fix(dbmanager): keep the pool alive across errors and restarts

Implements the fixes from audit/pkg/dbmanager.audit.md.

- Stop closing the shared *sql.DB to recover from errors. Adapter
  factories and the health checker no longer call Reconnect; Reconnect is
  atomic and operator-only.
- Postgres uses a custom driver.Connector: Reconnect retires pooled
  connections by generation without closing the pool, so held Bun/GORM
  handles keep working. Verified against a live server restart.
- Add TCP keepalive, TCP_USER_TIMEOUT, a bounded reuse ping and
  statement_timeout as a runtime parameter; drop the 2 min timeout floor.
- Health check pings without holding the connection lock.
- Listener: single goroutine pair, bounded Close without UNLISTEN, and
  serialised use of the pgx connection (fixes conn busy and a close race).
- Fix Connect/Close/Connect/Close panic, idempotent Connect, dial outside
  the manager lock, clean up on partial failure.
- SQLite: pin :memory: to one connection, pragmas via DSN.
- Escape credentials in Postgres/MSSQL/Mongo DSNs; sslmode defaults to
  prefer. Wire retry settings, publish metrics, fix logger calls.
- NewConnectionFromDB: Close is a no-op with a warning (caller owns the
  pool); Reconnect only pings.
- Document correct usage in the README; mark the audit with what was done.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

pkg/dbmanager/metrics.go

@@ -1,6 +1,8 @@
 package dbmanager
 
 import (
+	"sync"
+
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promauto"
 )
@@ -34,8 +36,8 @@ var (
 	)
 
 	// connectionWaitCount tracks how many times connections had to wait for availability
-	connectionWaitCount = promauto.NewGaugeVec(
-		prometheus.GaugeOpts{
+	connectionWaitCount = promauto.NewCounterVec(
+		prometheus.CounterOpts{
 			Name: "dbmanager_connection_wait_count",
 			Help: "Number of times connections had to wait for availability",
 		},
@@ -43,8 +45,8 @@ var (
 	)
 
 	// connectionWaitDuration tracks total time connections spent waiting
-	connectionWaitDuration = promauto.NewGaugeVec(
-		prometheus.GaugeOpts{
+	connectionWaitDuration = promauto.NewCounterVec(
+		prometheus.CounterOpts{
 			Name: "dbmanager_connection_wait_duration_seconds",
 			Help: "Total time connections spent waiting for availability",
 		},
@@ -61,8 +63,8 @@ var (
 	)
 
 	// connectionLifetimeClosed tracks connections closed due to max lifetime
-	connectionLifetimeClosed = promauto.NewGaugeVec(
-		prometheus.GaugeOpts{
+	connectionLifetimeClosed = promauto.NewCounterVec(
+		prometheus.CounterOpts{
 			Name: "dbmanager_connection_lifetime_closed_total",
 			Help: "Total connections closed due to exceeding max lifetime",
 		},
@@ -70,8 +72,8 @@ var (
 	)
 
 	// connectionIdleClosed tracks connections closed due to max idle time
-	connectionIdleClosed = promauto.NewGaugeVec(
-		prometheus.GaugeOpts{
+	connectionIdleClosed = promauto.NewCounterVec(
+		prometheus.CounterOpts{
 			Name: "dbmanager_connection_idle_closed_total",
 			Help: "Total connections closed due to exceeding max idle time",
 		},
@@ -114,13 +116,13 @@ func (m *connectionManager) PublishMetrics() {
 			connectionPoolSize.WithLabelValues(name, string(connStats.Type), "idle").Set(float64(connStats.Idle))
 			connectionPoolSize.WithLabelValues(name, string(connStats.Type), "in_use").Set(float64(connStats.InUse))
 
-			// Wait stats
-			connectionWaitCount.With(labels).Set(float64(connStats.WaitCount))
-			connectionWaitDuration.With(labels).Set(connStats.WaitDuration.Seconds())
-
-			// Lifetime/idle closed
-			connectionLifetimeClosed.With(labels).Set(float64(connStats.MaxLifetimeClosed))
-			connectionIdleClosed.With(labels).Set(float64(connStats.MaxIdleClosed))
+			// sql.DBStats values are cumulative, so add only the growth since
+			// the last publish to keep these true counters.
+			prev := lastPublished.swap(name, connStats)
+			connectionWaitCount.With(labels).Add(float64(connStats.WaitCount - prev.WaitCount))
+			connectionWaitDuration.With(labels).Add((connStats.WaitDuration - prev.WaitDuration).Seconds())
+			connectionLifetimeClosed.With(labels).Add(float64(connStats.MaxLifetimeClosed - prev.MaxLifetimeClosed))
+			connectionIdleClosed.With(labels).Add(float64(connStats.MaxIdleClosed - prev.MaxIdleClosed))
 		}
 	}
 }
@@ -134,3 +136,25 @@ func RecordReconnectAttempt(name string, dbType DatabaseType, success bool) {
 
 	reconnectAttempts.WithLabelValues(name, string(dbType), result).Inc()
 }
+
+// publishedStats remembers the cumulative pool stats last exported per
+// connection so counters can be advanced by the delta.
+type publishedStats struct {
+	mu   sync.Mutex
+	last map[string]ConnectionStats
+}
+
+var lastPublished = &publishedStats{last: make(map[string]ConnectionStats)}
+
+// swap stores cur and returns the previous value. A counter reset (a new pool
+// after Close+Connect) is treated as starting from zero.
+func (p *publishedStats) swap(name string, cur *ConnectionStats) ConnectionStats {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	prev := p.last[name]
+	if cur.WaitCount < prev.WaitCount || cur.MaxIdleClosed < prev.MaxIdleClosed || cur.MaxLifetimeClosed < prev.MaxLifetimeClosed {
+		prev = ConnectionStats{}
+	}
+	p.last[name] = *cur
+	return prev
+}