Alert System Changes: Fixes applied (feature kept, load bounded)

1. Removed docker.df() entirely
2. Poll only what rules need
   • docker_daemon → cheap ping
   • container_health / stack_health → listContainers
   • resource → host statfs only (no Docker)
   • Event-only rules → no poll timer
3. Poll floor 60s, default 120s (was 15s min / 60s default)
4. No overlapping polls (pollInFlight)
5. Event path: drop noisy actions (exec_*, attach, …), match only relevant rules, serialize work (bounded queue)
6. UI status shows whether poll is active

config.js

@@ -4,15 +4,16 @@
 
 export const CONFIG = {
   // Stats collection (server/services/stats.js)
-  // Prefer one-shot polls over perpetual stream:true attachments — each live
-  // stream forces dockerd to sample that container ~1Hz and is a host CPU hog.
-  // Override at runtime: PEARDOCK_STATS_INTERVAL_MS, PEARDOCK_STATS_CONCURRENCY.
+  // Round-robin one-shot samples — not full-fleet stream:true (Dozzle-style streams
+  // keep dockerd busy ~1Hz × N containers). See Settings → Performance.
   STATS: {
-    INTERVAL_MS: 2000,              // Min gap between allStats broadcasts
-    ACTIVE_INTERVAL_MS: 2000,       // Collection tick while peers are connected
+    INTERVAL_MS: 5000,              // Min gap between allStats broadcasts
+    ACTIVE_INTERVAL_MS: 5000,       // Idle gap after each sample batch completes
     IDLE_INTERVAL_MS: 5000,         // Reserved (peer-idle uses full pause)
-    CACHE_TTL_MS: 1000,             // Per-container broadcast cache TTL
-    CONCURRENCY: 6,                 // Max parallel one-shot stats to dockerd
+    CACHE_TTL_MS: 2000,             // Per-container broadcast cache TTL
+    CONCURRENCY: 2,                 // Max parallel one-shot stats to dockerd
+    SAMPLES_PER_TICK: 4,            // Max containers sampled per tick (round-robin)
+    LIST_INTERVAL_MS: 15000,        // How often to re-list containers
     SMOOTHING_FACTOR: 0.2,          // Client-side smoothing (if used)
   },
 

docs/SERVER.md

@@ -105,16 +105,22 @@ Subscribes to Engine events; fans out `push:dockerEvent` and may trigger list re
 
 Periodic container stats → `push:allStats`; history buffer for charts.
 
-**Important:** collection uses **one-shot** `stats({ stream: false })` with bounded concurrency (default 6), not perpetual per-container streams. Live `stream: true` attachments force dockerd to sample every running container ~1 Hz and routinely pin host load average. Stats only run while `peers.size > 0`. Operators can tune intervals and concurrency from the client under **Settings → Performance** (admin) without restarting the server.
+**Important:** peardock does **not** open perpetual per-container `stream: true` stats (Dozzle does, and [documents elevated dockerd CPU](https://dozzle.dev/guide/faq) as expected). Instead:
+
+1. **Round-robin one-shot** samples (`stream: false`) — at most `samplesPerTick` containers per batch  
+2. **Post-batch idle gap** — next batch waits `collectIntervalMs` after the previous finishes (no overlapping sweeps)  
+3. **Roster throttle** — `listContainers` only every `listIntervalMs`  
+4. **Peer-idle pause** — no sampling when `peers.size === 0`
 
 | Env / UI | Default | Meaning |
 |----------|---------|---------|
-| `PEARDOCK_STATS_INTERVAL_MS` / Settings → Performance | 2000 | Collection tick while peers connected |
-| `PEARDOCK_STATS_BROADCAST_MS` | 2000 | Min gap between `push:allStats` |
-| `PEARDOCK_STATS_CONCURRENCY` | 6 | Max parallel one-shot stats requests |
-| `PEARDOCK_STATS_CACHE_TTL_MS` | 1000 | Idle-container broadcast cache TTL |
+| Collect gap / `PEARDOCK_STATS_INTERVAL_MS` | 5000 | Idle after each sample batch |
+| Samples/tick / `PEARDOCK_STATS_SAMPLES_PER_TICK` | 4 | Max containers sampled per batch (main load knob) |
+| Concurrency / `PEARDOCK_STATS_CONCURRENCY` | 2 | Parallel one-shots within a batch |
+| List interval / `PEARDOCK_STATS_LIST_INTERVAL_MS` | 15000 | Roster refresh |
+| Broadcast / `PEARDOCK_STATS_BROADCAST_MS` | 5000 | Min gap between `push:allStats` |
 
-Runtime values persist in `peardock-stats.json` (cwd or `PEARDOCK_STATS_PATH`). RPC: `getStatsConfig` (viewer), `updateStatsConfig` (admin). The Settings → **Performance** tab applies changes live with no server restart.
+Runtime values persist in `peardock-stats.json` (cwd or `PEARDOCK_STATS_PATH`). RPC: `getStatsConfig` (viewer), `updateStatsConfig` (admin). Settings → **Performance** applies live. On hot hosts: raise collect gap, keep samples/tick at 2–4 and concurrency at 1–2.
 
