From 147a79962d12e6272f8c22e2a9317f732dbb7a59 Mon Sep 17 00:00:00 2001 From: zzstoatzz Date: Sun, 16 Aug 2026 08:14:06 +0000 Subject: [PATCH] handoff: compaction interval restored to 4h (config-only restart 08:13Z) Co-Authored-By: Claude Fable 5 --- HANDOFF.md | 9 +++++++-- 1 file(s) changed, 7 insertion(s)(+), 2 deletion(s)(-) diff --git a/HANDOFF.md b/HANDOFF.md --- a/HANDOFF.md +++ b/HANDOFF.md @@ -13,8 +13,13 @@ 07:13:35, tombstones rebuilt 07:16:22 (2,462,389 entries), compactor timer 86400s, site 200 throughout the startup window. The old ~2.5-min dark window is gone; fleet-health also gained a 6-min startup grace. - Restart reset both timers again: next compaction AND failed-repo retry - pass land ~07:12Z 2026-08-17 — that retry pass is the watch point + UPDATE 08:13Z: compaction interval restored to 4h (upstream's default; + the 24h was incident-#11 mitigation, and boot-relative timers + daily + deploys had starved compaction since the Aug 14 16:33Z drain — zero + passes since, verified in prometheus). Config-only restart at 08:13:17Z + (compose.yaml edit, backup kept). Retry stays 24h. First compaction + pass ~12:13Z 08-16, then 4-hourly; retry pass ~08:13Z 08-17 — that + retry pass is the watch point (pds_direct >> relay_fallback, no ingest dip); after a clean pass, consider restoring the 4h retry interval. Gate saga lesson recorded below the incident block: eight cycles, one invariant ("port answers ⇒ -- tangled.sh