version 5.1.20

_find_threshold() now returns the stripped prefix ("check_name") alongside
the ThresholdConfig, enabling a single generic entry (e.g. nagios_runner.status_code)
to cover all per-command metrics (check_disk_root_status_code, check_load_status_code,
…). The prefix is threaded through to _format_display() as {check_name}, with
{metric_name} also available in display templates. purge_stale_alerts() updated
to use generic matching so it does not incorrectly drop alerts on generic-matched
metrics. README updated with Display Format Templates and Generic Threshold
Matching sections.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

README.md

@@ -27,6 +27,7 @@ A lightweight daemon that listens for UDP heartbeat messages and acts on them: k
   - Configurable retention and backup management
 - **Plugin system for extensible monitoring** ✅
   - Collect system metrics (CPU, memory, disk, network)
+  - Monitor ZFS pool health, capacity, and I/O via `zpool(8)`
   - Execute existing Nagios monitoring plugins
   - Create custom plugins with simple Python classes
 - **Threshold alerting system** ✅
@@ -34,6 +35,8 @@ A lightweight daemon that listens for UDP heartbeat messages and acts on them: k
   - Hysteresis to prevent alert flapping
   - Automatic notifications on state changes
   - Re-notification for ongoing alerts
+- **Per-host watch flag** — set `watch: false` on any host to silence all notifications for that host without removing its configuration ✅
+- **Role-filtered dashboards** — Live Dashboard and Host Overview show only hosts where the logged-in user is owner or manager (admins see all) ✅
 - Modular codebase suitable for unit testing and CI ✅
 
 ---
@@ -61,12 +64,16 @@ Heartbeat includes a comprehensive plugin architecture that extends monitoring b
 - `network_monitor`: Monitors network interface statistics, bandwidth, and connections
 - `filesystem_info`: Collects mounted filesystem information (physical filesystems only by default)
 - `nagios_runner`: Executes Nagios monitoring plugins (check_disk, check_load, check_http, etc.)
+- `zfs_monitor`: Monitors ZFS pool health, capacity, fragmentation, dedup ratio, and cumulative I/O via `zpool(8)`
 
 ### Nagios Integration
 
 The `nagios_runner` plugin provides seamless integration with the vast Nagios plugin ecosystem. You can run any Nagios-compatible plugin and have the results automatically parsed and stored:
 
-- Executes plugins via subprocess with timeout protection
+- Executes plugins asynchronously (non-blocking) with timeout protection
+- Captures both stdout and stderr; if stdout is empty, stderr is used as the status message
+- Handles signal-killed processes (negative exit code → UNKNOWN status)
+- Validates absolute command paths at startup and warns on missing or non-executable files
 - Parses exit codes (OK/WARNING/CRITICAL/UNKNOWN)
 - Extracts performance data with thresholds
 - Reports aggregated status across all configured checks
@@ -147,9 +154,11 @@ Heartbeat includes a sophisticated threshold alerting system that monitors plugi
 - **Multi-level alerts**: WARNING and CRITICAL severity levels
 - **Flexible operators**: Support for >, >=, <, <=, ==, != comparisons
 - **Hysteresis**: Prevents alert flapping with configurable recovery thresholds
-- **Smart notifications**: Alerts only on state changes, not every check
+- **Smart notifications**: Alerts only on state changes, not every check; de-escalations (e.g. CRITICAL → WARNING) do not generate a notification
 - **Re-notifications**: Periodic reminders for ongoing alerts
+- **Short-duration suppression**: Recovery notifications are suppressed for down events under 4 seconds (avoids noise from transient blips)
 - **Journal integration**: All threshold events logged for audit trail
+- **`ping_monitor` thresholds**: Latency and packet-loss thresholds use the same format as all other plugin metrics
 
 ### Configuration
 
@@ -172,7 +181,8 @@ thresholds:
       warning: 80.0      # Warn when CPU > 80%
       critical: 90.0     # Critical when CPU > 90%
       operator: ">"
-      hysteresis: 0.1    # 10% hysteresis to prevent flapping
+      hysteresis: 0.02   # 2% hysteresis to prevent flapping
+      display: "(threshold: {op_symbol} {threshold_value}%)"  # optional
   
   memory_monitor:
     percent:
@@ -265,7 +275,59 @@ All plugin metrics can be thresholded:
 - **Memory**: percent, available_mb, swap_percent
 - **Disk**: Per-partition percent, free_gb, free_mb
 - **Network**: errors_total, dropped packets, connection counts
-- **Nagios**: exit_code mapping (0=OK, 1=WARNING, 2=CRITICAL)
+- **Nagios**: Any field emitted by `nagios_runner` (status_code, exit_code, performance data, …)
+
+### Display Format Templates
+
+Each threshold entry accepts an optional `display` field — a Python format string shown in notifications and on the Alerts dashboard:
+
+```yaml
+nagios_runner:
+  status_code:
+    warning: 1
+    critical: 2
+    operator: ">="
+    display: "{check_name}: exit {value} (expected < {threshold_value})"
+```
+
+Available variables:
+
+| Variable | Description |
+|---|---|
+| `{value}` | Current metric value |
+| `{threshold_value}` | Threshold that was crossed |
+| `{op_symbol}` | Comparison operator (`>`, `<`, `>=`, …) |
+| `{check_name}` | Prefix stripped by generic matching (see below) |
+| `{metric_name}` | Full field name within the plugin data |
+| any plugin field | Any other field present in the plugin's data |
+
+### Generic Threshold Matching
+
+When a metric name has no exact threshold entry, the server progressively strips leading underscore-separated segments and re-tries the lookup. This lets a single generic entry cover an entire family of metrics.
+
+The classic use case is `nagios_runner`, which names each metric after the command that produced it:
+
+```
+nagios_runner.check_disk_root_status_code    → no exact match
+nagios_runner.disk_root_status_code          → no match
+nagios_runner.root_status_code               → no match
+nagios_runner.status_code                    → matched ✓
+```
+
+Configure the generic threshold once:
+
+```yaml
+nagios_runner:
+  status_code:
+    warning: 1
+    critical: 2
+    operator: ">="
+    display: "{check_name}: exit {value}"
+```
+
+The stripped prefix (`check_disk_root` in the example above) is available as `{check_name}` in the display template, so you can identify which check triggered the alert without writing a separate threshold entry per command.
+
+Exact matches always take priority. A generic entry only applies when no specific one is defined.
 
