quic: the send-retention limits are QuicOptions, set when building the reactor - #275
Merged
Merged
Conversation
…e reactor SendRetentionBytes (the high-water, default 16 MiB, floored at 256 KiB) and SendRetentionCeilingBytes (the backstop, default 32 MiB - hard-wired at twice the high-water until now) move from the QuicEngine constructor to QuicOptions, and each connection takes them from its reactor. The Playground QUIC samples and the docs set both.
This was referenced Oct 4, 2026
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The two QUIC send-retention limits are now set where the reactor is built, in
QuicOptions, instead of onQuicEngine:SendRetentionBytes(default 16 MiB, floored at 256 KiB as before): unacknowledged response bytes a connection holds beforeCanQueueSendturns false and a streamed response waits for acks.SendRetentionCeilingBytes(default 32 MiB): past this the connection is closed - the backstop for a producer that ignores the wait. It was hard-wired at twice the high-water; it is its own knob now, never belowSendRetentionBytes.Each connection takes both from its reactor when it is accepted or opened, the way
ReadTimeoutMsalready reaches it; a client-only reactor (noQuicOptions) gets the defaults.Breaking:
QuicEngine'smaxSendRetentionBytesparameter is gone - a caller passing it moves the value toQuicOptions.SendRetentionBytes.The 13 QUIC Playground samples set both in their
QuicOptions;docs/how-ioxide-does-h3.htmland the generated panes follow.Why it came up: a buffered HTTP/3 response is handed over whole, so 32 concurrent 1 MiB responses on one connection cross the 32 MiB ceiling and the server closes every such connection (16 of 16 under h2load
-m 32; with one stream per connection the same responses work). The ceiling can be raised now; making the buffered path respectCanQueueSendis the real fix, and a separate change.Tests
quic: a connection pushes back at QuicOptions.SendRetentionBytes(new): with a 256 KiB high-water, 300 KiB queued turnsCanQueueSendfalse and the connection stays up.quic: the send-retention backstop tells the peer before it drops the connection: now sets its 512 KiB ceiling throughQuicOptions.TestServer.StartDatagramtakes an optionalquicOptionshook.All suites pass: E2E 232, Unit 60, Chaos 47, Http 44, Tls 151 (the 7 kTLS tests skip without sudo), File 4.
Bench
h2load (
--alpn-list=h3, 16 conns x 32 streams), interleaved 5 rounds with a second build of main as the control; median of the per-round ratio against main: