fix: Add configurable poll timeout and handle external Ceph storage

Changes:
1. Add MAX_POLL_TIMEOUT env var for large Proxmox clusters that need
   more than 3 minutes for polling (default: 3m, minimum: 30s)
2. Handle external Ceph storage gracefully - don't mark nodes unhealthy
   when Proxmox returns 'binary not installed' (e.g., for Ceph not
   managed by Proxmox)

Related to #965

internal/config/config.go

@@ -119,7 +119,8 @@ type Config struct {
 	GuestMetadataRetryBackoff       time.Duration `envconfig:"GUEST_METADATA_RETRY_BACKOFF" default:"30s" json:"guestMetadataRetryBackoff"`
 	GuestMetadataMaxConcurrent      int           `envconfig:"GUEST_METADATA_MAX_CONCURRENT" default:"4" json:"guestMetadataMaxConcurrent"`
 	DNSCacheTimeout                 time.Duration `envconfig:"DNS_CACHE_TIMEOUT" default:"5m" json:"dnsCacheTimeout"`
-	SSHPort                         int           `envconfig:"SSH_PORT" default:"22" json:"sshPort"` // Default SSH port for temperature monitoring
+	SSHPort                         int           `envconfig:"SSH_PORT" default:"22" json:"sshPort"`   // Default SSH port for temperature monitoring
+	MaxPollTimeout                  time.Duration `envconfig:"MAX_POLL_TIMEOUT" default:"3m" json:"-"` // Maximum poll timeout for large clusters (default 3m)
 
 	// Metrics retention settings (tiered storage)
 	// These control how long historical metrics are retained at each aggregation level.
@@ -594,6 +595,7 @@ func Load() (*Config, error) {
 		DiscoveryEnabled:                false,
 		DiscoverySubnet:                 "auto",
 		TemperatureMonitoringEnabled:    true,
+		MaxPollTimeout:                  3 * time.Minute, // Default max poll timeout for large clusters
 		EnableSensorProxy:               false,
 		EnvOverrides:                    make(map[string]bool),
 		AgentConnectURL:                 "",
@@ -1446,6 +1448,19 @@ func Load() (*Config, error) {
 			log.Info().Dur("timeout", d).Msg("Connection timeout overridden by CONNECTION_TIMEOUT env var")
 		}
 	}
+	if maxPollTimeout := os.Getenv("MAX_POLL_TIMEOUT"); maxPollTimeout != "" {
+		if d, err := time.ParseDuration(maxPollTimeout); err == nil {
+			if d >= 30*time.Second { // Minimum 30 seconds
+				cfg.MaxPollTimeout = d
+				cfg.EnvOverrides["maxPollTimeout"] = true
+				log.Info().Dur("timeout", d).Msg("Max poll timeout overridden by MAX_POLL_TIMEOUT env var")
+			} else {
+				log.Warn().Dur("value", d).Msg("MAX_POLL_TIMEOUT too low (minimum 30s), using default")
+			}
+		} else {
+			log.Warn().Str("value", maxPollTimeout).Msg("Invalid MAX_POLL_TIMEOUT value, using default")
+		}
+	}
 	if allowedOrigins := os.Getenv("ALLOWED_ORIGINS"); allowedOrigins != "" {
 		cfg.AllowedOrigins = allowedOrigins
 		cfg.EnvOverrides["allowedOrigins"] = true

internal/monitoring/monitor.go

@@ -4529,8 +4529,14 @@ func derivePollTimeout(cfg *config.Config) time.Duration {
 	if timeout < minTaskTimeout {
 		timeout = minTaskTimeout
 	}
-	if timeout > maxTaskTimeout {
-		timeout = maxTaskTimeout
+	// Use configurable max timeout from config (set via MAX_POLL_TIMEOUT env var)
+	// Falls back to hardcoded maxTaskTimeout if config is nil or MaxPollTimeout not set
+	maxTimeout := maxTaskTimeout
+	if cfg != nil && cfg.MaxPollTimeout > 0 {
+		maxTimeout = cfg.MaxPollTimeout
+	}
+	if timeout > maxTimeout {
+		timeout = maxTimeout
 	}
 	return timeout
 }

internal/monitoring/monitor_timeout_test.go

@@ -39,6 +39,22 @@ func TestDerivePollTimeout(t *testing.T) {
 			},
 			want: maxTaskTimeout,
 		},
+		{
+			name: "respects custom MaxPollTimeout",
+			cfg: &config.Config{
+				ConnectionTimeout: 3 * time.Minute,
+				MaxPollTimeout:    10 * time.Minute,
+			},
+			want: 6 * time.Minute, // 2 * ConnectionTimeout, still under MaxPollTimeout
+		},
+		{
+			name: "custom MaxPollTimeout caps at configured value",
+			cfg: &config.Config{
+				ConnectionTimeout: 10 * time.Minute,
+				MaxPollTimeout:    5 * time.Minute,
+			},
+			want: 5 * time.Minute, // Capped at MaxPollTimeout
+		},
 	}
 
 	for _, tt := range tests {

pkg/proxmox/cluster_client.go

@@ -724,6 +724,8 @@ func (cc *ClusterClient) executeWithFailover(ctx context.Context, fn func(*Clien
 			// PBS storage errors - Proxmox can't reach PBS, but node is still reachable
 			(strings.Contains(errStr, "500") && strings.Contains(errStr, "pbs-") && strings.Contains(errStr, "error fetching datastores")) ||
 			(strings.Contains(errStr, "500") && strings.Contains(errStr, "Can't connect to") && strings.Contains(errStr, ":8007")) ||
+			// External Ceph errors - Ceph not managed by Proxmox, but node is still reachable
+			(strings.Contains(errStr, "500") && strings.Contains(errStr, "binary not installed")) ||
 			// RRD data timeouts - secondary metric fetch failures, node is still working
 			(strings.Contains(errStr, "context deadline exceeded") && strings.Contains(errStr, "/rrddata")) ||
 			(strings.Contains(errStr, "context deadline exceeded") && strings.Contains(errStr, "/lxc/") && strings.Contains(errStr, "rrd")) ||