 ### Per-Host Threshold Profiles
 
@@ -363,9 +425,10 @@ Heartbeat includes a built-in HTTP/WebSocket server that provides both a REST AP
 ### Web Dashboards
 
 - **Login** (`/login`): Browser login form (shown automatically when auth is configured)
-- **Live View** (`/live`): Real-time host connectivity, latency, and messages
-- **Plugin Metrics** (`/plugins`): Browse and visualize metrics from all plugins
-- **Alerts Dashboard** (`/alerts`): Monitor active alerts with severity filtering
+- **Live View** (`/live`): Real-time host connectivity, latency, and messages; hostnames link directly to the Host Overview page
+- **Host Overview** (`/plugins/<host>`): Per-host plugin metrics with ZFS pool visualization; filtered to hosts where the logged-in user is owner or manager (admins see all)
+- **Alerts Dashboard** (`/alerts`): Monitor active alerts with severity filtering; alert count pie chart shown in the navigation bar
+- **Settings** (`/settings`): Server configuration, user management, and threshold configuration viewer
 
 ### API Endpoints
 
@@ -476,6 +539,10 @@ plugins:
 
 All monitoring plugins default to 5-minute (300 second) intervals, but can be customized as needed.
 
+**Connection retry:** If a server is temporarily unreachable, `hbc` retries `open()` indefinitely on every heartbeat interval. IPv6 connections that never succeeded during early startup are dropped after 3 consecutive failures (to handle hosts without IPv6 routing), while IPv4 connections always retry.
+
+**Daemon logging:** When running with `-d`, `hbc` routes all log output to syslog (`LOG_DAEMON` facility) after daemonizing. Without `-d`, logs go to stderr as usual.
+
 ### hbc_mini — single-file client (no external dependencies)
 
 `scripts/hbc_mini.py` is a self-contained version of the heartbeat client that requires only Python 3.8+ and no external packages. Copy it to any host and run it directly — no virtualenv, no `pip install`.
@@ -531,8 +598,10 @@ python3 hbc_mini.py -m "maintenance starting" your-server.example.com
 
 - No YAML config (use JSON instead)
 - No `filesystem_info` plugin
+- No `zfs_monitor` plugin (requires `zpool(8)` and the full plugin loader)
 - `cpu_monitor` does not report per-core usage or CPU frequency (no psutil)
 - Plugins cannot be loaded from external `.py` files — all plugins are compiled in
+- No IPv6 early-fail protection — connections that fail to open at startup are silently skipped rather than retried
 
 Everything else — heartbeat protocol, ACK/CMD/UPD handling, `hb_install.sh`-based self-update, daemonize, syslog — is identical to the full client.
 

hbd/__init__.py

@@ -14,4 +14,4 @@ Install options:
 """
 
 __all__ = ["__version__"]
-__version__ = "5.1.14"
+__version__ = "5.1.20"

hbd/client/main.py

@@ -56,6 +56,8 @@ class AsyncConnection:
         self.transport: Optional[asyncio.DatagramTransport] = None
         self.protocol: Optional[asyncio.DatagramProtocol] = None
         self._dead = False
+        self._ever_opened = False
+        self._open_fail_count = 0   # consecutive failures before first success
 
         self.logger = logging.getLogger(f"hbc.conn.{addr}")
 
@@ -73,6 +75,7 @@ class AsyncConnection:
                 lambda: HeartbeatProtocol(self),
                 family=self.af
             )
+            self._ever_opened = True
             self.logger.debug(f"Opened connection to {self.addr}:{self.port}")
             return True
         except Exception as e:
@@ -262,15 +265,51 @@ async def handle_update(conn: AsyncConnection, _msg: dict):  # pyright: ignore[r
 
 
 async def heartbeat_sender(conn: AsyncConnection, interval: int):
-    """Send periodic heartbeats.
+    """Send periodic heartbeats, retrying the connection if it is not open.
+
+    IPv6 connections that fail to open before their first successful send are
+    dropped after IPV6_EARLY_FAIL_LIMIT attempts so that a network without IPv6
+    does not keep a dead sender alive.  IPv4 connections are retried indefinitely.
 
     Args:
         conn: Connection to send on
         interval: Heartbeat interval in seconds
     """
     logger = logging.getLogger("hbc.heartbeat")
+    IPV6_EARLY_FAIL_LIMIT = 3
+
+    while running and not conn._dead:
+        # Ensure transport is open before attempting to send.
+        if not conn.transport:
+            opened = await conn.open()
+            if opened:
+                conn._open_fail_count = 0
+            else:
+                conn._open_fail_count += 1
+                # Drop an IPv6 connection that has never come up within the
+                # first few attempts — it is likely unavailable on this network.
+                if (not conn._ever_opened
+                        and conn.af == socket.AF_INET6
+                        and conn._open_fail_count >= IPV6_EARLY_FAIL_LIMIT):
+                    logger.warning(
+                        f"IPv6 connection to {conn.addr} unreachable after "
+                        f"{conn._open_fail_count} attempts, disabling"
+                    )
+                    conn._dead = True
+                    break
+                # Retry after the normal interval; IPv4 retries forever.
+                try:
+                    if shutdown_event:
+                        await asyncio.wait_for(shutdown_event.wait(), timeout=interval)
+                        break
+                    else:
+                        await asyncio.sleep(interval)
+                except asyncio.TimeoutError:
+                    pass
+                except asyncio.CancelledError:
+                    raise
+                continue
 
-    while running:
         try:
             msg = {
                 "acks": conn.ackcount,
@@ -279,19 +318,16 @@ async def heartbeat_sender(conn: AsyncConnection, interval: int):
             }
             await conn.sendto(msg, "HTB")
 
-        except Exception as e:
-            logger.error(f"Error sending heartbeat: {e}", exc_info=True)
         except asyncio.CancelledError:
             logger.debug("Heartbeat sender cancelled")
             raise
+        except Exception as e:
+            logger.error(f"Error sending heartbeat: {e}", exc_info=True)
 
         # Wait for next interval or shutdown event
         try:
             if shutdown_event:
-                await asyncio.wait_for(
-                    shutdown_event.wait(), 
-                    timeout=interval
-                )
+                await asyncio.wait_for(shutdown_event.wait(), timeout=interval)
                 break
             else:
                 await asyncio.sleep(interval)
@@ -427,16 +463,13 @@ async def cleanup(connections: List[AsyncConnection]):
     logger = logging.getLogger("hbc.cleanup")
     logger.info("Cleaning up connections")
     
-    for conn in connections:
+    target = next((c for c in connections if c.transport), connections[0] if connections else None)
+    if target:
         try:
-            msg = {
-                "shutdown": 1,
-                "acks": conn.ackcount
-            }
-            await conn.sendto(msg)
+            await target.sendto({"shutdown": 1, "acks": target.ackcount})
         except Exception as e:
             logger.error(f"Error sending shutdown: {e}")
-        
+    for conn in connections:
         conn.close()
     
     # Give messages time to send
@@ -481,12 +514,13 @@ async def async_main(args, config):
             addr = addr_info[4][0]
 
             conn = AsyncConnection(conn_id, addr, hb_port, af, iam)
-            if await conn.open():
+            if not await conn.open():
+                logger.warning(f"Initial open to {addr} failed, heartbeat sender will retry")
             connections.append(conn)
             conn_id += 1
 
     if not connections:
-        logger.error("No connections established")
+        logger.error("No connections established (DNS resolution failed for all hosts)")
         return 1
     
     logger.info(f"Created {len(connections)} connections")
@@ -501,8 +535,8 @@ async def async_main(args, config):
             boot_msg["msg"] = args.message
         
         boot_msg["acks"] = 0
-        for conn in connections:
-            await conn.sendto(boot_msg)
+        target = next((c for c in connections if c.transport), connections[0])
+        await target.sendto(boot_msg)
         
         if args.message and not args.daemon:
             # Message-only mode
@@ -702,7 +736,7 @@ def main(argv=None):
     
     # Daemonize if requested
     if args.daemon:
-        print("Daemonizing...")
+        logging.info("Daemonizing...")
         daemonize()
         _reconfigure_logging_for_daemon(log_level)
         logging.info(f"hbc starting, sending heartbeat to {', '.join(args.hosts)}")

hbd/client/plugins/cpu_monitor.py

@@ -119,6 +119,13 @@ class CPUMonitorPlugin(MonitorPlugin):
             except Exception as e:
                 self.logger.debug(f"Could not get CPU times: {e}")
 
+            # Uptime in seconds
+            try:
+                import time
+                data["uptime_seconds"] = int(time.time() - self.psutil.boot_time())
+            except Exception as e:
+                self.logger.debug(f"Could not get uptime: {e}")
+            
             self.logger.debug(
                 f"Collected CPU metrics: {data.get('cpu_percent', 'N/A')}% usage"
             )

hbd/client/plugins/memory_monitor.py

@@ -14,6 +14,24 @@ except ImportError:
 
 from hbd.client.plugin import MonitorPlugin
 
+
+def _zfs_arc_bytes() -> int:
+    """Return current ZFS ARC size in bytes, or 0 if ZFS is not present.
+
+    ZFS ARC is reclaimable but is not included in MemAvailable by the Linux
+    kernel (it is not in SReclaimable), so it would otherwise be counted as
+    used memory.
+    """
+    try:
+        with open("/proc/spl/kstat/zfs/arcstats") as fh:
+            for line in fh:
+                parts = line.split()
+                if len(parts) >= 3 and parts[0] == "size":
+                    return int(parts[2])
+    except (OSError, ValueError):
+        pass
+    return 0
+
 logger = logging.getLogger(__name__)
 
 
@@ -101,11 +119,21 @@ class MemoryMonitorPlugin(MonitorPlugin):
         
         # Virtual (physical) memory statistics
         vmem = psutil.virtual_memory()
+
+        # psutil's available already excludes page cache / file buffers
+        # (uses MemAvailable on Linux). Add ZFS ARC on top because the kernel
+        # does not include it in SReclaimable / MemAvailable even though it is
+        # reclaimable.
+        arc_bytes = _zfs_arc_bytes()
+        available = min(vmem.available + arc_bytes, vmem.total)
+        used = vmem.total - available
+        percent = round(used / vmem.total * 100, 1) if vmem.total else 0.0
+
         metrics['memory_total'] = vmem.total
-        metrics['memory_available'] = vmem.available
-        metrics['memory_used'] = vmem.used
+        metrics['memory_available'] = available
+        metrics['memory_used'] = used
         metrics['memory_free'] = vmem.free
-        metrics['memory_percent'] = vmem.percent
+        metrics['memory_percent'] = percent
         
         # Platform-specific memory details
         if hasattr(vmem, 'active'):

hbd/server/hbdclass.py

@@ -95,7 +95,7 @@ class Connection:
         if not Null:
             d["addr"] = self.addr
             if self.rtts[-1]:
-                d["rtt"] = "%0.1f" % self.rtts[-1]
+                d["rtt"] = "%d" % round(self.rtts[-1])
             elif self.state == Connection.UNKNOWN:
                 d["rtt"] = ""
             else:

hbd/server/http.py

@@ -154,6 +154,25 @@ async def start(
         lst = [h.jsons() for h in hosts]
         return web.json_response(json.loads("[" + ",".join(lst) + "]"))
 
+    async def api_alert_summary(request):
+        """GET /api/0/alert_summary — counts of ok/warning/critical hosts visible to caller."""
+        user, err = _require_auth(request)
+        if err:
+            return err
+        from .threshold import AlertLevel
+        critical = warning = ok = 0
+        for host in hbdclass.Host.hosts.values():
+            if not _can_operate_host(user, host):
+                continue
+            levels = {s.level for s in host.alert_states.values()}
+            if AlertLevel.CRITICAL in levels:
+                critical += 1
+            elif AlertLevel.WARNING in levels:
+                warning += 1
+            else:
+                ok += 1
+        return web.json_response({"critical": critical, "warning": warning, "ok": ok})
+
     async def api_messages(request):
         lst = data.msgs[-30:]
         return web.json_response(lst)
@@ -518,6 +537,7 @@ async def start(
                 hosts_with_plugins.append({
                     "name": hostname,
                     "plugins": list(host.plugin_data.keys()),
+                    "is_owner": _can_own_host(current_user, host),
                 })
 
         tmpl = env.get_template("plugins.html")
@@ -870,7 +890,7 @@ async def start(
         tmpl = env.get_template("settings.html")
         body = tmpl.render(
             title="Settings - Heartbeat",
-            sections=settings_mod.get_settings_sections(config),
+            sections=settings_mod.get_settings_sections(config, threshold_checker=threshold_checker),
             current_user=current_user.to_dict() if current_user else None,
             active_page="settings",
         )
@@ -893,6 +913,7 @@ async def start(
             web.get("/api/0/users/{username}/avatar", api_user_avatar),
             # Hosts
             web.get("/api/0/hosts", api_hosts),
+            web.get("/api/0/alert_summary", api_alert_summary),
             web.get("/api/0/messages", api_messages),
             web.get("/api/0/hosts/{hostname}/plugins", api_host_plugins),
             web.get("/api/0/hosts/{hostname}/plugins/{plugin_name}", api_host_plugin_detail),

hbd/server/main.py

@@ -101,9 +101,10 @@ async def reload_configuration(config_obj, config_path, components):
             access = config_mod.get_host_access(new_config, hostname)
             host.apply_access(access["owner"], access["managers"], access["monitors"])
 
-        # Reload threshold checker
+        # Reload threshold checker and prune alerts orphaned by the new config
         if 'threshold_checker' in components:
             components['threshold_checker'].reload(new_config)
+            components['threshold_checker'].purge_stale_alerts(hbdclass)
         
         # Note: Changes to the following require restart:
         # - hb_port, hbd_port, ws_port (already bound)
@@ -241,6 +242,10 @@ async def _run_async(config, config_path=None):
     )
     udp.restore_connection_timers(hbdclass, restore_ctx)
 
+    # Drop alert states that no longer have a matching threshold (stale after
+    # upgrade or config change between runs).
+    threshold_checker.purge_stale_alerts(hbdclass)
+
     # HTTP server (asyncio-based via aiohttp)
     try:
         http_task = asyncio.create_task(
@@ -250,6 +255,7 @@ async def _run_async(config, config_path=None):
                 config=config,
                 hbdclass=hbdclass,
                 tcss=None,
+                threshold_checker=threshold_checker,
                 verbose=config.get("verbose", False),
                 get_now=lambda: time.time(),
                 VER="",

hbd/server/settings.py

@@ -88,7 +88,7 @@ def _sanitize_channel(name, cfg):
 # Public API
 # ---------------------------------------------------------------------------
 
-def get_settings_sections(config: dict) -> list:
+def get_settings_sections(config: dict, threshold_checker=None) -> list:
     """Return ordered list of setting sections for the settings page.
 
