feat(agent) multi-driver db monitoring, hot-plug config, release pipeline

internal/collectors/postgres/postgres.go

@@ -0,0 +1,63 @@
+package postgres
+
+import (
+	"context"
+	"database/sql"
+	"time"
+
+	_ "github.com/lib/pq"
+
+	"bugs.watch/agent/internal/aggregate"
+)
+
+// Collector queries PostgreSQL status each sample interval.
+// All queries run under a 3-second timeout; any failure is silently skipped.
+type Collector struct {
+	db *sql.DB
+}
+
+func New(dsn string) (*Collector, error) {
+	db, err := sql.Open("postgres", dsn)
+	if err != nil {
+		return nil, err
+	}
+	db.SetMaxOpenConns(2)
+	db.SetMaxIdleConns(1)
+	db.SetConnMaxLifetime(5 * time.Minute)
+	return &Collector{db: db}, nil
+}
+
+func (c *Collector) Name() string { return "postgres" }
+
+func (c *Collector) Close() { c.db.Close() }
+
+func (c *Collector) Collect(window *aggregate.Window) {
+	ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+	defer cancel()
+
+	// pg_stat_database is readable by any user for their own database.
+	var numBackends, maxConn int
+	err := c.db.QueryRowContext(ctx,
+		`SELECT d.numbackends, s.setting::int
+		 FROM pg_stat_database d, pg_settings s
+		 WHERE d.datname = current_database() AND s.name = 'max_connections'`,
+	).Scan(&numBackends, &maxConn)
+	if err == nil && maxConn > 0 {
+		window.AddDBConnPct(float64(numBackends) / float64(maxConn) * 100)
+	}
+
+	// pg_stat_activity requires pg_monitor role (PG 10+); silently skip on error.
+	var activeQueries int
+	if err := c.db.QueryRowContext(ctx,
+		`SELECT count(*) FROM pg_stat_activity WHERE state = 'active'`,
+	).Scan(&activeQueries); err == nil {
+		window.AddDBThreadsRunning(float64(activeQueries))
+	}
+
+	var sizeBytes int64
+	if err := c.db.QueryRowContext(ctx,
+		`SELECT pg_database_size(current_database())`,
+	).Scan(&sizeBytes); err == nil {
+		window.SetDBSizeBytes(uint64(sizeBytes))
+	}
+}