huskies: merge 1231 story project-rebuild must not silently downgrade a live sled to the stale image binary

2026-07-19 20:47:19 +00:00
parent 3f05648d25
commit 7bbd34bc3a
6 changed files with 344 additions and 10 deletions
@@ -6,6 +6,17 @@ set -euo pipefail
 #
 # Run this after `script/docker_rebuild` or whenever you add a new stack.
 # Safe to re-run: each step re-tags the image with the latest layers.
+#
+# IMPORTANT (story 1231): this script is NOT part of `script/release`. The
+# huskies-project-* images bake whatever `huskies` binary happened to be built
+# locally the last time this ran — as of this comment that's 0.13.0, several
+# releases behind the current fleet artifact. `project-rebuild` self-heals a
+# sled that comes back on a stale baked binary (it re-upgrades it in place
+# from the gateway's published artifact), but that costs an extra
+# download+restart cycle every time. Run `script/build-project-images` after
+# every `script/release` — ideally as a step in the release flow itself — so
+# freshly rebuilt sleds start on a current binary instead of relying on the
+# self-heal.

 cd "$(dirname "$0")/.."