     Each section:
@@ -181,6 +181,41 @@ def get_settings_sections(config: dict) -> list:
             "notification_channels": attrs.get("notification_channels", []),
         })
 
+    # ---- Threshold configurations -----------------------------------------
+    def _tc_to_row(tc):
+        return {
+            "metric": tc.metric_path,
+            "operator": tc.operator.value,
+            "warning": tc.warning,
+            "critical": tc.critical,
+            "hysteresis": tc.hysteresis,
+            "count": tc.count,
+            "enabled": tc.enabled,
+        }
+
+    threshold_config_list = []
+    if threshold_checker is not None:
+        if threshold_checker.threshold_configs:
+            for cfg_name, cfg_metrics in sorted(threshold_checker.threshold_configs.items()):
+                # For the default config use the merged effective set;
+                # for named overrides use only the explicitly defined metrics
+                # (threshold_raw_configs) so inherited defaults are not repeated.
+                if cfg_name == "default":
+                    display_metrics = cfg_metrics
+                else:
+                    display_metrics = threshold_checker.threshold_raw_configs.get(cfg_name, cfg_metrics)
+                metrics = sorted(
+                    [_tc_to_row(tc) for tc in display_metrics.values()],
+                    key=lambda m: m["metric"],
+                )
+                threshold_config_list.append({"name": cfg_name, "metrics": metrics})
+        elif threshold_checker.thresholds:
+            metrics = sorted(
+                [_tc_to_row(tc) for tc in threshold_checker.thresholds.values()],
+                key=lambda m: m["metric"],
+            )
+            threshold_config_list.append({"name": "default", "metrics": metrics})
+
     # ---- Hosts summary ----------------------------------------------------
     hosts_list = []
     for hname, hcfg in (config.get("hosts") or {}).items():
@@ -312,6 +347,16 @@ def get_settings_sections(config: dict) -> list:
             "hosts": hosts_list,
             "fields": [],
         },
+        {
+            "id": "thresholds",
+            "title": "Threshold Configurations",
+            "description": "Named alert threshold sets. Each defines warning/critical levels per metric.",
+            "threshold_configs": threshold_config_list,
+            "fields": [
+                field("default_threshold_config", "Default config", "text",
+                      "Threshold config used for hosts with no explicit mapping."),
+            ],
+        },
         {
             "id": "runtime",
             "title": "Runtime",

hbd/server/templates/alerts.html

@@ -4,6 +4,11 @@
 
   <style>
 
+    body {
+      height: auto;
+      overflow-y: auto;
+    }
+
     .container {
       max-width: 1400px;
       margin: 0 auto;
@@ -400,6 +405,10 @@
         } else if (alert.threshold_value !== undefined && alert.threshold_value !== null && alert.operator) {
           valueText += ` <span class="threshold-info">(threshold: ${alert.operator} ${formatValue(alert.threshold_value)})</span>`;
         }
+        if (alert.recovery_threshold !== undefined && alert.recovery_threshold !== null) {
+          const recOp = (alert.operator === '>' || alert.operator === '>=') ? '<' : '>';
+          valueText += ` <span class="threshold-info" style="color:#888">(recovers ${recOp} ${formatValue(alert.recovery_threshold)})</span>`;
+        }
         
         // Build actions section
         let actionsHtml = '';

hbd/server/templates/head.html

@@ -126,11 +126,17 @@
       }
 
       /* Swiss railway clock — nav */
-      .nav-clock {
+      .nav-pie {
         flex-shrink: 0;
         line-height: 0;
         margin-left: auto;
         padding: 4px 4px 4px 0;
+      }
+      #alert-pie { display: block; cursor: default; }
+      .nav-clock {
+        flex-shrink: 0;
+        line-height: 0;
+        padding: 4px 4px 4px 0;
         cursor: pointer;
       }
       #swiss-clock { display: block; }

hbd/server/templates/live.html

@@ -236,6 +236,8 @@
       color: #ff9800;
       font-weight: 700;
     }
+    #ntable a.host-link { color: inherit; text-decoration: none; }
+    #ntable a.host-link:hover { text-decoration: underline; }
   </style>
   <script type="text/javascript">
     var cnt = 0;
@@ -245,11 +247,13 @@
     var HBD_VERSION = "{{ hbd_version }}";
 
     function hostNameHtml(data) {
+      var rawName = data.raw_name || data.name.replace(/<[^>]+>/g, '').replace('*', '').trim();
       var nameHtml = data.name;
       if (!data.hbc_version || data.hbc_version !== HBD_VERSION) {
         nameHtml += ' 🥀';
       }
-      return data.dyn ? '<b>' + nameHtml + '</b>' : nameHtml;
+      var display = data.dyn ? '<b>' + nameHtml + '</b>' : nameHtml;
+      return '<a class="host-link" href="/plugins#' + encodeURIComponent(rawName) + '">' + display + '</a>';
     }
 
     function setup() {
@@ -404,7 +408,7 @@
         );
         if (data.connections[i].state == "up") {
           state = '<span class="state-up">up</span>';
-          latency = Number.parseFloat(data.connections[i].rtts[0]).toFixed(2);
+          latency = String(Math.round(Number.parseFloat(data.connections[i].rtts[0])));
         } else {
           if (data.connections[i].state == "unknown") {
             state = "";
@@ -511,7 +515,7 @@
           <tbody id="ntablebody">
             {% for host in hosts %}
             <tr class="{% if host.alert_critical_unacked > 0 or host.alert_critical_acked > 0 %}row-critical{% elif host.alert_warning_unacked > 0 or host.alert_warning_acked > 0 %}row-warning{% endif %}">
-              <td data-name="{{ host.name }}">{{ host.name }}{% if not host.hbc_version or host.hbc_version != hbd_version %} 🥀{% endif %}</td>
+              <td data-name="{{ host.name }}"><a class="host-link" href="/plugins#{{ host.raw_name | urlencode }}">{{ host.name }}{% if not host.hbc_version or host.hbc_version != hbd_version %} 🥀{% endif %}</a></td>
               <td style="text-align: center; color: #ff9800; font-weight: bold;">
                 {%- set warning_unacked = host.alert_warning_unacked -%}
                 {%- set warning_acked = host.alert_warning_acked -%}

