@daemonitor/plugins 1.1.9 → 1.1.11

package/dist/plugins/monitoring/SnmpPlugin.js

@@ -204,20 +204,20 @@ async function snmpWalkIndexed(t, oid, timeout = 10000) {
 async function collectQnap(t) {
     const tempWarn = t.tempWarnC ?? 70;
     const volWarn = t.volumeWarnPct ?? 90;
-    const [model, cpuTemp, sysTemp, hdNumRaw, cpuLoads, storDescrs, memTotal, memFree, fanSpeeds] = await Promise.all([
-        snmpGet(t, QNAP.model),
-        snmpGet(t, QNAP.cpuTemp),
-        snmpGet(t, QNAP.sysTemp),
-        snmpGet(t, QNAP.hdNumber),
-        snmpWalk(t, "1.3.6.1.2.1.25.3.3.1.2"), // hrProcessorLoad (per core)
-        snmpWalkIndexed(t, "1.3.6.1.2.1.25.2.3.1.3"), // hrStorageDescr
-        snmpGet(t, QNAP.memTotal),
-        snmpGet(t, QNAP.memFree),
-        snmpWalk(t, "1.3.6.1.4.1.24681.1.2.15.1.3"), // SystemFanSpeed table
+    // The light scalar gets can run together, but the WALKS must be SERIALIZED:
+    // QNAP's SNMP agent silently drops rows (and whole tables) under concurrent
+    // sessions, producing partial reads that flip a failing NAS back to a false
+    // "healthy". 5-min cadence — sequential latency is fine.
+    const [model, cpuTemp, sysTemp, memTotal, memFree] = await Promise.all([
+        snmpGet(t, QNAP.model), snmpGet(t, QNAP.cpuTemp), snmpGet(t, QNAP.sysTemp),
+        snmpGet(t, QNAP.memTotal), snmpGet(t, QNAP.memFree),
     ]);
-    if (model == null && cpuTemp == null && !cpuLoads.length) {
+    if (model == null && cpuTemp == null) {
         return { name: t.name, ip: t.ip, kind: "qnap", reachable: false };
     }
+    const cpuLoads = await snmpWalk(t, "1.3.6.1.2.1.25.3.3.1.2"); // hrProcessorLoad (per core)
+    const storDescrs = await snmpWalkIndexed(t, "1.3.6.1.2.1.25.2.3.1.3"); // hrStorageDescr
+    const fanSpeeds = await snmpWalk(t, "1.3.6.1.4.1.24681.1.2.15.1.3"); // SystemFanSpeed table
     // CPU: average per-core load.
     const cores = cpuLoads.map(Number).filter((n) => !isNaN(n));
     const cpuPct = cores.length ? Math.round(cores.reduce((a, b) => a + b, 0) / cores.length) : undefined;
@@ -227,26 +227,46 @@ async function collectQnap(t) {
         ? Math.round(((memTotalMB - memFreeMB) / memTotalMB) * 100) : undefined;
     // Fans — each entry looks like "1835 RPM".
     const fans = fanSpeeds.map((f) => parseNum(f)).filter((n) => n != null && n > 0);
-    // Disks — iterate 1..hdNumber (bounded). Presence is judged by MODEL/CAPACITY:
-    // an empty bay reports model "--", but a FAILING disk can report an unreadable
-    // temp ("-- C/-- F") + negative status yet still advertise a real model. The old
-    // temp+status heuristic silently skipped those bad disks (e.g. an HDD in SMART
-    // "Warning" with status -9) — so the NAS looked healthy while QTS flagged it.
-    const hdN = Math.min(Number(hdNumRaw) || 0, 24);
+    // Disks — walk each column of the disk table ONCE (6 small walks) rather than
+    // 6 gets × N bays. QNAP's SNMP agent is flaky under concurrent-get bursts and
+    // silently drops rows (partial reads that hid disks); a per-column walk is a
+    // single session each — far gentler and returns every populated row atomically.
+    // Presence keys on MODEL/CAPACITY ("--" = empty bay), so a FAILING disk
+    // (unreadable temp, negative status) is still reported, not mistaken for empty.
+    const DT = "1.3.6.1.4.1.24681.1.2.11.1."; // disk table columns: 2=descr 3=temp 4=status 5=model 6=capacity 7=smart
+    // Serialized (not Promise.all) — concurrent column walks are what make the
+    // agent drop rows.
+    const wStatus = await snmpWalkIndexed(t, DT + "4");
+    const wTemp = await snmpWalkIndexed(t, DT + "3");
+    const wSmart = await snmpWalkIndexed(t, DT + "7");
+    const wDescr = await snmpWalkIndexed(t, DT + "2");
+    const wModel = await snmpWalkIndexed(t, DT + "5");
+    const wCap = await snmpWalkIndexed(t, DT + "6");
+    const col = (rows) => {
+        const m = {};
+        for (const r of rows)
+            m[r.idx] = r.val;
+        return m;
+    };
+    const cSt = col(wStatus), cTp = col(wTemp), cSm = col(wSmart), cDs = col(wDescr), cMd = col(wModel), cCap = col(wCap);
+    const slots = (Object.keys(cMd).length ? Object.keys(cMd) : Object.keys(cDs))
+        .sort((a, b) => Number(a) - Number(b));
     const disks = [];
-    for (let i = 1; i <= hdN; i++) {
-        const [st, tp, sm, ds, md, cap] = await Promise.all([
-            snmpGet(t, QNAP.hdStatus + i), snmpGet(t, QNAP.hdTemp + i), snmpGet(t, QNAP.hdSmart + i),
-            snmpGet(t, QNAP.hdDescr + i), snmpGet(t, QNAP.hdModel + i), snmpGet(t, QNAP.hdCapacity + i),
-        ]);
-        const model = md && md !== "--" ? md : undefined;
-        const capacity = cap && cap !== "--" ? cap : undefined;
+    for (const i of slots) {
+        const model = cMd[i] && cMd[i] !== "--" ? cMd[i] : undefined;
+        const capacity = cCap[i] && cCap[i] !== "--" ? cCap[i] : undefined;
         // Only a truly empty bay (no model AND no capacity) is skipped.
         if (!model && !capacity)
             continue;
-        const status = st != null ? Number(st) : undefined;
-        const tempC = parseC(tp);
-        disks.push({ slot: i, descr: ds || undefined, model, capacity, tempC, status, smart: sm && sm !== "--" ? sm : undefined });
+        const status = cSt[i] != null ? Number(cSt[i]) : undefined;
+        const tempC = parseC(cTp[i] || null);
+        disks.push({ slot: Number(i), descr: cDs[i] || undefined, model, capacity, tempC, status, smart: cSm[i] && cSm[i] !== "--" ? cSm[i] : undefined });
+    }
+    // A reachable QNAP always has ≥1 disk. Zero means the disk-table walks came
+    // back empty (agent timed out) — treat as an INCOMPLETE read, not a clean
+    // bill of health, so a real disk error isn't falsely cleared to green.
+    if (!disks.length) {
+        return { name: t.name, ip: t.ip, kind: "qnap", reachable: true, error: "SNMP read incomplete — disk table empty" };
     }
     // A present disk is bad only on a real error status, an explicitly-bad SMART
     // verdict, or overheating — not a blank "--" or an empty bay.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@daemonitor/plugins",
-  "version": "1.1.9",
+  "version": "1.1.11",
   "description": "daemonitor plugins",
   "types": "types/index.d.ts",
   "main": "dist/index.js",