Repository navigation
http-netty: reset cancelled HTTP/2 streams whose request body is still writing - #3678
Draft
bryce-anderson wants to merge 1 commit into
Draft
bryce-anderson wants to merge 1 commit into
bryce-anderson wants to merge 1 commit into
Conversation
Contributor
Author
|
This should be sequenced after #3676 since technically this can expand the number of cases where the ST stream counter is less than the netty stream counter. |
bryce-anderson
force-pushed
the
h2-cancel-reset-5e766195
branch
2 times, most recently
from
October 6, 2026 17:43
b329c8e to
60b31c9
Compare
…l writing #### Motivation An HTTP/2 client request cancelled after its response finished but before its request body did was never reset. A gRPC client does this without any user cancel when a server rejects a client-streaming or bidirectional call with a Trailers-Only response while the request stream is still open. The server kept waiting for the rest of the body, and the stream held a concurrent-stream slot on both peers until the server reset it or the connection closed. Enough of these made the client open extra connections. The client's close-on-cancel skipped HTTP/2 because it once ran on the parent connection, where closing failed every stream (apple#1307), but it now runs on each stream. #### Modifications - The client's close-on-cancel now applies to HTTP/2 too, where it closes only the cancelled stream. #### Result Cancelling an HTTP/2 request whose request body is still streaming resets its stream. Behavior change: an HTTP/2 server now receives RST_STREAM(CANCEL) for these streams, including after it has sent its full response. A server that counts such resets against a rapid-reset limit, as the ServiceTalk server does, can reach that limit sooner, for example when it rejects many streaming gRPC calls on one connection. If that happens, raise the server's limit on received resets.
bryce-anderson
force-pushed
the
h2-cancel-reset-5e766195
branch
from
October 6, 2026 18:16
60b31c9 to
6289389
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
An HTTP/2 client request cancelled after its response finished but before its request body did was never reset. The server kept waiting for the rest of the body, and the stream held a concurrent-stream slot on both peers until the server reset it or the connection closed. Enough of these made the client open extra connections. The client's close-on-cancel skipped HTTP/2 because it once ran on the parent connection, where closing failed every stream (#1307), but it now runs on each stream.
Modifications
Result
A cancelled HTTP/2 request no longer leaves its stream open.