hbd/server/templates/nav.html

@@ -11,6 +11,9 @@
     {% endif %}
     <a href="/about"{% if active_page == "about" %} class="active"{% endif %}>About</a>
   </div>
+  <div class="nav-pie" title="Host alert status">
+    <canvas id="alert-pie" width="44" height="44"></canvas>
+  </div>
   <div class="nav-clock" title="Click for full-screen clock">
     <canvas id="swiss-clock" width="44" height="44"></canvas>
   </div>
@@ -42,4 +45,52 @@
       });
     }
   })();
+
+  function drawAlertPie(critical, warning, ok) {
+    var canvas = document.getElementById('alert-pie');
+    if (!canvas) return;
+    var ctx = canvas.getContext('2d');
+    var SIZE = canvas.width;
+    var R = SIZE / 2;
+    ctx.clearRect(0, 0, SIZE, SIZE);
+    var total = critical + warning + ok;
+    if (total === 0) {
+      ctx.beginPath();
+      ctx.arc(R, R, R - 1, 0, Math.PI * 2);
+      ctx.fillStyle = '#ccc';
+      ctx.fill();
+      return;
+    }
+    var slices = [
+      { value: critical, color: '#e53935' },
+      { value: warning,  color: '#ffb300' },
+      { value: ok,       color: '#43a047' }
+    ];
+    var start = -Math.PI / 2;
+    slices.forEach(function(s) {
+      if (s.value === 0) return;
+      var sweep = (s.value / total) * Math.PI * 2;
+      ctx.beginPath();
+      ctx.moveTo(R, R);
+      ctx.arc(R, R, R - 1, start, start + sweep);
+      ctx.closePath();
+      ctx.fillStyle = s.color;
+      ctx.fill();
+      start += sweep;
+    });
+  }
+
+  function updateAlertPie() {
+    fetch('/api/0/alert_summary').then(function(r) {
+      if (!r.ok) return;
+      return r.json();
+    }).then(function(d) {
+      if (d) drawAlertPie(d.critical || 0, d.warning || 0, d.ok || 0);
+    }).catch(function() {});
+  }
+
+  document.addEventListener('DOMContentLoaded', function() {
+    updateAlertPie();
+    setInterval(updateAlertPie, 30000);
+  });
 </script>

hbd/server/templates/plugins.html

@@ -131,6 +131,52 @@
       text-overflow: ellipsis;
     }
 
+    .host-action-btn {
+      font-size: 0.75em;
+      font-weight: bold;
+      padding: 3px 10px;
+      border-radius: 4px;
+      border: none;
+      cursor: pointer;
+      text-decoration: none;
+      white-space: nowrap;
+    }
+    .host-action-btn.update-btn {
+      background: #e3f2fd;
+      color: #1565c0;
+    }
+    .host-action-btn.update-btn:hover { background: #bbdefb; }
+    .host-action-btn.delete-btn {
+      background: #ffebee;
+      color: #c62828;
+    }
+    .host-action-btn.delete-btn:hover { background: #ffcdd2; }
+
+    /* ── Action result toast ───────────────────────────────────── */
+    #action-toast {
+      position: fixed;
+      bottom: 24px;
+      left: 50%;
+      transform: translateX(-50%) translateY(20px);
+      background: #323232;
+      color: #fff;
+      padding: 12px 22px;
+      border-radius: 6px;
+      font-size: 0.9em;
+      max-width: 480px;
+      text-align: center;
+      opacity: 0;
+      pointer-events: none;
+      transition: opacity 0.25s, transform 0.25s;
+      z-index: 9000;
+      white-space: pre-wrap;
+    }
+    #action-toast.show {
+      opacity: 1;
+      transform: translateX(-50%) translateY(0);
+    }
+    #action-toast.error { background: #c62828; }
+
     /* ── Host body ──────────────────────────────────────────────── */
 
     .host-body {
@@ -379,6 +425,12 @@
               <span class="nagios-badge" id="nagios-badge-{{ host.name }}">—</span>
               {% endif %}
               <span class="os-label" id="os-label-{{ host.name }}"></span>
+              {% if host.is_owner %}
+              <button class="host-action-btn update-btn"
+                      onclick="event.stopPropagation(); hostAction(this, '/u?h={{ host.name }}')">Update</button>
+              <button class="host-action-btn delete-btn"
+                      onclick="event.stopPropagation(); hostDelete(this, '{{ host.name }}')">Delete</button>
+              {% endif %}
             </div>
           </div>
 
@@ -1156,12 +1208,68 @@
       // ── Init ────────────────────────────────────────────────────────────────
 
       document.addEventListener('DOMContentLoaded', () => {
+        // If a host fragment is in the URL, expand and scroll to that host;
+        // otherwise expand the first host as before.
+        const hash = window.location.hash;
+        if (hash) {
+          const hostname = decodeURIComponent(hash.slice(1));
+          const card = document.querySelector(`.host-card[data-hostname="${hostname}"]`);
+          if (card) {
+            card.classList.remove('collapsed');
+            fetchHostGlance(hostname);
+            setTimeout(() => card.scrollIntoView({ behavior: 'smooth', block: 'start' }), 150);
+            return;
+          }
+        }
         const first = document.querySelector('.host-card');
         if (first) {
           first.classList.remove('collapsed');
           fetchHostGlance(first.dataset.hostname);
         }
       });
+      // ── Host action helpers ──────────────────────────────────────
+
+      let _toastTimer = null;
+      function showToast(msg, isError) {
+        const t = document.getElementById('action-toast');
+        t.textContent = msg;
+        t.classList.toggle('error', !!isError);
+        t.classList.add('show');
+        clearTimeout(_toastTimer);
+        _toastTimer = setTimeout(() => t.classList.remove('show'), 4000);
+      }
+
+      async function hostAction(btn, url) {
+        btn.disabled = true;
+        try {
+          const res = await fetch(url);
+          const text = await res.text();
+          showToast(text, !res.ok);
+        } catch (e) {
+          showToast('Request failed: ' + e.message, true);
+        } finally {
+          btn.disabled = false;
+        }
+      }
+
+      async function hostDelete(btn, hostname) {
+        if (!confirm('Delete host ' + hostname + '?')) return;
+        btn.disabled = true;
+        try {
+          const res = await fetch('/d?h=' + encodeURIComponent(hostname));
+          const text = await res.text();
+          showToast(text, !res.ok);
+          if (res.ok) {
+            const card = document.querySelector(`.host-card[data-hostname="${hostname}"]`);
+            if (card) card.remove();
+          }
+        } catch (e) {
+          showToast('Request failed: ' + e.message, true);
+          btn.disabled = false;
+        }
+      }
     </script>
+
+    <div id="action-toast"></div>
   </body>
 </html>

hbd/server/templates/settings.html

@@ -254,6 +254,17 @@
     .host-bool { text-align: center; }
     .dot-yes { color: #2e7d32; font-size: 1.1em; }
     .dot-no  { color: #ddd;    font-size: 1.1em; }
+
+    /* ---- Threshold configurations ---- */
+    .thresh-config { margin: 12px 20px 20px; }
+    .thresh-config-name {
+      font-weight: 600; font-size: 0.9em; color: #1a237e;
+      margin-bottom: 6px;
+    }
+    .mini-table .warn  { color: #e65100; font-weight: 600; }
+    .mini-table .crit  { color: #b71c1c; font-weight: 600; }
+    .mini-table .dim   { color: #aaa; }
+    .mini-table .metric-path { font-family: monospace; font-size: 0.88em; }
   </style>
 
