Say where an answer no notification can give is given, and never offer to silence data loss
mesh/merge-gate pass: builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step; every machine composes with the change as it…
mesh/repo-check pass: its merge-check.sh passed
mesh/delivery-group group fix/needs-you-from-the-console delivering: 1 of 2 delivered
mesh/delivery superseded: a newer delivery to the same trunk took over its walk

A "Needs you" with no button left the operator guessing where to act, and a click
could silence a condition that says data is gone. The place is named in the
glossary's word, the mesh MCP server (hq ADR 0258).
This commit is contained in:
jochen
2026-10-08 15:03:11 +02:00
parent 45ae1d0112
commit 2e6a9b1efc
2 changed files with 34 additions and 11 deletions
+18 -3
View File
@@ -68,7 +68,7 @@ func TestADeliveryWaitingNeedsNothingUntilItsBoundThenOffersStartAndStop(t *test
f.waits[0].since = now.Add(-5 * time.Hour)
got = watchWaits(f)
plainExample(t, got[0], "openrazer delivery waiting to start",
"Needs you: start it, or stop it. The change to openrazer is merged and built, and mesh-delivery (the "+
"Needs you: start it, or stop it, from the mesh MCP server; this notification cannot do it. The change to openrazer is merged and built, and mesh-delivery (the "+
"module that decides when a delivery goes out) has not let it start for 5 hours, so mesh-delivery may "+
"be stuck.")
@@ -88,7 +88,7 @@ func TestAModuleUnhealthyAsksForARestartInWords(t *testing.T) {
Resource: "openrazer-daemon", Target: "openrazer-daemon.service",
Reason: "failed in the account's own service manager (exit-code)", Since: time.Now()}})
plainExample(t, o, "openrazer not working on g14",
"Needs you: restart its service openrazer-daemon on g14; if it fails again, the details say why. "+
"Needs you: restart its service openrazer-daemon on g14 from the mesh MCP server; this notification cannot do it. "+
"openrazer on g14 is not healthy: its service openrazer-daemon stopped with an error. It clears as soon "+
"as it runs again.")
// An account waiting for a new login (ADR 0252) asks for the login, held to the plain rule.
@@ -149,7 +149,7 @@ func TestADeliveryHeldAsksForReleaseOrStopInWords(t *testing.T) {
got := stalledObservations([]stalledLine{{ID: "novox/hq@055550802096", State: "held", For: "36h2m6s",
Bound: "24h0m0s", H2: "none: the state is the operator's", Says: "it waits for the operator"}})
plainExample(t, got[0], "Delivery of hq held for 36 hours",
"Needs you: release it, or stop it. A delivery of hq has been held for 36 hours, past its limit.",
"Needs you: release it, or stop it, from the mesh MCP server; this notification cannot do it. A delivery of hq has been held for 36 hours, past its limit.",
)
}
@@ -204,3 +204,18 @@ func TestAnOperatorsAnswerIsNoHandRepair(t *testing.T) {
t.Fatalf("a silence by hand stopped counting: %+v", got)
}
}
// **Data loss is never silenced from a notification** (ADR 0258): data missing or shrunk offers no answer,
// and says where it is silenced.
func TestDataLossOffersNoSilence(t *testing.T) {
for _, kind := range []string{kindDataMissing, kindDataShrank} {
w := plainWordings[kind](conditions.Observation{Scope: conditions.ScopeMachine, ID: "ace.immich.library",
Machine: "ace", Kind: kind, Severity: conditions.Urgent})
if len(w.Actions) != 0 || !strings.HasSuffix(w.Needs, FromMeshMCPServer) {
t.Errorf("%s: %+v", kind, w)
}
if why, ok := conditions.PlainWords(w, "ace"); !ok {
t.Errorf("%s: %s", kind, why)
}
}
}