Large, streamed request body may result in noSignalReceived preconditionFailure crash
daveanderson opened this issue · 6 comments
Describe the bug
When streaming a large (20+ MB) request body to a controller that writes it to a file on disk the Request.Body.AsyncSequenceDelegate method produceMore0() encounters a noSignalReceived case and the server exits with a preconditionFailure().
Having the server spontaneously exit in production is highly undesirable.
It's not completely clear why noSignalReceived is occurring, but experimentally replacing the preconditionFailure() with break appears to allow the server to wait for additional data to arrive from the client, allowing the remainder of the body to be received and stored.
Please consider adjusting the implementation of the Request.Body.AsyncSequenceDelegate method produceMore0() such that the server will not spontaneously terminate when streaming a large request body.
To Reproduce
Steps to reproduce the behavior:
- A route with configured with
body: .stream - A controller method that handles the 20+ MB streamed body using the
for try await byteBuffer in body {}pattern. - As the streamed bytes are received at some point the
Request.Body.AsyncSequenceDelegatemethodproduceMore0()encounters anoSignalReceivedcase and the server exits with apreconditionFailure().
Environment
- Vapor Framework version: 4.92.2
Please test with v4.92.3 https://github.com/vapor/vapor/releases/tag/4.92.3
Hmm if it's hitting https://github.com/vapor/vapor/blob/main/Sources/Vapor/Concurrency/RequestBody%2BConcurrency.swift#L34 then I don't think the latest release will help. It could be the same bug as #2985
(Wrong button!)
Vapor v4.92.3 does not resolve the produceMore0() preconditionFailure().
Also, this does appear to be a the same issue as #2985