From fc50b66be91033d2240cf22d2372bb00019fbe3d Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 18 Nov 2026 12:04:43 -0800 Subject: [PATCH 1/2] chore(quest): claim quest/m1/precut-fixes/relay-backoff-backport.md Quest-Claim: aa4df95b0798fe5d From 7263e15d368560c7233b9ca06f60cf852dce37a6 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 18 Nov 2026 12:47:46 -0800 Subject: [PATCH 2/2] quest: finish relay-backoff-backport, landed on release in #9104 Co-Authored-By: Claude Opus 5.5 --- .../m1/precut-fixes/relay-backoff-backport.md | 27 ------------------- 1 file changed, 27 deletions(-) delete mode 100644 quest/m1/precut-fixes/relay-backoff-backport.md diff --git a/quest/m1/precut-fixes/relay-backoff-backport.md b/quest/m1/precut-fixes/relay-backoff-backport.md deleted file mode 100644 index 53dde7f40e..0000000000 --- a/quest/m1/precut-fixes/relay-backoff-backport.md +++ /dev/null @@ -1,27 +0,0 @@ -# [XS] Backport the relay redial backoff reset to the next release - -## Goal - -A relay on the next `release` patch redials a cluster peer from the first -backoff ceiling once an attempt connected, so a flap after a long outage no -longer waits out 30 s. - -## Plan - -Decided by the maintainer when the PR that finished `relay-config-loud` was -picked: ✅ backport only the redial backoff reset to `release`, and leave the -two startup refusals on `main` (rejected: backporting all of it, which would -let a patch stop a working relay from booting on a config that ran before; no -backport, which leaves a flapping peer waiting out 30 s on `release`). - -Port the backoff reset from that PR as a separate PR onto `release`, per -RELEASE.md, keeping the published API. -`rs/moq-relay/src/cluster.rs` `run_remote` on `release` has the same dead -`Ok(()) if elapsed >= stable_threshold` branch, because `run_remote_once` only -returns an error. Report whether the attempt connected (any observation with -`connects > 0`) and pass it to `Redial::next`, which starts over at the first -ceiling when it did, including on the immediate retry after the cluster token -rotated. - -Do not port `cluster.api.root` without `cluster.api.url` or `cluster.token` -with no peer to dial as startup errors: those stay on `main`.