Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Why did I build KDash in Rust

Why did I build KDash in Rust

Why did I build a Kubernetes Dashboard in Rust? I’ll explain the reason behind it and show what KDash does and how it differs from existing tools. I’ll also be deep-diving into the challenges I faced and the reason I choose to do it in Rust.

Deepu K Sasidharan

October 04, 2022
Tweet

More Decks by Deepu K Sasidharan

Other Decks in Programming

Transcript

  1. @deepu105 @oktaDev Why did I build a Kubernetes Dashboard in

    Rust? Deepu K Sasidharan @deepu105 | deepu.tech
  2. @deepu105 @oktaDev Hi, I’m Deepu K Sasidharan ➔ JHipster co-lead

    developer ➔ Java Champion ➔ Creator of KDash, JDL Studio ➔ Developer Advocate @ Auth0 by Okta ➔ Polyglot dev, OSS aficionado, author, speaker @deepu105 deepu.tech deepu105 deepu05
  3. @deepu105 @oktaDev KDash • A simple terminal dashboard for Kubernetes

    • Provides resource views, Node metrics, logs and common view features • Resources utilizations for nodes, pods and namespaces based on metrics server. • Context switching • Easy full keyboard navigation and themes • Open Source, MIT license https://github.com/kdash-rs/kdash
  4. @deepu105 @oktaDev Why another dashboard Kdash is obviously inspired by

    k9s, so why build another dashboard? • User experience: I wanted the stuff that is most used easily available and a workflow suitable for daily use • Better navigation: I wanted a better overview and keyboard shortcuts for everything • Better metrics: I wanted to see better resource metrics for the cluster • PoC for Rust tooling: I wanted to build something useful in Rust for the k8s space • Speed: While Go based tooling can be almost as fast, I wanted to build something faster and efficient with Rust. Why not? ◦ KDash uses way less memory/cpu than k9s and even Kubectl watch
  5. @deepu105 @oktaDev Rust = High level general purpose language •

    Multi-paradigm, ideal for functional, imperative and even OOP • Modern tooling • Ideal for systems programming, embedded, web servers and more • Memory safe • Concurrent • No garbage collection • Performance focused • Awesome community • Most loved language in Stack Overflow survey for 6 years in a row
  6. @deepu105 @oktaDev “Rust throws around some buzz words in its

    docs, but they are not just marketing buzz, they actually mean it with full sincerity and they actually matter a lot”
  7. @deepu105 @oktaDev Ownership and borrowing • No garbage collection or

    any runtime memory management • Memory is managed using lifetimes of variables using a borrow checker at compile time • No pause times, no runtime overhead • Efficient and very low memory usage • Reference counting available when needed
  8. @deepu105 @oktaDev Safety guarantee • Memory safe • Null safe

    • Type safe • Thread safe Rust is safe by default and you can write unsafe code only within unsafe code blocks
  9. @deepu105 @oktaDev Memory safety “About 70% of all CVEs at

    Microsoft are memory safety issues. Two-thirds of Linux kernel vulnerabilities come from memory safety issues” • No free after use errors • No invalid pointer access • No double free errors • No undefined behaviours • Memory safety is ensured by ownership mechanism • Unsafe mode for manual memory management and memory unsafe code https://deepu.tech/memory-management-in-programming/
  10. @deepu105 @oktaDev Null safety “NULL is the most dreaded feature

    and the worst invention in programming” • No concept of NULL • Optional monad is used for presence and absence of data
  11. @deepu105 @oktaDev Thread safety “Fearless concurrency” • Threads, coroutines and

    asynchronous concurrency • Mutex and ARC for shared data concurrency • Channels for message passing concurrency • Data race is not possible in Rust • No need for thread synchronization • Memory and Type safety ensures thread safety
  12. @deepu105 @oktaDev Zero cost abstractions “What you don’t use, you

    don’t pay for. And further: What you do use, you couldn’t hand code any better.” – Bjarne Stroustrup • Your programming style or paradigm does not affect performance • Number of abstractions does not affect performance as the compiler always translates your code to the best possible machine code • You could not get more performance from hand written optimizations • In-built abstractions are often more performant than hand written code • Compiler produces identical assembly code for almost all variations
  13. @deepu105 @oktaDev Immutable by default • Variable are immutable by

    default, including references • Mutations needs to explicitly declared at all stages using the mut keyword, like var declaration and method signatures • Variables can be passed by value or reference, mutable or immutable
  14. @deepu105 @oktaDev Pattern matching • First class support • Can

    be used for control flow in if, switch, while, for statements • Can be used for error handling, optionals and so on • Can be used for value assignments and for code blocks
  15. @deepu105 @oktaDev Advanced generics, traits and types • Advanced generics

    ◦ Generics in types, structs, enums and functions ◦ No performance impact due to zero cost abstractions ◦ Generics with lifetime annotations • Traits for shared behaviour ◦ Default implementation for traits ◦ Placeholders for traits, operator overloading ◦ Trait bounds for Generics ◦ Multiple and compound trait bounds • Type aliasing and great type inference
  16. @deepu105 @oktaDev Macros • Meta programming • Great for non

    generic reusable code • Custom behaviours • Declarative macros and Procedural macros
  17. @deepu105 @oktaDev Tooling and community • Hands down, one of

    the best compilers out there and great backward compatibility • One of the best tooling you can find in terms of features and developer experience. Cargo is one stop shop for Rust tooling, build, compilation, formatting, linting, and so on • One of the best documentation, which is shipped with the tooling • A very diverse and vibrant community ◦ Community formed from other languages hence bringing in best of many ◦ Very welcoming and helpful ◦ Rapidly maturing ecosystem with growing number of libraries and use cases ◦ Has a forum which is used more than stack overflow for Rust • Big names like Google, Apple, Microsoft, Amazon and Facebook are already behind rust and investing it. • It’s on path to become the second supported language in Linux development. • Use case has already extended to embedded, web assembly, kubernetes, web development, game development and even client side • It’s only a matter of time until you can do any use case in Rust
  18. @deepu105 @oktaDev The downsides • Complexity • Steep learning curve

    • Young and maturing • Many ways to do the same thing (kind of like JS)
  19. @deepu105 @oktaDev High level language compromise • Safety • Speed

    • Abstractions With Rust we can get all three. Hence Rust is a high level language with performance and memory efficiency closest to a low level language. The only tradeoff you will make with Rust is the learning curve.
  20. @deepu105 @oktaDev Major Libraries used • tui-rs: Library to build

    rich terminal user interfaces and dashboards • Crossterm: Terminal manipulation library • Tokio: An event-driven, non-blocking async I/O platform • kube-rs (k8s-openapi, Openssl): Kubernetes client for Rust • kubectl-view-allocations: Library to fetch resource allocations for k8s • clap: Command line argument parser • rust-clipboard: cross-platform library for OS-level clipboard management • duct.rs: Library for running child processes • Serde: Serialization and deserialization library
  21. @deepu105 @oktaDev Data architecture • Event driven ◦ Async I/O

    using dedicated channels • Multi threaded ◦ Single data-store using ARC + Mutex ◦ Separate threads for UI, streams, network and CMD • Non-blocking UI ◦ Only bottleneck is network I/O with kube-api-server
  22. @deepu105 @oktaDev Challenges • Log streaming was blocking due to

    how the kube-rs lib works ◦ Solution: Seperate thread for log streaming with tokio-streams ◦ Some quirks exists and needs to be investigated and fixed • UI was blocking and not very smooth ◦ Solution: Seperate thread for network calls and event driven I/O • Kube-rs version updates are coupled to another library used ◦ No proper solution yet
  23. @deepu105 @oktaDev Thank You Deepu K Sasidharan @deepu105 | deepu.tech

    https://deepu.tech/tags#rust https://developer.auth0.com