 ### Metrics (`services/metrics.js`)
 

index.html

@@ -2906,33 +2906,43 @@ services:
             <p class="small text-muted mb-2">Changes apply live on the server — no restart and no Save button.</p>
             <div class="row g-3 mt-1">
               <div class="col-md-3">
-                <label class="form-label" for="stats-collect-sec">Collect interval (sec)</label>
-                <input type="number" id="stats-collect-sec" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="60" step="1" value="2">
-                <div class="form-text">How often the server starts a stats sweep while a client is connected. Higher = less dockerd CPU.</div>
+                <label class="form-label" for="stats-collect-sec">Idle gap after batch (sec)</label>
+                <input type="number" id="stats-collect-sec" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="60" step="1" value="5">
+                <div class="form-text">Wait this long <em>after</em> a sample batch finishes before the next. Higher = less dockerd CPU.</div>
               </div>
               <div class="col-md-3">
-                <label class="form-label" for="stats-broadcast-sec">Broadcast interval (sec)</label>
-                <input type="number" id="stats-broadcast-sec" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="60" step="1" value="2">
-                <div class="form-text">Min gap between <code>push:allStats</code> updates to clients.</div>
+                <label class="form-label" for="stats-samples-per-tick">Samples per tick</label>
+                <input type="number" id="stats-samples-per-tick" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="100" step="1" value="4">
+                <div class="form-text">Containers sampled each tick (round-robin). <strong>Main load control</strong> — do not set to fleet size.</div>
               </div>
               <div class="col-md-3">
                 <label class="form-label" for="stats-concurrency">Sample concurrency</label>
-                <input type="number" id="stats-concurrency" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="32" step="1" value="6">
-                <div class="form-text">Max parallel one-shot stats requests to dockerd. Keep modest on busy hosts.</div>
+                <input type="number" id="stats-concurrency" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="32" step="1" value="2">
+                <div class="form-text">Max parallel one-shot stats to dockerd within a tick. Keep at 1–2 on busy hosts.</div>
               </div>
               <div class="col-md-3">
-                <label class="form-label" for="stats-cache-ttl-ms">Idle cache TTL (ms)</label>
-                <input type="number" id="stats-cache-ttl-ms" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="0" max="30000" step="100" value="1000">
-                <div class="form-text">Reuse last sample for idle containers when broadcasting (0 = always fresh).</div>
+                <label class="form-label" for="stats-broadcast-sec">Broadcast interval (sec)</label>
+                <input type="number" id="stats-broadcast-sec" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="1" max="60" step="1" value="5">
+                <div class="form-text">Min gap between <code>push:allStats</code> to clients (can reuse last samples).</div>
               </div>
             </div>
             <div class="row g-3 mt-1">
+              <div class="col-md-3">
+                <label class="form-label" for="stats-list-sec">List containers every (sec)</label>
+                <input type="number" id="stats-list-sec" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="2" max="120" step="1" value="15">
+                <div class="form-text">Roster refresh rate (start/stop detection for the sampler).</div>
+              </div>
+              <div class="col-md-3">
+                <label class="form-label" for="stats-cache-ttl-ms">Idle cache TTL (ms)</label>
+                <input type="number" id="stats-cache-ttl-ms" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="0" max="30000" step="100" value="2000">
+                <div class="form-text">Reuse last sample for idle containers when broadcasting.</div>
+              </div>
               <div class="col-md-3">
                 <label class="form-label" for="stats-warn-containers">Large-fleet warn threshold</label>
                 <input type="number" id="stats-warn-containers" class="form-control bg-dark text-white stats-live-field" data-min-role="admin" min="10" max="10000" step="1" value="80">
-                <div class="form-text">Log a one-time warning when this many containers are running.</div>
+                <div class="form-text">One-time server log when this many containers are running.</div>
               </div>
-              <div class="col-md-9 d-flex align-items-end flex-wrap gap-2">
+              <div class="col-md-3 d-flex align-items-end flex-wrap gap-2">
                 <button type="button" class="btn btn-outline-secondary btn-sm" id="stats-reset-defaults" data-min-role="admin" title="Reset to server defaults">
                   <i class="fas fa-undo me-1"></i>Reset to defaults
                 </button>
@@ -3098,7 +3108,8 @@ services:
               </div>
               <div class="col-md-2">
                 <label class="form-label" for="alerts-poll-ms">Health poll (sec)</label>
-                <input type="number" id="alerts-poll-ms" class="form-control bg-dark text-white alerts-live-field" data-min-role="admin" min="15" max="3600" value="60">
+                <input type="number" id="alerts-poll-ms" class="form-control bg-dark text-white alerts-live-field" data-min-role="admin" min="60" max="3600" value="120" title="Minimum 60s — lower values thrash Docker">
+                <div class="form-text">Min 60s. Only runs for daemon / health / stack / disk rules (not pure event rules).</div>
               </div>
               <div class="col-md-2">
                 <label class="form-label" for="alerts-rate-limit">Max / minute</label>

server/services/alerts.js

@@ -3,7 +3,14 @@
  *
  * Fully customizable rules + webhook channels (Discord, Slack, Teams,
  * generic HTTP, ntfy, Gotify, Telegram). Evaluates Docker events and
- * periodic health polls even when no client is connected.
+ * light periodic health polls even when no client is connected.
+ *
+ * Host-load rules (keep this cheap on dockerd):
+ * - Never call docker.df() (expensive, unused).
+ * - Poll only kinds that have enabled rules (skip listContainers when unused).
+ * - Min poll interval 60s (default 120s); never overlap polls.
+ * - Docker event path filters noise (exec_*, attach, …) and serializes work.
+ * - Event-only rules need no Docker polling.
  *
  * Config (mode 0600), first match wins:
  *   PEARDOCK_ALERTS_PATH
@@ -45,10 +52,36 @@ const FILE = resolveAlertsFilePath()
 const LEGACY_CWD_FILE = path.join(process.cwd(), 'peardock-alerts.json')
 
 const HISTORY_MAX = 200
-const DEFAULT_POLL_MS = 60_000
+/** Default health poll — 2 minutes (was 60s). */
+const DEFAULT_POLL_MS = 120_000
+/** Hard floor — sub-minute listContainers/ping loops thrash dockerd. */
+const MIN_POLL_MS = 60_000
+const MAX_POLL_MS = 3_600_000
+/** Drop oldest under event flood rather than unbounded memory. */
+const EVENT_QUEUE_MAX = 200
 const SEVERITIES = ['info', 'warning', 'critical']
 const SEV_RANK = { info: 1, warning: 2, critical: 3 }
 
+/**
+ * Docker event actions that never match default peardock alert rules and are
+ * extremely chatty (terminals, attach, top). Skip before evaluateRules.
+ */
+const NOISY_DOCKER_ACTIONS = new Set([
+  'exec_create',
+  'exec_start',
+  'exec_die',
+  'exec_detach',
+  'attach',
+  'detach',
+  'resize',
+  'top',
+  'export',
+  'commit',
+  'copy',
+  'archive-path',
+  'extract-to-dir',
+])
+
 const CHANNEL_TYPES = new Set([
   'discord',
   'slack',
@@ -85,6 +118,11 @@ let deliveriesThisMinute = 0
 let deliveriesMinuteStart = Date.now()
 let lastDaemonOk = true
 let started = false
+/** Prevent overlapping pollHealth (setInterval + slow listContainers). */
+let pollInFlight = false
+/** Serialize docker-event alert evaluation under noisy fleets. */
+let eventQueue = []
+let eventDraining = false
 
 function newId(prefix) {
   return `${prefix}_${Date.now().toString(36)}_${randomBytes(3).toString('hex')}`
@@ -316,8 +354,8 @@ function normalizeConfig(raw) {
     version: 1,
     enabled: r.enabled !== false,
     pollIntervalMs: Math.max(
-      15_000,
-      Math.min(3_600_000, Number(r.pollIntervalMs) || DEFAULT_POLL_MS)
+      MIN_POLL_MS,
+      Math.min(MAX_POLL_MS, Number(r.pollIntervalMs) || DEFAULT_POLL_MS)
     ),
     minSeverity: normalizeSeverity(r.minSeverity, 'info'),
     rateLimitPerMinute: Math.max(1, Math.min(300, Number(r.rateLimitPerMinute) || 40)),
@@ -1031,141 +1069,260 @@ async function evaluateRules(kind, ctx) {
 }
 
 /**
- * Called from Docker event stream.
- * @param {object} event
+ * @param {string} kind
+ * @returns {boolean}
  */
-export async function onDockerEvent(event) {
-  if (!event || !config.enabled) return
-  try {
-    await evaluateRules('docker_event', event)
-    // Health status often arrives as container health_status event
-    const action = String(event.Action || event.status || '')
-    if (event.Type === 'container' && action.startsWith('health_status')) {
-      const health = action.includes(':')
-        ? action.split(':').slice(1).join(':').trim()
-        : event.status
-      await evaluateRules('container_health', {
-        id: event.id || event.Actor?.ID,
-        name: event.Actor?.Attributes?.name,
-        labels: event.Actor?.Attributes,
-        health: health || event.Actor?.Attributes?.health_status,
-      })
-    }
-  } catch (err) {
-    logger.warn('alerts: onDockerEvent failed', { error: err.message })
+function hasEnabledRuleKind(kind) {
+  return config.rules.some(
+    (r) =>
+      r.enabled &&
+      r.kind === kind &&
+      severityAtLeast(r.severity, config.minSeverity)
+  )
+}
+
+/**
+ * Kinds that require a periodic Docker / host poll (not pure events).
+ */
+function pollNeeds() {
+  return {
+    daemon: hasEnabledRuleKind('docker_daemon'),
+    list: hasEnabledRuleKind('container_health') || hasEnabledRuleKind('stack_health'),
+    resource: hasEnabledRuleKind('resource'),
   }
 }
 
+/**
+ * Whether this Docker event could fire any enabled alert rule.
+ * Drops terminal/exec noise that was previously evaluated on every event.
+ * @param {object} event
+ */
+function isAlertRelevantEvent(event) {
+  if (!event) return false
+  const type = String(event.Type || event.type || '')
+  const actionRaw = String(event.Action || event.status || '')
+  const action = actionRaw.split(':')[0]
+
+  if (NOISY_DOCKER_ACTIONS.has(action)) return false
+
+  // Healthcheck transitions → container_health rules
+  if (
+    type === 'container' &&
+    actionRaw.startsWith('health_status') &&
+    hasEnabledRuleKind('container_health')
+  ) {
+    return true
+  }
+
+  if (!hasEnabledRuleKind('docker_event')) return false
+
+  for (const rule of config.rules) {
+    if (!rule.enabled || rule.kind !== 'docker_event') continue
+    if (!severityAtLeast(rule.severity, config.minSeverity)) continue
+    const match = rule.match || {}
+    if (Array.isArray(match.types) && match.types.length && !match.types.includes(type)) {
+      continue
+    }
+    if (
+      Array.isArray(match.actions) &&
+      match.actions.length &&
+      !match.actions.includes(action)
+    ) {
+      continue
+    }
+    return true
+  }
+  return false
+}
+
+/**
+ * Process one Docker event for alert rules (awaited).
+ * @param {object} event
+ */
+async function processDockerEvent(event) {
+  await evaluateRules('docker_event', event)
+  const action = String(event.Action || event.status || '')
+  if (event.Type === 'container' && action.startsWith('health_status')) {
+    const health = action.includes(':')
+      ? action.split(':').slice(1).join(':').trim()
+      : event.status
+    await evaluateRules('container_health', {
+      id: event.id || event.Actor?.ID,
+      name: event.Actor?.Attributes?.name,
+      labels: event.Actor?.Attributes,
+      health: health || event.Actor?.Attributes?.health_status,
+    })
+  }
+}
+
+async function drainEventQueue() {
+  if (eventDraining) return
+  eventDraining = true
+  try {
+    while (eventQueue.length) {
+      const ev = eventQueue.shift()
+      if (!ev || !config.enabled) continue
+      try {
+        await processDockerEvent(ev)
+      } catch (err) {
+        logger.warn('alerts: onDockerEvent failed', { error: err.message })
+      }
+    }
+  } finally {
+    eventDraining = false
+    // Work may have been enqueued while draining
+    if (eventQueue.length) {
+      drainEventQueue().catch(() => {})
+    }
+  }
+}
+
+/**
+ * Called from Docker event stream (fire-and-forget from events.js).
+ * Cheap sync filter + bounded queue; never piles concurrent evaluate/fire work.
+ * @param {object} event
+ */
+export function onDockerEvent(event) {
+  if (!event || !config.enabled) return
+  if (!isAlertRelevantEvent(event)) return
+  if (eventQueue.length >= EVENT_QUEUE_MAX) {
+    eventQueue.shift()
+  }
+  eventQueue.push(event)
+  drainEventQueue().catch((err) => {
+    logger.debug('alerts: event drain error', { error: err?.message })
+  })
+}
+
+/**
+ * Periodic health poll. Skips Docker work for disabled rule kinds.
+ * Never calls docker.df (expensive and unused).
+ */
 async function pollHealth() {
   if (!config.enabled) return
-
-  // Docker daemon
-  try {
-    await docker.ping()
-    lastDaemonOk = true
-    await evaluateRules('docker_daemon', { ok: true })
-  } catch (err) {
-    lastDaemonOk = false
-    await evaluateRules('docker_daemon', {
-      ok: false,
-      error: err.message || String(err),
-    })
-    return // skip further polls if docker is down
+  if (pollInFlight) {
+    logger.debug('alerts: poll skipped (previous still running)')
+    return
+  }
+  const needs = pollNeeds()
+  if (!needs.daemon && !needs.list && !needs.resource) {
+    // Only docker_event / peardock rules — pure event path, no polling work
+    return
   }
 
-  // One listContainers for both container_health and stack_health (half the socket work)
-  /** @type {object[]} */
-  let containers = []
+  pollInFlight = true
   try {
-    containers = (await docker.listContainers({ all: true })) || []
-  } catch (err) {
-    logger.debug('alerts: container list poll failed', { error: err.message })
-  }
-
-  // Container health
-  try {
-    for (const c of containers) {
-      const health = c.Status?.match(/\((healthy|unhealthy|health: starting)\)/i)?.[1]
-        || c.State // running/exited — Health may be in inspect only
-      // Prefer Health from inspect-lite if present
-      const h =
-        c.Health?.Status ||
-        (typeof health === 'string' ? health.toLowerCase().replace('health: ', '') : null)
-      if (!h || h === 'none') continue
-      const name = (c.Names?.[0] || '').replace(/^\//, '') || c.Id?.slice(0, 12)
-      await evaluateRules('container_health', {
-        id: c.Id,
-        name,
-        labels: c.Labels,
-        health: h === 'starting' ? 'starting' : h,
-      })
-    }
-  } catch (err) {
-    logger.debug('alerts: container health poll failed', { error: err.message })
-  }
-
-  // Stack / compose project health (group by com.docker.compose.project)
-  try {
-    /** @type {Map<string, { name: string, running: number, total: number, unhealthy: number }>} */
-    const stacks = new Map()
-    for (const c of containers) {
-      const project =
-        c.Labels?.['com.docker.compose.project'] ||
-        c.Labels?.['com.docker.stack.namespace']
-      if (!project) continue
-      let s = stacks.get(project)
-      if (!s) {
-        s = { name: project, running: 0, total: 0, unhealthy: 0 }
-        stacks.set(project, s)
+    if (needs.daemon) {
+      try {
+        await docker.ping()
+        lastDaemonOk = true
+        await evaluateRules('docker_daemon', { ok: true })
+      } catch (err) {
+        lastDaemonOk = false
+        await evaluateRules('docker_daemon', {
+          ok: false,
+          error: err.message || String(err),
+        })
+        return // skip list/resource if docker is down
       }
-      s.total += 1
-      if (c.State === 'running') s.running += 1
-      const st = String(c.Status || '')
-      if (/\(unhealthy\)/i.test(st)) s.unhealthy += 1
     }
-    for (const s of stacks.values()) {
-      const degraded = s.running < s.total || s.unhealthy > 0
-      await evaluateRules('stack_health', {
-        name: s.name,
-        status: degraded ? 'degraded' : 'ok',
-        running: s.running,
-        desired: s.total,
-        message: degraded
-          ? `${s.running}/${s.total} running${s.unhealthy ? `, ${s.unhealthy} unhealthy` : ''}`
-          : 'all services running',
-      })
-    }
-  } catch (err) {
-    logger.debug('alerts: stack poll failed', { error: err.message })
-  }
 
-  // Disk via system df
-  try {
-    if (typeof docker.df === 'function') {
-      const df = await docker.df()
-      // Docker df doesn't give host disk %; approximate via layers size vs nothing.
-      // Prefer host free space from / if available via system info
+    /** @type {object[]} */
+    let containers = []
+    if (needs.list) {
+      try {
+        containers = (await docker.listContainers({ all: true })) || []
+      } catch (err) {
+        logger.debug('alerts: container list poll failed', { error: err.message })
+      }
     }
-    // Host disk from Node (server filesystem)
-    try {
-      const { statfsSync } = await import('fs')
-      if (typeof statfsSync === 'function') {
-        const st = statfsSync('/')
-        const total = Number(st.blocks) * Number(st.bsize)
-        const free = Number(st.bfree) * Number(st.bsize)
-        if (total > 0) {
-          const usedPct = ((total - free) / total) * 100
-          await evaluateRules('resource', {
-            diskPercent: usedPct,
-            mount: '/',
+
+    if (needs.list && hasEnabledRuleKind('container_health')) {
+      try {
+        for (const c of containers) {
+          const health =
+            c.Status?.match(/\((healthy|unhealthy|health: starting)\)/i)?.[1] || c.State
+          const h =
+            c.Health?.Status ||
+            (typeof health === 'string'
+              ? health.toLowerCase().replace('health: ', '')
+              : null)
+          if (!h || h === 'none') continue
+          // Skip pure state strings that aren't healthcheck results
+          if (h === 'running' || h === 'exited' || h === 'created' || h === 'paused') {
+            continue
+          }
+          const name = (c.Names?.[0] || '').replace(/^\//, '') || c.Id?.slice(0, 12)
+          await evaluateRules('container_health', {
+            id: c.Id,
+            name,
+            labels: c.Labels,
+            health: h === 'starting' ? 'starting' : h,
           })
         }
+      } catch (err) {
+        logger.debug('alerts: container health poll failed', { error: err.message })
       }
-    } catch {
-      // statfs not available on all platforms
     }
-  } catch (err) {
-    logger.debug('alerts: resource poll failed', { error: err.message })
+
+    if (needs.list && hasEnabledRuleKind('stack_health')) {
+      try {
+        /** @type {Map<string, { name: string, running: number, total: number, unhealthy: number }>} */
+        const stacks = new Map()
+        for (const c of containers) {
+          const project =
+            c.Labels?.['com.docker.compose.project'] ||
+            c.Labels?.['com.docker.stack.namespace']
+          if (!project) continue
+          let s = stacks.get(project)
+          if (!s) {
+            s = { name: project, running: 0, total: 0, unhealthy: 0 }
+            stacks.set(project, s)
+          }
+          s.total += 1
+          if (c.State === 'running') s.running += 1
+          const st = String(c.Status || '')
+          if (/\(unhealthy\)/i.test(st)) s.unhealthy += 1
+        }
+        for (const s of stacks.values()) {
+          const degraded = s.running < s.total || s.unhealthy > 0
+          await evaluateRules('stack_health', {
+            name: s.name,
+            status: degraded ? 'degraded' : 'ok',
+            running: s.running,
+            desired: s.total,
+            message: degraded
+              ? `${s.running}/${s.total} running${s.unhealthy ? `, ${s.unhealthy} unhealthy` : ''}`
+              : 'all services running',
+          })
+        }
+      } catch (err) {
+        logger.debug('alerts: stack poll failed', { error: err.message })
+      }
+    }
+
+    // Host disk only — never docker.df() (expensive Engine inventory scan).
+    if (needs.resource) {
+      try {
+        const { statfsSync } = await import('fs')
+        if (typeof statfsSync === 'function') {
+          const st = statfsSync('/')
+          const total = Number(st.blocks) * Number(st.bsize)
+          const free = Number(st.bfree) * Number(st.bsize)
+          if (total > 0) {
+            const usedPct = ((total - free) / total) * 100
+            await evaluateRules('resource', {
+              diskPercent: usedPct,
+              mount: '/',
+            })
+          }
+        }
+      } catch (err) {
+        logger.debug('alerts: resource poll failed', { error: err?.message })
+      }
+    }
+  } finally {
+    pollInFlight = false
   }
 }
 
@@ -1175,7 +1332,12 @@ function armPoll() {
     pollTimer = null
   }
   if (!config.enabled) return
-  const ms = config.pollIntervalMs || DEFAULT_POLL_MS
+  const needs = pollNeeds()
+  if (!needs.daemon && !needs.list && !needs.resource) {
+    logger.debug('alerts: poll timer off (no poll-backed rules enabled)')
+    return
+  }
+  const ms = Math.max(MIN_POLL_MS, config.pollIntervalMs || DEFAULT_POLL_MS)
   pollTimer = setInterval(() => {
     pollHealth().catch((err) => {
       logger.debug('alerts: poll error', { error: err.message })
@@ -1228,6 +1390,9 @@ export function stopAlerts() {
     clearInterval(pollTimer)
     pollTimer = null
   }
+  eventQueue = []
+  eventDraining = false
+  pollInFlight = false
   if (saveTimer) {
     clearTimeout(saveTimer)
     saveTimer = null
@@ -1333,6 +1498,8 @@ export function upsertAlertRule(input) {
   if (idx >= 0) config.rules[idx] = rule
   else config.rules.push(rule)
   flushSave()
+  // Rule kind/enable changes which Docker polls we need
+  armPoll()
   logger.debug('alerts: rule upserted live', {
     id: rule.id,
     enabled: rule.enabled,
@@ -1345,6 +1512,7 @@ export function deleteAlertRule(id) {
   const before = config.rules.length
   config.rules = config.rules.filter((r) => r.id !== id)
   flushSave()
+  armPoll()
   return before !== config.rules.length
 }
 
@@ -1354,6 +1522,7 @@ export function listAlertHistory(limit = 50) {
 }
 
 export function getAlertsStatus() {
+  const needs = pollNeeds()
   return {
     enabled: config.enabled,
     channelCount: config.channels.filter((c) => c.enabled).length,
@@ -1361,6 +1530,10 @@ export function getAlertsStatus() {
     lastDaemonOk,
     historyCount: history.length,
     pollIntervalMs: config.pollIntervalMs,
+    pollActive: Boolean(pollTimer),
+    pollInFlight,
+    pollNeeds: needs,
+    eventQueueDepth: eventQueue.length,
     quietHoursActive: inQuietHours(),
     rateLimitPerMinute: config.rateLimitPerMinute,
   }

