> Deployment with Rust backend programs in general can be less than ideal due to having to use Dockerfiles,
Actually, in my experience, Rust is one of the best languages for ease of deployment (for much the same reason as Go). Rust/Cargo produces self-contained statically linked binaries. Rust/Cargo also has a real nice cross-compiling story. Often my deployment will be to build the binary and then basically scp the binary and supporting files (such as html/config) to the target. You don't really need Docker.
As a seasoned python developer deployment of a pure Rust program is much simpler than Python. The only issue that I ever had was figuring out how to cross compile for a raspberry pi, but even that took me less time than the typical python dependency problem that may happen when you e.g. devlope on Ubuntu and deploy on Debian.
When Rust deployment gets hairy, it is typically the fault of non-Rust-dependencies.
Is it safe to expose to the internet[0]? I’m still looking for a Rust web framework that supports eg timeouts for idle connections so that it doesn’t crash and burn when it runs out of sockets after a few hours of serving production traffic (let alone being able to survive an intentional slowloris attack) :(
This isn't trolling, it's my genuine ignorance- I thought Actix was the goto web framework for Rust? I don't do a whole lot of Rust and when I do it isn't http router type work so I'm out of the loop here.
Looking at stats on crates.io for a few frameworks
- axum: 100k installs/day
- warp: 25k installs/day
- actix: 15k installs/day
- rocket: 7k installs/day
- tide: 1k installs/day
(For comparison, hyper - the HTTP library that most web frameworks use under the hood - is at 200k/day, many of which are using it for client rather than server functions. So unless there is another not-hyper-based framework out there, then I’d be moderately confident guessing that axum has >50% of the server market)
Actix used to be the fastest (maybe still is? all the frameworks are so fast it doesn't even matter that much anymore) and most popular framework.
Then a bit of drama happened about the hard life of maintaining an OSS project with users criticising you (the usual).
Rocket seemed to gain popularity for a bit but then there was some drama with the maintainer disappearing / not releasing 0.5 for a long time.
Meanwhile, quietly, the guys who did tokio (the most popular async runtime) released axum, which is compatible with tower middlewares. So they had tons of features pretty early.
I switched from actix to axum around 0.4 and updates have been very sensible and kept improving the DX.
One year ago it was still the standard and I think most production workload still are using actix (we are). When you dig into actix you feel the pre-async decisions (acceptor threads, worker threads, futures not send).
But for a lot of usecases this model is faster than Axum work stealing. I am not a super huge fan of hyper (on which Axum is based), I salute the work (it is impressive) but it is very clunky to use, the code is hard to read and usage is super opiniated.
I think Actix used to be the goto web framework at the time is because it was the most complete, it's still a solid choice. What I think made Axum more popular aside from the features added is that it's part of the tokio-rs projects.
I used to use Actix for a web project, but the Websocket story was severely lacking. Actix is non-work-stealing multi-threaded (to achieve "performance", I presume) and thus it has to use a non-async Actor model for inter-Websocket interactions. This meant that I could not use an async DB pool or an async HTTP requests library within my Websocket code.
Alternatively, Axum is built by the same people that made tokio, is work-stealing multi-threaded, and the Websocket code is async be default. It is just overall a better library.
When I looked into it while starting a new project about a year ago, Axum seemed like the framework with the most inertia. I think there was some concern a year or two back about whether Actix would continue to be maintained, but it looks healthy now.
They're probably both fine choices, I chose Axum because the middleware architecture was easier to grok and at the time I felt that staying under the tokio umbrella would be safer long term.
I've been using Tide with great success lately. Also - for my own projects I haven't used docker lately and just dump the binary with it's configs to the debian box with a systemd configured and so far it's been painless, ymmv.
Tide is fantastic and has IMO the best ergonomics (and could be improved to be even better), except it's essentially abandonware at this point. I'd love to see it continue and thrive, but there's many PR's that have been ready to land for a long time gathering dust, and plenty of open issues that are at a standstill because there's simply no momentum or direction.
Not blaming anyone - the maintainers don't owe us anything - it just wouldn't be my crate of choice if I was starting today. If any of the maintainers read this, shoot me a message because I'd love to help out and get the ball rolling again on tide.
jbandela1|2 years ago
Actually, in my experience, Rust is one of the best languages for ease of deployment (for much the same reason as Go). Rust/Cargo produces self-contained statically linked binaries. Rust/Cargo also has a real nice cross-compiling story. Often my deployment will be to build the binary and then basically scp the binary and supporting files (such as html/config) to the target. You don't really need Docker.
logankeenan|2 years ago
I wrote a quick blog post in the past describing what the parent comment is talking about.
https://logankeenan.com/posts/deploy-your-rust-project-to-an...
atoav|2 years ago
When Rust deployment gets hairy, it is typically the fault of non-Rust-dependencies.
NewJazz|2 years ago
Shish2k|2 years ago
[0] I’m ideally hoping for a framework which ticks the same boxes that Go’s `net/http` was ticking in 2016: https://blog.cloudflare.com/exposing-go-on-the-internet/
Klonoar|2 years ago
Or at least I can say - I’ve not experienced issues doing this in an app serving 100m in a day. YMMV etc.
Wonnk13|2 years ago
Shish2k|2 years ago
- axum: 100k installs/day
- warp: 25k installs/day
- actix: 15k installs/day
- rocket: 7k installs/day
- tide: 1k installs/day
(For comparison, hyper - the HTTP library that most web frameworks use under the hood - is at 200k/day, many of which are using it for client rather than server functions. So unless there is another not-hyper-based framework out there, then I’d be moderately confident guessing that axum has >50% of the server market)
jokethrowaway|2 years ago
Rocket seemed to gain popularity for a bit but then there was some drama with the maintainer disappearing / not releasing 0.5 for a long time.
Meanwhile, quietly, the guys who did tokio (the most popular async runtime) released axum, which is compatible with tower middlewares. So they had tons of features pretty early.
I switched from actix to axum around 0.4 and updates have been very sensible and kept improving the DX.
Nowadays Axum is at the top for downloads number: https://www.arewewebyet.org/topics/frameworks/
Sytten|2 years ago
But for a lot of usecases this model is faster than Axum work stealing. I am not a super huge fan of hyper (on which Axum is based), I salute the work (it is impressive) but it is very clunky to use, the code is hard to read and usage is super opiniated.
quincepie|2 years ago
goddtriffin|2 years ago
Alternatively, Axum is built by the same people that made tokio, is work-stealing multi-threaded, and the Websocket code is async be default. It is just overall a better library.
sampullman|2 years ago
They're probably both fine choices, I chose Axum because the middleware architecture was easier to grok and at the time I felt that staying under the tokio umbrella would be safer long term.
spacedimp|2 years ago
https://spacedimp.com/blog/using-rust-axum-postgresql-and-to...
boredumb|2 years ago
mdtusz|2 years ago
Not blaming anyone - the maintainers don't owe us anything - it just wouldn't be my crate of choice if I was starting today. If any of the maintainers read this, shoot me a message because I'd love to help out and get the ball rolling again on tide.