websockets initial messages delayed multiple seconds

[rukai]

• I have looked for existing issues (including closed) about this

Bug Report

Version

axum 0.6 and latest git

Platform

Linux memes 6.2.12-arch1-1 #1 SMP PREEMPT_DYNAMIC Thu, 20 Apr 2023 16:11:55 +0000 x86_64 GNU/Linux

Crates

axum

Description

Please refer to this repo that reproduces the issues https://github.com/rukai/axum/tree/websockets_bug_report
Run
axum/examples/websockets/src> cargo run --bin example-websockets
and
axum/examples/websockets/src> cargo run --bin example-client

I create a websocket handler on the server which responds with many messages for each message that the client sends it.

The server sends each message 1 second apart, but the client receives the first 3 messages all at once after 3 seconds. From that point things proceed reasonably
Instead of this behavior I expect the initial messages to arrive immediately after being sent.

In my application I had these much more severe scenarios occur:

• no messages arriving until the server closes the connection after about 30s
• no messages arriving until the client has sent 14 messages over about 14s

but I couldnt recreate these severe cases in a simple scenario.
My application is running over a network which may be making the issue worse than the simple localhost test in the reproduced example.

Possibly this is a tungstenite issue, but figured I'd start by reporting it here.

[davidpdrsn]

Either I don't understand what you mean or I cannot reproduce it.

This is what I see

2023-04-25_15-41-39.mov

[rukai]

Looks like you have main checked out instead of the branch I linked.
These are the changes to the example: f1cbd6b

[davidpdrsn]

Ah I see. I just git cloned your fork, but I see it now. I tried setting no_delay on the tcp stream but that didn't change it.

It's most likely a thing with tungstenite though. Axum's web sockets are a very thin wrapper around tungstenite.

I'll close this for now but feel free to ask follow up questions.

[rukai]

Thankyou for at least confirming you can reproduce it, I've now reproduced it without axum involved and will take it up with the tungstenite project.