server/services/stats.js

@@ -1,15 +1,18 @@
 /**
  * Container stats collection and broadcast.
  *
- * Uses **one-shot** Docker stats (stream:false) with bounded concurrency —
- * not perpetual `stats({ stream: true })` attachments.
+ * Strategy (tuned against dockerd load, informed by Dozzle's approach):
  *
- * Why: each live stats stream forces dockerd to sample cgroups ~1 Hz for that
- * container for as long as the stream is open. With N running containers that
- * is N concurrent streams and is a well-known host CPU / load-average killer.
- * One-shot polls only work when we ask, and we can throttle concurrency.
- *
- * Collection only runs while peers.size > 0 (paused when idle).
+ * - **One-shot** `stats({ stream: false })` — no perpetual stream:true attachments.
+ *   Dozzle uses long-lived streams for all running containers and documents
+ *   elevated dockerd CPU as expected; we intentionally avoid that.
+ * - **Round-robin batches** — each tick samples at most `samplesPerTick`
+ *   containers (not the entire fleet). Full-fleet one-shot sweeps are often
+ *   *worse* than streams when sweep time > interval (dockerd never idles).
+ * - **Post-sweep spacing** — next tick is scheduled only after the previous
+ *   finishes + collectIntervalMs (not a fixed setInterval that overlaps work).
+ * - **Roster cache** — listContainers is not called every tick.
+ * - **Peer-idle pause** — no collection when peers.size === 0.
  */
 import fs from 'fs'
 import path from 'path'
@@ -28,6 +31,8 @@ export const STATS_LIMITS = Object.freeze({
   collectIntervalMs: { min: 1000, max: 60_000 },
   broadcastIntervalMs: { min: 1000, max: 60_000 },
   concurrency: { min: 1, max: 32 },
+  samplesPerTick: { min: 1, max: 100 },
+  listIntervalMs: { min: 2000, max: 120_000 },
   cacheTtlMs: { min: 0, max: 30_000 },
   warnContainers: { min: 10, max: 10_000 },
 })
@@ -45,10 +50,13 @@ function clampInt(value, min, max, fallback) {
 
 /**
  * Built-in defaults, then env overrides (bootstrap before disk / UI).
+ * Conservative defaults keep dockerd idle between batches on busy hosts.
  * @returns {{
  *   collectIntervalMs: number,
  *   broadcastIntervalMs: number,
  *   concurrency: number,
+ *   samplesPerTick: number,
+ *   listIntervalMs: number,
  *   cacheTtlMs: number,
  *   warnContainers: number,
  * }}
@@ -57,29 +65,41 @@ export function defaultStatsConfig() {
   const collectIntervalMs = clampInt(
     envInt(
       'PEARDOCK_STATS_INTERVAL_MS',
-      CONFIG.STATS?.ACTIVE_INTERVAL_MS ?? CONFIG.STATS?.INTERVAL_MS ?? 2000
+      CONFIG.STATS?.ACTIVE_INTERVAL_MS ?? CONFIG.STATS?.INTERVAL_MS ?? 5000
     ),
     STATS_LIMITS.collectIntervalMs.min,
     STATS_LIMITS.collectIntervalMs.max,
-    2000
+    5000
   )
   const broadcastIntervalMs = clampInt(
-    envInt('PEARDOCK_STATS_BROADCAST_MS', CONFIG.STATS?.INTERVAL_MS ?? 2000),
+    envInt('PEARDOCK_STATS_BROADCAST_MS', CONFIG.STATS?.INTERVAL_MS ?? 5000),
     STATS_LIMITS.broadcastIntervalMs.min,
     STATS_LIMITS.broadcastIntervalMs.max,
-    2000
+    5000
   )
   const concurrency = clampInt(
-    envInt('PEARDOCK_STATS_CONCURRENCY', CONFIG.STATS?.CONCURRENCY ?? 6),
+    envInt('PEARDOCK_STATS_CONCURRENCY', CONFIG.STATS?.CONCURRENCY ?? 2),
     STATS_LIMITS.concurrency.min,
     STATS_LIMITS.concurrency.max,
-    6
+    2
+  )
+  const samplesPerTick = clampInt(
+    envInt('PEARDOCK_STATS_SAMPLES_PER_TICK', CONFIG.STATS?.SAMPLES_PER_TICK ?? 4),
+    STATS_LIMITS.samplesPerTick.min,
+    STATS_LIMITS.samplesPerTick.max,
+    4
+  )
+  const listIntervalMs = clampInt(
+    envInt('PEARDOCK_STATS_LIST_INTERVAL_MS', CONFIG.STATS?.LIST_INTERVAL_MS ?? 15_000),
+    STATS_LIMITS.listIntervalMs.min,
+    STATS_LIMITS.listIntervalMs.max,
+    15_000
   )
   const cacheTtlMs = clampInt(
-    envInt('PEARDOCK_STATS_CACHE_TTL_MS', CONFIG.STATS?.CACHE_TTL_MS ?? 1000),
+    envInt('PEARDOCK_STATS_CACHE_TTL_MS', CONFIG.STATS?.CACHE_TTL_MS ?? 2000),
     STATS_LIMITS.cacheTtlMs.min,
     STATS_LIMITS.cacheTtlMs.max,
-    1000
+    2000
   )
   const warnContainers = clampInt(
     envInt('PEARDOCK_STATS_WARN_CONTAINERS', 80),
@@ -91,6 +111,8 @@ export function defaultStatsConfig() {
     collectIntervalMs,
     broadcastIntervalMs,
     concurrency,
+    samplesPerTick,
+    listIntervalMs,
     cacheTtlMs,
     warnContainers,
   }
@@ -120,6 +142,8 @@ function saveStatsDisk() {
       collectIntervalMs: runtime.collectIntervalMs,
       broadcastIntervalMs: runtime.broadcastIntervalMs,
       concurrency: runtime.concurrency,
+      samplesPerTick: runtime.samplesPerTick,
+      listIntervalMs: runtime.listIntervalMs,
       cacheTtlMs: runtime.cacheTtlMs,
       warnContainers: runtime.warnContainers,
     }
@@ -138,7 +162,7 @@ function saveStatsDisk() {
  */
 export function normalizeStatsConfig(partial = {}, base = defaultStatsConfig()) {
   const src = partial && typeof partial === 'object' ? partial : {}
-  // Accept seconds from UI as collectIntervalSec / broadcastIntervalSec
+  // Accept seconds from UI as collectIntervalSec / broadcastIntervalSec / listIntervalSec
   let collectMs = src.collectIntervalMs
   if (collectMs == null && src.collectIntervalSec != null) {
     collectMs = Number(src.collectIntervalSec) * 1000
@@ -147,6 +171,10 @@ export function normalizeStatsConfig(partial = {}, base = defaultStatsConfig())
   if (broadcastMs == null && src.broadcastIntervalSec != null) {
     broadcastMs = Number(src.broadcastIntervalSec) * 1000
   }
+  let listMs = src.listIntervalMs
+  if (listMs == null && src.listIntervalSec != null) {
+    listMs = Number(src.listIntervalSec) * 1000
+  }
   return {
     collectIntervalMs: clampInt(
       collectMs ?? base.collectIntervalMs,
@@ -166,6 +194,18 @@ export function normalizeStatsConfig(partial = {}, base = defaultStatsConfig())
       STATS_LIMITS.concurrency.max,
       base.concurrency
     ),
+    samplesPerTick: clampInt(
+      src.samplesPerTick ?? base.samplesPerTick,
+      STATS_LIMITS.samplesPerTick.min,
+      STATS_LIMITS.samplesPerTick.max,
+      base.samplesPerTick
+    ),
+    listIntervalMs: clampInt(
+      listMs ?? base.listIntervalMs,
+      STATS_LIMITS.listIntervalMs.min,
+      STATS_LIMITS.listIntervalMs.max,
+      base.listIntervalMs
+    ),
     cacheTtlMs: clampInt(
       src.cacheTtlMs ?? base.cacheTtlMs,
       STATS_LIMITS.cacheTtlMs.min,
@@ -210,13 +250,17 @@ export function getStatsConfig() {
     collectIntervalMs: runtime.collectIntervalMs,
     broadcastIntervalMs: runtime.broadcastIntervalMs,
     concurrency: runtime.concurrency,
+    samplesPerTick: runtime.samplesPerTick,
+    listIntervalMs: runtime.listIntervalMs,
     cacheTtlMs: runtime.cacheTtlMs,
     warnContainers: runtime.warnContainers,
     // Convenience for UI
     collectIntervalSec: Math.round(runtime.collectIntervalMs / 1000),
     broadcastIntervalSec: Math.round(runtime.broadcastIntervalMs / 1000),
-    active: Boolean(intervalHandle),
+    listIntervalSec: Math.round(runtime.listIntervalMs / 1000),
+    active: collectionLoopActive,
     peers: peers.size,
+    runningTracked: rosterRunning.length,
     limits: { ...STATS_LIMITS },
     path: STATS_FILE,
   }
@@ -232,35 +276,21 @@ export function updateStatsConfig(partial = {}, opts = {}) {
   ensureStatsConfigLoaded()
   const base = opts.replace ? defaultStatsConfig() : { ...runtime }
   const next = normalizeStatsConfig(partial, base)
-  const intervalChanged = next.collectIntervalMs !== runtime.collectIntervalMs
   runtime = next
   if (opts.persist !== false) saveStatsDisk()
-  if (intervalChanged && intervalHandle) {
-    rearmCollectInterval()
-  }
+  // Interval is applied on the next schedule after a tick completes
   logger.info('Stats config updated', {
     collectIntervalMs: runtime.collectIntervalMs,
     broadcastIntervalMs: runtime.broadcastIntervalMs,
     concurrency: runtime.concurrency,
+    samplesPerTick: runtime.samplesPerTick,
+    listIntervalMs: runtime.listIntervalMs,
     cacheTtlMs: runtime.cacheTtlMs,
     persist: opts.persist !== false,
   })
   return getStatsConfig()
 }
 
