From b6a2296c71d6b12b3ffe421eb3a4d0462945b50a Mon Sep 17 00:00:00 2001 From: jochen Date: Sat, 21 Nov 2026 18:14:40 +0100 Subject: [PATCH] Design 47 H: H-A-289, the end's write waits on no other part --- 03-DESIGN/01-to-be/47-milestones/H-lead-answers.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md index 9b54664f..01deb190 100644 --- a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md @@ -1661,3 +1661,10 @@ Where two preparers disagreed, the chosen one is named. pipeline found in that state. The end's outcome is superseded when every mover left, failed when a step ended could-not-run or failed, passed otherwise; the journal says which. In the runner. ADR 0682, H-A-223, H-A-243, P3 (no slot held for work nobody waits on). +- **H-A-289 (the end's write waits on no other part, 18:14)** Amends H-A-288 with the cause met: main.117 judged its + end at 17:45:19 but never wrote it, because the runner computed every mover's widens inside the end's + compare-and-set loop (126 movers, each asking grants-for and cut after five seconds: about eleven minutes per try) + while the pipeline's key kept changing, so every write was refused and everything read again until the restart. + The reads an end needs (widens, the catalog revision, the holders) are made once, before the compare-and-set, and + shared by every mover and every retry of the write; a retry rereads only the pipeline's key. P3, P-NATS-native (a + write to own state waits on no other part).