   <body>
@@ -394,6 +405,49 @@
             {% endif %}
             {% endif %}
 
+            {# ---- Threshold configurations section ---- #}
+            {% if section.id == "thresholds" %}
+            {% if section.threshold_configs %}
+            {% for tc in section.threshold_configs %}
+            <div class="thresh-config">
+              <div class="thresh-config-name">{{ tc.name }}</div>
+              {% if tc.metrics %}
+              <div style="overflow-x: auto;">
+                <table class="mini-table">
+                  <thead>
+                    <tr>
+                      <th>Metric</th>
+                      <th>Op</th>
+                      <th>Warning</th>
+                      <th>Critical</th>
+                      <th>Hysteresis</th>
+                      <th>Count</th>
+                    </tr>
+                  </thead>
+                  <tbody>
+                    {% for m in tc.metrics %}
+                    <tr {% if not m.enabled %} style="opacity:0.45"{% endif %}>
+                      <td class="metric-path">{{ m.metric }}</td>
+                      <td>{{ m.operator or '>' }}</td>
+                      <td class="warn">{{ m.warning if m.warning is not none else '—' }}</td>
+                      <td class="crit">{{ m.critical if m.critical is not none else '—' }}</td>
+                      <td class="dim">{{ '%.0f%%' % (m.hysteresis * 100) if m.hysteresis else '—' }}</td>
+                      <td class="dim">{{ m.count }}</td>
+                    </tr>
+                    {% endfor %}
+                  </tbody>
+                </table>
+              </div>
+              {% else %}
+              <span class="val-empty">No thresholds defined.</span>
+              {% endif %}
+            </div>
+            {% endfor %}
+            {% else %}
+            <div class="field-row"><span class="val-empty">No threshold configurations defined.</span></div>
+            {% endif %}
+            {% endif %}
+
             {# ---- Hosts section ---- #}
             {% if section.id == "hosts" %}
             {% if section.hosts %}

hbd/server/threshold.py

@@ -57,6 +57,7 @@ class AlertState:
         self.last_notification = None
         self.threshold_value = None  # The threshold value that triggered alert
         self.operator = None  # The comparison operator (>, <, >=, etc.)
+        self.hysteresis: Optional[float] = None  # Hysteresis fraction used for recovery
         self.formatted_message = None  # Formatted display message for UI
         self.acknowledged = False  # Whether alert has been acknowledged
         self.acknowledged_at = None  # Timestamp when acknowledged
@@ -152,6 +153,15 @@ class AlertState:
         if self.formatted_message is not None:
             result["formatted_message"] = self.formatted_message
 
+        # Compute and expose the recovery threshold so the UI can display it
+        if (self.hysteresis and self.threshold_value is not None
+                and self.operator is not None):
+            ha = abs(self.threshold_value * self.hysteresis)
+            if self.operator in ('>', '>='):
+                result["recovery_threshold"] = round(self.threshold_value - ha, 4)
+            elif self.operator in ('<', '<='):
+                result["recovery_threshold"] = round(self.threshold_value + ha, 4)
+
         return result
     
     def __setstate__(self, state):
@@ -159,6 +169,8 @@ class AlertState:
         self.__dict__.update(state)
         if not hasattr(self, 'consecutive_count'):
             self.consecutive_count = 0
+        if not hasattr(self, 'hysteresis'):
+            self.hysteresis = None
 
     def acknowledge(self):
         """Acknowledge this alert to stop reminder notifications."""
@@ -546,7 +558,7 @@ class ThresholdChecker:
             critical = threshold_config.get("critical")
             operator = threshold_config.get("operator", ">")
             display = threshold_config.get("display", "(threshold: {op_symbol} {threshold_value})")
-            hysteresis = threshold_config.get("hysteresis", 0.1)  # 10% default
+            hysteresis = threshold_config.get("hysteresis", 0.02)  # 2% default
             enabled = threshold_config.get("enabled", True)
             
             if warning is None and critical is None:
@@ -649,7 +661,7 @@ class ThresholdChecker:
         warning = rtt_thresholds.get("warning")
         critical = rtt_thresholds.get("critical")
         operator = rtt_thresholds.get("operator", ">")
-        hysteresis = rtt_thresholds.get("hysteresis", 0.1)  # 10% default
+        hysteresis = rtt_thresholds.get("hysteresis", 0.02)  # 2% default
         enabled = rtt_thresholds.get("enabled", True)
         display = rtt_thresholds.get("display")
         count = rtt_thresholds.get("count", 1)
@@ -794,6 +806,12 @@ class ThresholdChecker:
         elif new_level == AlertLevel.WARNING and threshold.warning is not None:
             threshold_value = threshold.warning
 
+        # Keep hysteresis on the state so the UI can show the recovery threshold
+        if new_level != AlertLevel.OK:
+            alert_state.hysteresis = threshold.hysteresis
+        else:
+            alert_state.hysteresis = None
+
         # Update state and check for changes
         old_level = alert_state.level
         if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
@@ -803,6 +821,36 @@ class ThresholdChecker:
             self._check_pending_or_renotify(host_name, alert_state, metric_path, value, threshold, None)
 
         return None
+    def _find_threshold(
+        self, thresholds: Dict[str, "ThresholdConfig"], metric_path: str
+    ) -> Tuple[Optional["ThresholdConfig"], Optional[str]]:
+        """Return (threshold, check_name) for *metric_path*, falling back to suffix matches.
+
+        Allows generic thresholds like ``nagios_runner.status_code`` to match
+        fully-qualified paths like ``nagios_runner.check_disk_root_status_code``.
+        The exact match is always tried first; then successive leading
+        underscore-delimited segments are stripped from the field name until
+        a match is found or no segments remain.
+
+        Returns:
+            (ThresholdConfig, None) for an exact match.
+            (ThresholdConfig, "check_disk_root") for a suffix match — the second
+            element is the stripped prefix, available as ``{check_name}`` in
+            display format templates.
+            (None, None) when no threshold is found.
+        """
+        if metric_path in thresholds:
+            return thresholds[metric_path], None
+        plugin, sep, field = metric_path.partition(".")
+        if not sep:
+            return None, None
+        parts = field.split("_")
+        for i in range(1, len(parts)):
+            candidate = plugin + "." + "_".join(parts[i:])
+            if candidate in thresholds:
+                return thresholds[candidate], "_".join(parts[:i])
+        return None, None
+
     def check_plugin_data(
         self,
         host_name: str,
@@ -831,11 +879,10 @@ class ThresholdChecker:
         for metric_name, value in data.items():
             metric_path = f"{plugin_name}.{metric_name}"
 
