novox/mesh-controller
6
0
Fork 0
Code Issues Pull Requests 3 Releases Wiki Activity

Gate a release plan's first machine and roll a failed build back there (hq ADR 0236)

Browse Source 
A build that reported applied was sent everywhere; one that then did nothing, served
no tools or broke its machine's word reached every machine. Now the first machine is
judged by the component's health (the core's definitions, as doctor probes H-*, or a
module's own) three times over two minutes within ten; a failing gate puts the previous
build back there once, marks the build, and says it as a condition and an event.
Upgrades roll out by default; the bus is a planned step; a module deleted at its
source is not built (the public-acme plan failure).
This commit is contained in:
jochen
2026-10-06 18:56:54 +02:00
parent 81f497e5dd
commit d9289ef6d4
41 changed files with 3297 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cmd/mesh-controller/plan_retry.go
+9
View File
@@ -178,6 +178,15 @@ func retryRefusal(p inventory.Plan, plans []inventory.Plan) error {
return fmt.Errorf("nothing in tier %d of %s failed to build or stopped rolling out — it stopped at: %s",
p.Tier, p.ID, p.Note)
}
// **A build that failed its gate is not sent again** (novox/hq ADR 0235): it was put back on its first
// machine, and retrying would judge the build the mesh put back, or send the failed one by hand.
for _, m := range stopped {
if g := p.Modules[m].Gate; g != nil && g.Verdict == inventory.GateFailed {
return fmt.Errorf("%s failed its gate on %s (%s) and was put back: a build that failed its gate is not "+
"sent again — a newer merge, or `rebuild %s`, makes a new build, judged at the gate again",
m, strings.Join(g.Machines, ", "), g.Why, m)
}
}
// **A rollout is retried unless the module has moved on**: a newer plan holding it sends — or
// sent — a newer build, and sending this one again would put the older build back on its machines.
for _, m := range stopped {