-/** Rearm setInterval after collect interval change (while peers connected). */
-function rearmCollectInterval() {
-  if (!intervalHandle) return
-  clearInterval(intervalHandle)
-  intervalHandle = setInterval(() => {
-    statsTick().catch(() => {})
-  }, runtime.collectIntervalMs)
-  if (typeof intervalHandle.unref === 'function') intervalHandle.unref()
-  logger.debug('Stats collect interval rearmed', {
-    intervalMs: runtime.collectIntervalMs,
-  })
-}
-
 export function getStatsFilePath() {
   return STATS_FILE
 }
@@ -270,7 +300,10 @@ const containerStats = {}
 const statsCache = new Map()
 const containerActivity = new Map()
 
-let intervalHandle = null
+/** @type {ReturnType<typeof setTimeout>|null} */
+let loopTimer = null
+/** True while peer-connected collection loop is armed */
+let collectionLoopActive = false
 /** Peer-registry unsubscribe; set while stats service is armed */
 let unsubPeers = null
 /** True after startStatsBroadcast(); false after stopStatsBroadcast() */
@@ -282,6 +315,14 @@ let dockerBackoffUntil = 0
 let tickInFlight = false
 let largeFleetWarned = false
 
+/** Cached running containers from last listContainers */
+let rosterRunning = []
+/** @type {Set<string>} */
+let rosterAllIds = new Set()
+let rosterFetchedAt = 0
+/** Round-robin cursor into rosterRunning */
+let rrIndex = 0
+
 /**
  * Docker Engine CPU % (same idea as `docker stats`).
  * One-shot stats populate precpu_stats after ~1s wait inside the engine.
@@ -580,30 +621,41 @@ export function destroyStatsForContainer(id) {
 }
 
 /**
- * List once, sample running containers with bounded one-shot stats.
- * No perpetual streams — dockerd only does work while we request samples.
+ * Refresh container roster from Docker (throttled).
+ * @param {boolean} [force]
  */
-async function collectContainerStats() {
-  // Single list (all) — derive running from State; was 2× listContainers/tick
+async function refreshRoster(force = false) {
+  const now = Date.now()
+  if (!force && rosterFetchedAt && now - rosterFetchedAt < runtime.listIntervalMs) {
+    return
+  }
+
   const all = await docker.listContainers({ all: true })
-  const running = all.filter((c) => String(c.State || '').toLowerCase() === 'running')
+  const running = all
+    .filter((c) => String(c.State || '').toLowerCase() === 'running')
+    .sort((a, b) => String(a.Id).localeCompare(String(b.Id)))
   const allIds = new Set(all.map((c) => c.Id))
   const runningIds = new Set(running.map((c) => c.Id))
 
+  rosterRunning = running
+  rosterAllIds = allIds
+  rosterFetchedAt = now
+  if (rrIndex >= running.length) rrIndex = 0
+
   if (running.length >= runtime.warnContainers && !largeFleetWarned) {
     largeFleetWarned = true
     logger.warn(
-      'Large running fleet for stats collection — dockerd load scales with container count',
+      'Large running fleet for stats collection — dockerd load scales with samples/tick',
       {
         running: running.length,
+        samplesPerTick: runtime.samplesPerTick,
         concurrency: runtime.concurrency,
         intervalMs: runtime.collectIntervalMs,
-        hint: 'Settings → Performance, or PEARDOCK_STATS_INTERVAL_MS / PEARDOCK_STATS_CONCURRENCY',
+        hint: 'Settings → Performance: raise interval, lower samples/tick & concurrency',
       }
     )
   }
 
-  // Ensure entries + drop gone / zero-out stopped
   for (const containerInfo of running) {
     ensureStatsEntry(containerInfo)
   }
@@ -625,9 +677,35 @@ async function collectContainerStats() {
       }
     }
   }
+}
 
-  // Bounded parallel one-shot samples (only running)
-  await mapPool(running, runtime.concurrency, async (containerInfo) => {
+/**
+ * Pick next round-robin batch of running containers (at most samplesPerTick).
+ * @param {object[]} running
+ * @returns {object[]}
+ */
+function nextSampleBatch(running) {
+  if (!running.length) return []
+  const n = Math.min(runtime.samplesPerTick, running.length)
+  const batch = []
+  for (let i = 0; i < n; i++) {
+    batch.push(running[(rrIndex + i) % running.length])
+  }
+  rrIndex = (rrIndex + n) % running.length
+  return batch
+}
+
+/**
+ * Throttled roster + round-robin one-shot samples.
+ * Dockerd cost ≈ samplesPerTick × ~1s / concurrency per tick, then idle for interval.
+ */
+async function collectContainerStats() {
+  await refreshRoster(false)
+
+  const batch = nextSampleBatch(rosterRunning)
+  if (!batch.length) return
+
+  await mapPool(batch, runtime.concurrency, async (containerInfo) => {
     if (peers.size === 0) return
     const statsData = containerStats[containerInfo.Id]
     if (!statsData) return
@@ -654,17 +732,26 @@ function destroyAllStatsEntries() {
   statsCache.clear()
   containerActivity.clear()
   largeFleetWarned = false
+  rosterRunning = []
+  rosterAllIds = new Set()
+  rosterFetchedAt = 0
+  rrIndex = 0
+}
+
+function clearLoopTimer() {
+  if (loopTimer) {
+    clearTimeout(loopTimer)
+    loopTimer = null
+  }
 }
 
