hub v0.132.0: the System page (versions + OS updates with the ring/switch/approve buttons, R-852), the Hosts Proxmox/kernel column, the operator-approved Docker engine release (2 healthy ring-0 nights), the crash-guard events (R-851); evidence audits/os-docker-crash-2026-10-04

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
2026-10-04 16:16:02 +02:00
parent bee277ffad
commit 175ecfcdd2
45 changed files with 2198 additions and 12 deletions
@@ -0,0 +1,67 @@
+package api
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+	"time"
+)
+
+func crashBody(boots []string, tripped bool, trippedAt, rearmedAt, bootID string, oops bool) []byte {
+	q := func(ss []string) string {
+		if len(ss) == 0 {
+			return "[]"
+		}
+		return `["` + strings.Join(ss, `","`) + `"]`
+	}
+	return []byte(fmt.Sprintf(`{"system":{"pve_version":"pve-manager/9.0.11/x","facts":{"host":{"oops_this_boot":%v,"kernel_panic":10,
+"crash_guard":{"unclean_boots":%s,"unclean_boots_24h":%d,"tripped":%v,"tripped_at":%q,"tripped_reason":"2 unclean boots within 60 minutes",
+"rearmed_at":%q,"rearmed_by":"operator","boot_id":%q}},"guest":{}}}}`, oops, q(boots), len(boots), tripped, trippedAt, rearmedAt, bootID))
+}
+
+func countEv(t *testing.T, h *Handler, cust, typ string) int {
+	t.Helper()
+	ev, err := h.store.GetEventsByType(cust, typ, time.Now().Add(-time.Hour))
+	if err != nil {
+		t.Fatal(err)
+	}
+	return len(ev)
+}
+
+// The first report only records history (no flood); each NEW unclean boot is one operator event + one household line;
+// the trip is one alarm; a repeat report is quiet. Red-proof: drop the `known[b]` skip and the repeat report emits again.
+func TestCheckCrash_EventsOncePerFact(t *testing.T) {
+	h, st, _ := newTestHandler(t)
+	seedHost(t, st, "h1", "c1", "K1")
+	old := []string{"2026-10-01T10:00:00Z"}
+	h.checkCrash("h1", "c1", crashBody(old, false, "", "", "b0", false))
+	if n := countEv(t, h, "c1", EventHostCrashRestart); n != 0 {
+		t.Fatalf("first sighting flooded %d events", n)
+	}
+	two := append(old, "2026-10-04T15:40:00Z")
+	h.checkCrash("h1", "c1", crashBody(two, false, "", "", "b1", false))
+	h.checkCrash("h1", "c1", crashBody(two, false, "", "", "b1", false)) // the same report again
+	if countEv(t, h, "c1", EventHostCrashRestart) != 1 || countEv(t, h, "c1", EventHostRestartedAfterCrash) != 1 {
+		t.Fatalf("restart events: op=%d household=%d", countEv(t, h, "c1", EventHostCrashRestart), countEv(t, h, "c1", EventHostRestartedAfterCrash))
+	}
+	three := append(two, "2026-10-04T15:45:00Z")
+	h.checkCrash("h1", "c1", crashBody(three, true, "2026-10-04T15:45:30Z", "", "b2", false))
+	h.checkCrash("h1", "c1", crashBody(three, true, "2026-10-04T15:45:30Z", "", "b2", false))
+	if countEv(t, h, "c1", EventCrashGuardTripped) != 1 {
+		t.Fatalf("tripped events = %d", countEv(t, h, "c1", EventCrashGuardTripped))
+	}
+	h.checkCrash("h1", "c1", crashBody(three, false, "", "2026-10-04T16:10:00Z", "b2", true))
+	if countEv(t, h, "c1", EventCrashGuardRearmed) != 1 || countEv(t, h, "c1", EventKernelOops) != 1 {
+		t.Fatal("re-arm / oops not announced")
+	}
+}
+
+// An agent older than v0.142.0 (no system stanza) never produces a crash event.
+func TestCheckCrash_NoStanzaIsQuiet(t *testing.T) {
+	h, st, _ := newTestHandler(t)
+	seedHost(t, st, "h1", "c1", "K1")
+	h.checkCrash("h1", "c1", []byte(`{"host":{}}`))
+	if st.CrashSeen("h1") != "" {
+		t.Fatal("a report without the stanza recorded crash state")
+	}
+}