-            if metric_path not in thresholds:
+            threshold, check_name = self._find_threshold(thresholds, metric_path)
+            if threshold is None:
                 continue
 
-            threshold = thresholds[metric_path]
-            
             # Get or create alert state
             if metric_path not in alert_states:
                 alert_states[metric_path] = AlertState(metric_path)
@@ -855,13 +902,15 @@ class ThresholdChecker:
             elif new_level == AlertLevel.WARNING and threshold.warning is not None:
                 threshold_value = threshold.warning
 
+            alert_state.hysteresis = threshold.hysteresis if new_level != AlertLevel.OK else None
+
             # Update state and check for changes
             old_level = alert_state.level
             if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
                 state_changes.append((metric_path, old_level, new_level, value))
-                self._apply_grace(host_name, alert_state, metric_path, old_level, new_level, value, threshold, data)
+                self._apply_grace(host_name, alert_state, metric_path, old_level, new_level, value, threshold, data, check_name=check_name, metric_name=metric_name)
             elif new_level != AlertLevel.OK:
-                self._check_pending_or_renotify(host_name, alert_state, metric_path, value, threshold, data)
+                self._check_pending_or_renotify(host_name, alert_state, metric_path, value, threshold, data, check_name=check_name, metric_name=metric_name)
 
         # Check nested metrics (e.g., partition data in disk_monitor)
         self._check_nested_metrics(
@@ -921,6 +970,8 @@ class ThresholdChecker:
                     elif new_level == AlertLevel.WARNING and threshold.warning is not None:
                         threshold_value = threshold.warning
 
+                    alert_state.hysteresis = threshold.hysteresis if new_level != AlertLevel.OK else None
+
                     old_level = alert_state.level
                     if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
                         state_changes.append((metric_path, old_level, new_level, value))
@@ -937,6 +988,8 @@ class ThresholdChecker:
         value: Any,
         threshold: ThresholdConfig,
         plugin_data: Optional[Dict[str, Any]] = None,
+        check_name: Optional[str] = None,
+        metric_name: Optional[str] = None,
     ):
         """Trigger a notification for an alert state change.
         
@@ -975,7 +1028,9 @@ class ThresholdChecker:
                     value=display_value,
                     threshold_value=threshold_value,
                     op_symbol=op_symbol,
-                    plugin_data=plugin_data
+                    plugin_data=plugin_data,
+                    check_name=check_name,
+                    metric_name=metric_name,
                 )
                 message = f"{metric_path} = {display_value} {threshold_info}"
             else:
@@ -988,7 +1043,9 @@ class ThresholdChecker:
                     value=display_value,
                     threshold_value=threshold_value,
                     op_symbol=op_symbol,
-                    plugin_data=plugin_data
+                    plugin_data=plugin_data,
+                    check_name=check_name,
+                    metric_name=metric_name,
                 )
                 message = f"{metric_path} = {display_value} {threshold_info}"
             else:
@@ -1005,7 +1062,9 @@ class ThresholdChecker:
                 value=display_value,
                 threshold_value=threshold_value,
                 op_symbol=op_symbol,
-                plugin_data=plugin_data
+                plugin_data=plugin_data,
+                check_name=check_name,
+                metric_name=metric_name,
             )
 
         return lvl, message, formatted_threshold_msg
@@ -1058,15 +1117,21 @@ class ThresholdChecker:
         threshold_value: float,
         op_symbol: str,
         plugin_data: Optional[Dict[str, Any]] = None,
+        check_name: Optional[str] = None,
+        metric_name: Optional[str] = None,
     ) -> str:
         """Format the display string using available data.
 
-        Args:
-            display_format: Format string from threshold config
-            value: Current metric value
-            threshold_value: Threshold value that was exceeded
-            op_symbol: Comparison operator symbol
-            plugin_data: Optional dictionary of plugin data fields
+        Available template variables:
+            {value}           - current metric value
+            {threshold_value} - threshold that was exceeded
+            {op_symbol}       - comparison operator (>, <, >=, <=, ==, !=)
+            {check_name}      - prefix stripped for generic threshold match
+                                (e.g. "check_disk_root" when metric
+                                "check_disk_root_status_code" matched generic
+                                threshold "status_code")
+            {metric_name}     - field name within the plugin data dict
+            Any key from plugin_data is also available.
 
         Returns:
             Formatted display string
@@ -1078,6 +1143,12 @@ class ThresholdChecker:
             'op_symbol': op_symbol,
         }
 
+        # Add generic-match context variables when available
+        if check_name is not None:
+            format_context['check_name'] = check_name
+        if metric_name is not None:
+            format_context['metric_name'] = metric_name
+
         # Add all plugin data fields if available
         if plugin_data:
             format_context.update(plugin_data)
@@ -1111,17 +1182,22 @@ class ThresholdChecker:
         value: Any,
         threshold: ThresholdConfig,
         plugin_data: Optional[Dict[str, Any]],
+        check_name: Optional[str] = None,
+        metric_name: Optional[str] = None,
     ) -> None:
         """Handle a state-change transition with grace-period logic.
 
-        Transitioning INTO alert: defers the notification for grace_seconds.
+        Transitioning INTO alert (worsening): defers the notification for grace_seconds.
+        De-escalation within alert states (e.g. CRITICAL→WARNING): no new notification;
+          the metric is still alerting so no RECOVER was sent.
         Transitioning TO OK:
           - Still in grace window (pending_since set): suppresses both the alert
             and the recovery — the spike never warranted a page.
           - Past grace: fires the RECOVER notification normally.
         """
         lvl, message, formatted_msg = self._trigger_notification(
-            host_name, metric_path, old_level, new_level, value, threshold, plugin_data
+            host_name, metric_path, old_level, new_level, value, threshold, plugin_data,
+            check_name=check_name, metric_name=metric_name,
         )
         alert_state.formatted_message = formatted_msg
 
@@ -1134,12 +1210,20 @@ class ThresholdChecker:
                 alert_state.pending_since = None
             else:
                 self._send_notification(host_name, lvl, message, metric_path, old_level, new_level, value)
-        else:
+        elif new_level.value > old_level.value:
+            # Worsening (OK→WARNING, OK→CRITICAL, WARNING→CRITICAL): schedule notification.
             alert_state.pending_since = time.time()
             logger.debug(
                 "Alert deferred (%.0fs grace): %s on %s = %s",
                 self.grace_seconds, metric_path, host_name, value,
             )
+        else:
+            # De-escalation within alert states (e.g. CRITICAL→WARNING): metric is still
+            # alerting but did not recover, so no new notification.
+            logger.debug(
+                "De-escalation %s→%s for %s on %s, no notification",
+                old_level.name, new_level.name, metric_path, host_name,
+            )
 
     def _check_pending_or_renotify(
         self,
@@ -1149,6 +1233,8 @@ class ThresholdChecker:
         value: Any,
         threshold: ThresholdConfig,
         plugin_data: Optional[Dict[str, Any]],
+        check_name: Optional[str] = None,
+        metric_name: Optional[str] = None,
     ) -> None:
         """Called when alert level is unchanged and non-OK.
 
