passwd-as-service

crate source docs

Passwd-as-service is a webservice that provides a query API into the /etc/passwd and /etc/group files. It's not meant to be a 'real' service; it's a didactic coding sample meant for potential employers, based off a coding challenge from Brain Corp. See the file api.pdf for details on the exact challenge.

• Documentation
• Installation
• Configuration
• Testing
• Design Decisions

Documentation

Clone the source and run cargo doc --open, or visit the docs on docs.rs

Installation

• Install Rust using rustup.

• Install the nightly toolchain

  rustup install nightly
  

• Install via crates.io

  cargo +nightly install password-as-service
  

  Alternatively, prebuilt binaries for 64-bit OSX, Windows, and Linux Ubuntu/Debianare available in [./bin].

Configuration

[ROCKET_PORT=PORT] passwd-as-service [PATH_TO_USERS] [PATH_TO_GROUPS]

• ROCKET_PORT is the port to serve from. This can also be set as an environment variable.
• PATH_TO_USERS is the location of the users file: defaults to "/etc/passwd".
• PATH_TO_GROUPS is the location of the users file: defaults to "/etc/group" You must use both of PATH_TO_USERS and PATH_TO_GROUPS, or neither.

Testing

• Clone the repository

  cd ~/rust
  git clone gitlab.com/efronlicht/passwd-as-service
  

• Navigate to the project directory

  cd ~/rust/passwd-ass-service
  

• Run Tests

  Cargo

  cargo test
  

Test Coverage (Ubuntu/Debian)

• Install Tarpaulin:

  # install the lib-ssl development libraries
  apt-get install libssl-dev pkg-config cmake zlib1g-dev
  # set cargo to use rust's `nightly` toolchain
  rustup default nightly
  # OR:
  rustup override set nightly
  # install tarpaulin
  RUSTFLAGS="--cfg procmacro2_semver_exempt" cargo install cargo-tarpaulin
  

• Get Coverage

  $ cargo tarpaulin
  
    [INFO tarpaulin] Running Tarpaulin
    [INFO tarpaulin] Building project
    [INFO tarpaulin] Launching test
    [INFO tarpaulin] running /home/efron/rust/passwd-as-service/target/debug/deps/bin-f54ce995ec16a011
  
    running 0 tests
  
    test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
  
    [INFO tarpaulin] Launching test
    [INFO tarpaulin] running /home/efron/rust/passwd-as-service/target/debug/deps/lib-1087fab0c4512719
  
    running 6 tests
    ......
    test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
  
    [INFO tarpaulin] Coverage Results:
    || Tested/Total Lines:
    || src/api/mod.rs: 14/14
    || src/api/tests.rs: 84/86
    || src/lib.rs: 6/20
    || src/main.rs: 0/3
    || src/model/group.rs: 11/18
    || src/model/user.rs: 24/26
    ||
    83.23% coverage, 139/167 lines covered
  

Design Decisions

Language: Rust

• Positives
  • Extremely strong static typing and guarantees about concurrency. No race conditions.
  • No garbage collection and powerful optimizing compiler make it run blazing fast.
  • Easy to compile to a portable binary and deploy in containers.
  • Best-in-class good documentation and code quality tools
  • Deployment is easy, and low usage of resources makes running on rented servers cheap.
  • Cargo is an extremely good package and dependency manger.
  • My personal favorite programming language.
• Negatives
  • High learning curve.
  • Difficult to write.
  • Immature ecosystem for web development comparied to competitors like JavaScript, Go, & Python.
  • Strictness not always necessary for 'quick and dirty' jobs.
  • Slow compilation times.

Database: None, w/ a level of in-memory caching.

This is a toy project, dealing with very small amounts of data. Even the largest linux system is unlikely to have more than ten thousand users or groups; we simply don't need a relational database to query this kind of data. We have a single source of truth (the flat files /etc/passwd/ and /etc/group themselves.

• Positives

  • Removes a lot of overhead and disk usage.
  • Cheaper containerization: Most relational databases clock out at over 200mb and can use considerable memory that's a lot of resources for something we simply don't need.
  • Much simpler deployment; we can just distribute a binary, with zero dependencies.

• Negatives:

  • Relational databasses like PostgreSQL and MYSQL are pretty much made for this kind of query.
  • If the project expands we'll probably need to use a database eventually.
  • Reinventing the wheel.

Framework: Rocket.

Rocket is an extremely exciting framework for writing servers in Rust.

Positives

• Very little boilerplate.
• Strong typing & powerful custom macros allow you to create readable, testable APIs.
• Very fast.

Negatives

• Immature.
• Requires nightly rust.
• Changes to Rust could leave the Rocket stranded.

Deployment: Statically linked binaries

• Fast
• Portable
• Much smaller than wrapping an entire interpreter.
• Can't inject DLLs (because there are no dynamic libararies)

• Must be recompiled every time, unlike interpreted languages.
• Larger than using dynamic links.

Continuous Integration & Source Control: Git w/. Gitlab

Upsides:

• Simpler than most
• Git is the most common source control tool by a country mile
• Gitlab's continuous integration is much better for small or medium sized projects than the heavyweight solutions used for most github projects.

Negatives:

• Github is much more discoverable.