at least N instances.
Earlier quoted context omitted.
Wait, I'm confused. What standing charge? What exactly did you get charged for? I've been using DO for a few years and I have no idea what you are referring to. When I delete my unused resources, I don't get charged.
He probably forgot to turn off the instances and kept getting charged.
I am so glad to see this. I was looking to deploy an app and the choice is either Heroku or manage your own server which I don't want to do. Heroku gives instant deployment for the most common types of apps (python/java/ruby). It's PaaS done right, it's fantastic. You should really have a look if you're not aware of it, it's only $7 for a starter app. Problem is, scaling up is about $50 per gigabyte of memory which m…