@@ -1158,7 +1244,8 @@ class ThresholdChecker:
         if alert_state.pending_since is not None:
             if time.time() - alert_state.pending_since >= self.grace_seconds:
                 lvl, message, formatted_msg = self._trigger_notification(
-                    host_name, metric_path, AlertLevel.OK, alert_state.level, value, threshold, plugin_data
+                    host_name, metric_path, AlertLevel.OK, alert_state.level, value, threshold, plugin_data,
+                    check_name=check_name, metric_name=metric_name,
                 )
                 alert_state.formatted_message = formatted_msg
                 self._send_notification(
@@ -1167,7 +1254,7 @@ class ThresholdChecker:
                 alert_state.pending_since = None
             # else: still within grace window, do nothing
         else:
-            self._check_renotify(host_name, alert_state, metric_path, value, threshold, plugin_data)
+            self._check_renotify(host_name, alert_state, metric_path, value, threshold, plugin_data, check_name=check_name, metric_name=metric_name)
 
     def _check_renotify(
         self,
@@ -1177,6 +1264,8 @@ class ThresholdChecker:
         value: Any,
         threshold: ThresholdConfig,
         plugin_data: Optional[Dict[str, Any]] = None,
+        check_name: Optional[str] = None,
+        metric_name: Optional[str] = None,
     ):
         """Check if we should send a repeat notification.
         
@@ -1223,7 +1312,9 @@ class ThresholdChecker:
                     value=value,
                     threshold_value=threshold_value,
                     op_symbol=op_symbol,
-                    plugin_data=plugin_data
+                    plugin_data=plugin_data,
+                    check_name=check_name,
+                    metric_name=metric_name,
                 )
                 message = f"REMINDER ({alert_state.level.name}): {host_name} - {metric_path} = {value} {threshold_info}, ongoing for {int(now - alert_state.since)}s"
             else:
@@ -1244,6 +1335,26 @@ class ThresholdChecker:
             alert_state.last_notification = now
             alert_state.notification_count += 1
     
+    def purge_stale_alerts(self, hbdclass) -> None:
+        """Remove alert states that have no matching threshold configuration.
+
+        Called after startup (pickle restore) and after each config reload so
+        that alerts orphaned by configuration changes do not linger forever.
+        Alerts whose metric_path is not present in the current threshold config
+        for that host are silently dropped.
+        """
+        for hostname, host in hbdclass.Host.hosts.items():
+            if not host.alert_states:
+                continue
+            configured = self.get_thresholds_for_host(hostname)
+            stale = [mp for mp in host.alert_states if self._find_threshold(configured, mp)[0] is None]
+            for mp in stale:
+                logger.info(
+                    "Purging stale alert state for %s / %s (no threshold configured)",
+                    hostname, mp,
+                )
+                del host.alert_states[mp]
+
     def get_active_alerts(self, alert_states: Dict[str, AlertState]) -> list:
         """
         Get all currently active (non-OK) alerts.

hbd/server/udp.py

@@ -440,8 +440,12 @@ def handle_datagram(msg: dict, addr, transport, ctx: dict):
         if not newh:
             if d == 0 or lasts == "unknown":
                 m = "%s is up" % (conn.afam)
+            elif d < 4:
+                # Transient blip (likely client restart) — skip log and notification
+                m = None
             else:
                 m = "%s back after being %s for %s" % (conn.afam, lasts, dur(d))
+            if m:
                 eventlog(uname, "RECOVER", m)
                 if host.watched:
                     asyncio.create_task(notify_mod.send_notification(

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "hbd"
-version = "5.1.14"
+version = "5.1.20"
 description = "Heartbeat monitoring system — client (hbc) and server (hbd)"
 readme = "README.md"
 requires-python = ">=3.11"

scripts/hbc_mini.py

@@ -41,7 +41,7 @@ from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple
 
 # updated by scripts/bumpminor.sh
-__version__ = "5.1.14"
+__version__ = "5.1.20"
 
 # ---------------------------------------------------------------------------
 # Protocol  (mirrors hbd/common/proto.py)
@@ -487,6 +487,12 @@ class CPUMonitorPlugin(MonitorPlugin):
         except Exception:
             pass
 
+        try:
+            with open("/proc/uptime") as fh:
+                data["uptime_seconds"] = int(float(fh.read().split()[0]))
+        except Exception:
+            pass
+
         return data
 
 
@@ -535,6 +541,20 @@ class MemoryMonitorPlugin(MonitorPlugin):
         total = mi.get("MemTotal", 0)
         avail = mi.get("MemAvailable", mi.get("MemFree", 0))
         free = mi.get("MemFree", 0)
+
+        # ZFS ARC is reclaimable but not included in MemAvailable; add it.
+        arc_kb = 0
+        try:
+            with open("/proc/spl/kstat/zfs/arcstats") as _f:
+                for _line in _f:
+                    _p = _line.split()
+                    if len(_p) >= 3 and _p[0] == "size":
+                        arc_kb = int(_p[2]) // 1024
+                        break
+        except (OSError, ValueError):
+            pass
+
+        avail = min(avail + arc_kb, total)
         used = total - avail
         data: Dict[str, Any] = {
             "memory_total": total * 1024,
@@ -1052,8 +1072,8 @@ async def _async_main(args, cfg: Dict[str, Any]) -> int:
         if args.message:
             bmsg["service"] = "service"
             bmsg["msg"] = args.message
-        for c in connections:
-            await c.sendto(bmsg)
+        target = next((c for c in connections if c._transport), connections[0])
+        await target.sendto(bmsg)
         if args.message and not args.daemon:
             await asyncio.sleep(0.3)
             for c in connections:
@@ -1085,11 +1105,13 @@ async def _async_main(args, cfg: Dict[str, Any]) -> int:
         pass
 
     log.info("shutting down")
-    for conn in connections:
+    target = next((c for c in connections if c._transport), connections[0] if connections else None)
+    if target:
         try:
-            await conn.sendto({"shutdown": 1, "acks": conn.ackcount})
+            await target.sendto({"shutdown": 1, "acks": target.ackcount})
         except Exception:
             pass
+    for conn in connections:
         conn.close()
     await asyncio.sleep(0.3)
     for plugin in plugins: