[DISCUSS] Release Apache Iggy (incubating) v0.6.0

[spetz]

Hello, Apache Iggy (incubating) Community,

This is a call for a discussion to release Apache Iggy (incubating) version 0.6.0.

Along with the Iggy Server, we will also release the SDKs and the remaining tooling (CLIs, Connectors and Web UI) which are part of the repository.

Below is the current list of changes (there are also some outstanding PRs, that we'd like to complete before starting a vote)

server-0.5.0...master

The most noticeable feature is the total rewrite of the server's core to be based on a thread-per-core, shared-nothing design, using io_uring runtime by compio. Moreover, Iggy will now also support WebSocket transport protocol to make it a possible fit for edge devices or web applications.

Kindly provide your feedback on this release plan. Following the conclusion of the discussion, we will update the version in the repository and initiate the release process.

Thanks,
Piotr

[kparisa]

Lot of cool features, can't wait for 0.6 release. Looking forward to seeing the benchmark results before and after io_uring. WebSocket protocol enables a lot of real-time use cases.

[hubcio]

Good plan, so +1 from my side. There will be also plenty of minor releases (direct_io, connection transfer and other non-breaking improvements) - we're on good track.

[numinnex]

+1 from my side, it's been quite a while since our last release and it's a pretty big one.

[ariesdevil]

AFAIK, Rust's poll-based asynchronous design is inherently incompatible with io-completed designs. Currently, all async solutions based on io_uring are problematic. https://tonbo.io/blog/async-rust-is-not-safe-with-io-uring

Tokio adopted a compromise solution.

I understand that this type of issue rarely occurs due to the efforts of the library author, but when it does happen, it can be catastrophic and challenging to reproduce and debug.

[numinnex]

We are aware of those limitations (that rust futures are built for poll io not completion io), but what you said isn't entirely true, monoio (which we previously used) addressed that issue in this pr and compio afair doesn't suffer from this problem at all.

It's worth noting that this is not the end of our journey with exploring io_uring, we want to go beyond what compio currently has to offer, as the project is structured in a way where it's pretty easy to fork and extend.

[ariesdevil]

Make sense, let’s move on.

[lukaszzborek]

+1
this will be amazing release

[T1B0]

+1 let's go to release 🚀

[krishvishal]

+1 from my side.

[mmodzelewski]

+1
Can't wait for the release 🤩

[spetz]

I'm happy to announce that the voting thread has just started #2403