 /**
- * Pause collection: stop interval and drop in-memory state.
+ * Pause collection: stop loop and drop in-memory state.
  * Safe to call when already paused.
  */
 export function pauseStatsCollection() {
-  if (intervalHandle) {
-    clearInterval(intervalHandle)
-    intervalHandle = null
-  }
+  collectionLoopActive = false
+  clearLoopTimer()
   destroyAllStatsEntries()
   tickInFlight = false
   logger.debug('Stats collection paused (no peers)')
@@ -769,6 +856,33 @@ async function statsTick() {
   }
 }
 
+/**
+ * Run one stats tick, then schedule the next after collectIntervalMs.
+ * Spacing is measured from tick *completion* so long sweeps cannot overlap.
+ */
+function runLoopTick() {
+  if (!collectionLoopActive || !serviceArmed || peers.size === 0) {
+    collectionLoopActive = false
+    clearLoopTimer()
+    return
+  }
+  statsTick()
+    .catch(() => {})
+    .finally(() => {
+      if (!collectionLoopActive || !serviceArmed || peers.size === 0) {
+        collectionLoopActive = false
+        clearLoopTimer()
+        return
+      }
+      clearLoopTimer()
+      loopTimer = setTimeout(() => {
+        loopTimer = null
+        runLoopTick()
+      }, runtime.collectIntervalMs)
+      if (typeof loopTimer.unref === 'function') loopTimer.unref()
+    })
+}
+
 /**
  * Resume collection when at least one peer is connected.
  * Idempotent; kicks an immediate tick so first client is not waiting a full interval.
@@ -776,20 +890,19 @@ async function statsTick() {
 export function resumeStatsCollection() {
   if (!serviceArmed) return
   if (peers.size === 0) return
-  if (!intervalHandle) {
-    intervalHandle = setInterval(() => {
-      statsTick().catch(() => {})
-    }, runtime.collectIntervalMs)
-    if (typeof intervalHandle.unref === 'function') intervalHandle.unref()
-    logger.debug('Stats collection resumed', {
-      peers: peers.size,
-      intervalMs: runtime.collectIntervalMs,
-      concurrency: runtime.concurrency,
-    })
-  }
-  // Immediate kick so first connected client gets samples ASAP
+  if (collectionLoopActive) return
+  collectionLoopActive = true
   lastBroadcast = 0
-  statsTick().catch(() => {})
+  logger.debug('Stats collection resumed', {
+    peers: peers.size,
+    intervalMs: runtime.collectIntervalMs,
+    concurrency: runtime.concurrency,
+    samplesPerTick: runtime.samplesPerTick,
+    listIntervalMs: runtime.listIntervalMs,
+  })
+  // Force fresh roster on first tick after resume
+  rosterFetchedAt = 0
+  runLoopTick()
 }
 
 function onPeerCountChange(size, prevSize) {
@@ -842,9 +955,9 @@ export function stopStatsBroadcast() {
   pauseStatsCollection()
 }
 
-/** @returns {boolean} whether the collect interval is currently running */
+/** @returns {boolean} whether the collect loop is currently armed */
 export function isStatsCollectionActive() {
-  return Boolean(intervalHandle)
+  return collectionLoopActive
 }
 
 /** Test/ops helpers — same shape as getStatsConfig subset */
@@ -854,6 +967,8 @@ export function getStatsRuntimeConfig() {
     collectIntervalMs: runtime.collectIntervalMs,
     broadcastIntervalMs: runtime.broadcastIntervalMs,
     concurrency: runtime.concurrency,
+    samplesPerTick: runtime.samplesPerTick,
+    listIntervalMs: runtime.listIntervalMs,
     cacheTtlMs: runtime.cacheTtlMs,
     warnContainers: runtime.warnContainers,
   }

test/alerts.test.js

@@ -90,6 +90,49 @@ test('alerts update global settings persists', async (t) => {
   mod2.stopAlerts()
 })
 
+test('alerts poll interval clamps to min 60s', async (t) => {
+  withTempAlerts(t)
+  const mod = await import('../server/services/alerts.js?' + Date.now() + 'pollclamp')
+  mod.startAlerts()
+  mod.updateAlertsConfig({ pollIntervalMs: 5_000 })
+  t.is(mod.getAlertsConfig().pollIntervalMs, 60_000, 'floor is 60s')
+  mod.updateAlertsConfig({ pollIntervalMs: 180_000 })
+  t.is(mod.getAlertsConfig().pollIntervalMs, 180_000)
+  mod.stopAlerts()
+})
+
+test('alerts ignore noisy docker events (exec/attach)', async (t) => {
+  withTempAlerts(t)
+  const mod = await import('../server/services/alerts.js?' + Date.now() + 'noisy')
+  mod.startAlerts()
+  // Fire noisy events — should not throw and should not enqueue work
+  mod.onDockerEvent({
+    Type: 'container',
+    Action: 'exec_create',
+    id: 'abc',
+    Actor: { Attributes: { name: 'x' } },
+  })
+  mod.onDockerEvent({
+    Type: 'container',
+    Action: 'attach',
+    id: 'abc',
+    Actor: { Attributes: { name: 'x' } },
+  })
+  const st = mod.getAlertsStatus()
+  t.is(st.eventQueueDepth, 0, 'noisy events dropped before queue')
+  // Relevant event should queue / process
+  mod.onDockerEvent({
+    Type: 'container',
+    Action: 'die',
+    id: 'dead1',
+    Actor: { Attributes: { name: 'boom' } },
+  })
+  // Allow microtask drain
+  await new Promise((r) => setTimeout(r, 30))
+  t.ok(mod.listAlertHistory(5).length >= 1, 'die event fires alert history')
+  mod.stopAlerts()
+})
+
 test('protocol exposes alert methods and push', async (t) => {
   const { Methods, MethodRoles, Pushes, Roles } = await import('../shared/protocol.js')
   t.is(Methods.getAlertsConfig, 'getAlertsConfig')

test/stats-config.test.js

@@ -17,12 +17,16 @@ test('normalizeStatsConfig clamps out-of-range values', (t) => {
     collectIntervalMs: 50,
     broadcastIntervalMs: 999_999,
     concurrency: 100,
+    samplesPerTick: 0,
+    listIntervalMs: 100,
     cacheTtlMs: -5,
     warnContainers: 1,
   })
   t.is(n.collectIntervalMs, STATS_LIMITS.collectIntervalMs.min)
   t.is(n.broadcastIntervalMs, STATS_LIMITS.broadcastIntervalMs.max)
   t.is(n.concurrency, STATS_LIMITS.concurrency.max)
+  t.is(n.samplesPerTick, STATS_LIMITS.samplesPerTick.min)
+  t.is(n.listIntervalMs, STATS_LIMITS.listIntervalMs.min)
   t.is(n.cacheTtlMs, STATS_LIMITS.cacheTtlMs.min)
   t.is(n.warnContainers, STATS_LIMITS.warnContainers.min)
 })
@@ -31,11 +35,15 @@ test('normalizeStatsConfig accepts seconds aliases from UI', (t) => {
   const n = normalizeStatsConfig({
     collectIntervalSec: 5,
     broadcastIntervalSec: 3,
+    listIntervalSec: 20,
     concurrency: 4,
+    samplesPerTick: 8,
   })
   t.is(n.collectIntervalMs, 5000)
   t.is(n.broadcastIntervalMs, 3000)
+  t.is(n.listIntervalMs, 20000)
   t.is(n.concurrency, 4)
+  t.is(n.samplesPerTick, 8)
 })
 
 test('defaultStatsConfig is within limits', (t) => {

ui/alerts-settings.js

@@ -100,7 +100,7 @@ function renderAlertsGlobal(config, status) {
     const min = document.getElementById('alerts-min-severity')
     if (min) min.value = config.minSeverity || 'info'
     const poll = document.getElementById('alerts-poll-ms')
-    if (poll) poll.value = String(Math.round((config.pollIntervalMs || 60000) / 1000))
+    if (poll) poll.value = String(Math.round((config.pollIntervalMs || 120000) / 1000))
     const rate = document.getElementById('alerts-rate-limit')
     if (rate) rate.value = String(config.rateLimitPerMinute || 40)
     const qh = document.getElementById('alerts-quiet-enabled')
@@ -120,10 +120,14 @@ function renderAlertsGlobal(config, status) {
 
   const statusEl = document.getElementById('alerts-engine-status')
   if (statusEl && status) {
+    const pollLabel = status.pollActive
+      ? `poll ${Math.round((status.pollIntervalMs || 0) / 1000)}s`
+      : 'poll off'
     const parts = [
       status.enabled ? 'Enabled' : 'Disabled',
       `${status.channelCount} channel(s)`,
       `${status.ruleCount} rule(s)`,
+      pollLabel,
       status.lastDaemonOk ? 'Docker OK' : 'Docker DOWN',
       status.quietHoursActive ? 'Quiet hours' : null,
     ].filter(Boolean)
@@ -373,7 +377,7 @@ function readGlobalConfigFromUi() {
     enabled: document.getElementById('alerts-enabled')?.value !== '0',
     minSeverity: document.getElementById('alerts-min-severity')?.value || 'info',
     pollIntervalMs:
-      Math.max(15, Number(document.getElementById('alerts-poll-ms')?.value) || 60) * 1000,
+      Math.max(60, Number(document.getElementById('alerts-poll-ms')?.value) || 120) * 1000,
     rateLimitPerMinute: Number(document.getElementById('alerts-rate-limit')?.value) || 40,
     includeHostname: document.getElementById('alerts-include-hostname')?.value !== '0',
     quietHours: {

ui/ops-app.js

@@ -2518,7 +2518,9 @@ function isServerStatsLiveField(el) {
   return (
     id === 'stats-collect-sec' ||
     id === 'stats-broadcast-sec' ||
+    id === 'stats-samples-per-tick' ||
     id === 'stats-concurrency' ||
+    id === 'stats-list-sec' ||
     id === 'stats-cache-ttl-ms' ||
     id === 'stats-warn-containers'
   )

ui/stats-settings.js

@@ -75,34 +75,44 @@ function renderStatsForm(config) {
     const collect = document.getElementById('stats-collect-sec')
     if (collect) {
       collect.value = String(
-        config.collectIntervalSec ?? Math.round((config.collectIntervalMs || 2000) / 1000)
+        config.collectIntervalSec ?? Math.round((config.collectIntervalMs || 5000) / 1000)
       )
     }
     const broadcast = document.getElementById('stats-broadcast-sec')
     if (broadcast) {
       broadcast.value = String(
-        config.broadcastIntervalSec ?? Math.round((config.broadcastIntervalMs || 2000) / 1000)
+        config.broadcastIntervalSec ?? Math.round((config.broadcastIntervalMs || 5000) / 1000)
       )
     }
+    const samples = document.getElementById('stats-samples-per-tick')
+    if (samples) samples.value = String(config.samplesPerTick ?? 4)
     const conc = document.getElementById('stats-concurrency')
-    if (conc) conc.value = String(config.concurrency ?? 6)
+    if (conc) conc.value = String(config.concurrency ?? 2)
+    const listSec = document.getElementById('stats-list-sec')
+    if (listSec) {
+      listSec.value = String(
+        config.listIntervalSec ?? Math.round((config.listIntervalMs || 15000) / 1000)
+      )
+    }
     const cache = document.getElementById('stats-cache-ttl-ms')
-    if (cache) cache.value = String(config.cacheTtlMs ?? 1000)
+    if (cache) cache.value = String(config.cacheTtlMs ?? 2000)
     const warn = document.getElementById('stats-warn-containers')
     if (warn) warn.value = String(config.warnContainers ?? 80)
 
     const active = config.active ? 'collecting' : 'idle (no peers or paused)'
     setEngineStatus(
-      `${active} · every ${config.collectIntervalSec ?? '?'}s · ×${config.concurrency ?? '?'}`
+      `${active} · gap ${config.collectIntervalSec ?? '?'}s · ${config.samplesPerTick ?? '?'} samples · ×${config.concurrency ?? '?'}`
     )
     const meta = document.getElementById('stats-runtime-meta')
     if (meta) {
       const parts = [
         `peers=${config.peers ?? 0}`,
-        `collect=${config.collectIntervalMs}ms`,
-        `broadcast=${config.broadcastIntervalMs}ms`,
+        `running=${config.runningTracked ?? '?'}`,
+        `gap=${config.collectIntervalMs}ms`,
+        `samples/tick=${config.samplesPerTick}`,
         `concurrency=${config.concurrency}`,
-        `cacheTtl=${config.cacheTtlMs}ms`,
+        `list=${config.listIntervalMs}ms`,
+        `broadcast=${config.broadcastIntervalMs}ms`,
       ]
       if (config.path) parts.push(`file=${config.path}`)
       meta.textContent = parts.join(' · ')
@@ -117,13 +127,17 @@ function renderStatsForm(config) {
 function readFormPartial() {
   const collectSec = Number(document.getElementById('stats-collect-sec')?.value)
   const broadcastSec = Number(document.getElementById('stats-broadcast-sec')?.value)
+  const samplesPerTick = Number(document.getElementById('stats-samples-per-tick')?.value)
   const concurrency = Number(document.getElementById('stats-concurrency')?.value)
+  const listSec = Number(document.getElementById('stats-list-sec')?.value)
   const cacheTtlMs = Number(document.getElementById('stats-cache-ttl-ms')?.value)
   const warnContainers = Number(document.getElementById('stats-warn-containers')?.value)
   return {
     collectIntervalSec: Number.isFinite(collectSec) ? collectSec : undefined,
     broadcastIntervalSec: Number.isFinite(broadcastSec) ? broadcastSec : undefined,
+    samplesPerTick: Number.isFinite(samplesPerTick) ? samplesPerTick : undefined,
     concurrency: Number.isFinite(concurrency) ? concurrency : undefined,
+    listIntervalSec: Number.isFinite(listSec) ? listSec : undefined,
     cacheTtlMs: Number.isFinite(cacheTtlMs) ? cacheTtlMs : undefined,
     warnContainers: Number.isFinite(warnContainers) ? warnContainers : undefined,
   }