Just for some news feed like contents - it might be troublesome to launch an instance of the single purpose.
Looking into Google spreadsheet as API endpoint.
We started with API Gateway + Lambda, but recently migrated to Google Cloud Functions. The overhead to spin up new funcs w/ endpoints is trivial, as is rolling in CORS and even some auth if you really want (we're using Firebase Auth). It was a big deal paradigm shift for us. Whether that's because it's a fun product or because I'm just awful at AWS is, well...
I don't really find Pythonanywhere useful for anything other than hobby sites. I mean no gracefull restart, no caching on static assets and slow SSL encryption (not cached on the web server).
PythonAnywhere dev here: fair point re: the caching of static assets, but could you give a bit more information about the SSL and the graceful restart? All of that should work fine.
I also run on DO but haven't deploy with docker. What are the greatest benefits in your opinion? (I only run little experiences for now, but would also like to run small client backends).
I have run DO + Docker and the reason I do it is because it makes deployments dead simple. Through up Ngnix as a reverse proxy to the port that is exposed through docker and you're done. I'm deploying an Elixir/Phoenix app, so it makes dependency management dead simple too with not having to worry about versions of Elixir/Phoenix and it makes changing droplets a breeze too, since the only pre-reqs to install are docker.
The biggest benefit for me is isolation. I keep the main system clean. Just install fish shell and glances there. And all messy stuff required by projects is isolated in individual containers. They can be rebuilt from scratch if needed and don't pollute the main system (except for well-defined data folders).
Ditto. Dedicated hardware at packet.net, OVH, ServerStadium, HostVirtual. Recommend WebHostingTalk.com forums offers sections for good deals on dedicated servers, VPS.
Using docker / dokku
That's a lot less simple as now you suddenly have multiple machines to handle (even if they're virtual). Whenever I really want something like that, I'll usually use systemd-nspawn or docker depending on the use case.
Kubernetes on Azure and GKE for main projects (high traffic)
If your projects are lightweight / low traffic, check out Flynn. You get to push code directly to flynn and skip the docker workflow (push commit to github -> docker builds image -> kubernetes updates pods, this takes time ofcourse). Flynn can also create DB for you.
IBM Bluemix used to have a free tier for containers, which you could use to run tiny apps quite easily and cheaply, using the native Docker CLI. They've since moved to Kubernetes and still have a free tier, but not sure how that works.
I'd probably recommend a DigitalOcean VPS though. Cheap, very stable (had a couple VPS for a year with 100% uptime, as far as I know), stellar support and a lot of new functionality being added lately - block storage, firewall, load balancing. DO is one of the best services I've paid for.
Disclosure: I used to work at IBM about a year ago, in the container service team.
We use a combination of Lambda, ECS (for slightly larger services), and Heroku to host them. And, we use Mashape's Kong in front of them for authentication / authorization, and logging (see https://blog.arachnys.com/how-we-solved-load-testing-scalabi...). We found it far easier to configure than API Gateway.
Formerly Heroku, but now only for toy projects: their approach to pen testing and the insecurity of their Postgres setup count heavily against them, and they're expensive (I keep meaning to write a blog on these points ...)
I've been impressed with Google Apps Script (https://script.google.com) for little things. Define a doGet (or doPost) function, and it will be called for HTTP requests.
63 comments
[ 2.5 ms ] story [ 123 ms ] thread[1] https://zeit.co/now#features
Most big cloud vendors offer API management through their services.
If your projects are lightweight / low traffic, check out Flynn. You get to push code directly to flynn and skip the docker workflow (push commit to github -> docker builds image -> kubernetes updates pods, this takes time ofcourse). Flynn can also create DB for you.
Or you can just use firebase.
For static content that can be (re)published as needed, I'd use S3 + CloudFront and a scheduled or triggered Lambda that handles the publishing.
I'd probably recommend a DigitalOcean VPS though. Cheap, very stable (had a couple VPS for a year with 100% uptime, as far as I know), stellar support and a lot of new functionality being added lately - block storage, firewall, load balancing. DO is one of the best services I've paid for.
Disclosure: I used to work at IBM about a year ago, in the container service team.
https://stdlib.com/
I've seen the founder setup some impressing services in a matter of minutes.
I love their docker-like CLI.
Formerly Heroku, but now only for toy projects: their approach to pen testing and the insecurity of their Postgres setup count heavily against them, and they're expensive (I keep meaning to write a blog on these points ...)
https://developers.google.com/apps-script/guides/content has a tiny example where they fetch an RSS feed, transform it, and serve the result.
You also have very easy access to Google services, so you can read from and write to Google spreadsheets, Drive, Gmail, etc.