novox/hq
3
0
Fork 0
Code Issues 5 Pull Requests 1 Releases Wiki Activity

Issue 188: what the live fault was and how it was resolved

Browse Source
This commit is contained in:
jschoubben
2026-10-01 18:14:37 +02:00
parent 187442ec7b
commit a92e4bf121
1 changed files with 8 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
04-ISSUES/188-a-refusal-inside-on-the-network-drops-a-machine-silently/00-report.md
+8 -3
View File
@@ -43,6 +43,11 @@ where it happens and discovered where it hurts.
## Resolved in the live mesh, 2026-10-01
By hand: `pin novox acme-ca novox step-ca`, the provider the previous controller had in fact bound
the proxy to (read from its plan), after which the control node resolved, every seat read as held
and the roll-out proceeded. The design fault above stands.
The refusal itself was a fault of mesh-controller 195, already corrected on main by its author's
hotfix (196) when the control node was found refusing; the running controller was the one build in
between. Found by running the previous image and main's image as one-shots beside the running one
and reading which resolved. Resolved by pushing the control node from a one-shot of main's image, as
the recipe for a controller that cannot roll itself says. A pin naming the proxy's issuer
(`step-ca`, the one the previous plan had bound) was made first and kept; it changes nothing. The
design fault above stands and is fixed by mesh-controller PR `fix/a-machine-not-on-the-network-is-said`:
the dropped machine and the resolver's words are said where the drop happens.