refactor: decompose monitor.go and sqlstore.go into focused files

Split monitor.go (1131→350 lines) into alerts.go, checks.go, sites.go, and maintenance.go by concern. Split sqlstore.go (833→470 lines) into sqlstore_alerts.go, sqlstore_history.go, and sqlstore_maintenance.go by domain. Tests move with their implementation. Pure reorganization — no behavioral changes.
2026-06-27 16:38:53 -04:00
parent edbc2beddd
commit 9eaddbb6e8
18 changed files with 2671 additions and 2553 deletions
@@ -5,7 +5,6 @@ import (
 	"crypto/rand"
 	"database/sql"
 	"encoding/hex"
-	"encoding/json"
 	"fmt"
 	"strings"
 	"time"
@@ -214,40 +213,6 @@ func (s *SQLStore) GetSiteByName(ctx context.Context, name string) (models.SiteC
 	return st, err
 }

-func (s *SQLStore) unmarshalSettings(raw string) (map[string]string, error) {
-	decrypted, err := s.decryptSettings(raw)
-	if err != nil {
-		return nil, fmt.Errorf("decrypt settings: %w", err)
-	}
-	var m map[string]string
-	if err := json.Unmarshal([]byte(decrypted), &m); err != nil {
-		return nil, fmt.Errorf("unmarshal settings: %w", err)
-	}
-	return m, nil
-}
-
-func (s *SQLStore) marshalSettings(settings map[string]string) (string, error) {
-	jsonBytes, err := json.Marshal(settings)
-	if err != nil {
-		return "", err
-	}
-	return s.encryptSettings(string(jsonBytes))
-}
-
-func (s *SQLStore) GetAlertByName(ctx context.Context, name string) (models.AlertConfig, error) {
-	var a models.AlertConfig
-	var settingsRaw string
-	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE name = ?"), name).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
-	if err != nil {
-		return a, err
-	}
-	a.Settings, err = s.unmarshalSettings(settingsRaw)
-	if err != nil {
-		return a, fmt.Errorf("alert %q: %w", name, err)
-	}
-	return a, nil
-}
-
 func (s *SQLStore) AddSiteReturningID(ctx context.Context, site models.SiteConfig) (int, error) {
 	token := ""
 	if site.Type == "push" {
@@ -274,89 +239,6 @@ func (s *SQLStore) AddSiteReturningID(ctx context.Context, site models.SiteConfi
 	return int(id), err
 }

-func (s *SQLStore) AddAlertReturningID(ctx context.Context, name, aType string, settings map[string]string) (int, error) {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return 0, err
-	}
-	if s.dollar {
-		var id int
-		err := s.db.QueryRowContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?) RETURNING id"), name, aType, stored).Scan(&id)
-		return id, err
-	}
-	result, err := s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
-	if err != nil {
-		return 0, err
-	}
-	id, err := result.LastInsertId()
-	return int(id), err
-}
-
-func (s *SQLStore) GetAllAlerts(ctx context.Context) ([]models.AlertConfig, error) {
-	rows, err := s.db.QueryContext(ctx, "SELECT id, name, type, settings FROM alerts")
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var alerts []models.AlertConfig
-	for rows.Next() {
-		var a models.AlertConfig
-		var settingsRaw string
-		if err := rows.Scan(&a.ID, &a.Name, &a.Type, &settingsRaw); err != nil {
-			return alerts, err
-		}
-		a.Settings, err = s.unmarshalSettings(settingsRaw)
-		if err != nil {
-			return alerts, fmt.Errorf("alert %q: %w", a.Name, err)
-		}
-		alerts = append(alerts, a)
-	}
-	return alerts, rows.Err()
-}
-
-func (s *SQLStore) GetAlert(ctx context.Context, id int) (models.AlertConfig, error) {
-	var a models.AlertConfig
-	var settingsRaw string
-	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE id = ?"), id).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
-	if err != nil {
-		return a, err
-	}
-	a.Settings, err = s.unmarshalSettings(settingsRaw)
-	if err != nil {
-		return a, fmt.Errorf("alert %d: %w", id, err)
-	}
-	return a, nil
-}
-
-func (s *SQLStore) AddAlert(ctx context.Context, name, aType string, settings map[string]string) error {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return err
-	}
-	_, err = s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
-	return err
-}
-
-func (s *SQLStore) UpdateAlert(ctx context.Context, id int, name, aType string, settings map[string]string) error {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return err
-	}
-	_, err = s.db.ExecContext(ctx, s.q("UPDATE alerts SET name=?, type=?, settings=? WHERE id=?"), name, aType, stored, id)
-	return err
-}
-
-func (s *SQLStore) DeleteAlert(ctx context.Context, id int) error {
-	if _, err := s.db.ExecContext(ctx, s.q("UPDATE sites SET alert_id = 0 WHERE alert_id = ?"), id); err != nil {
-		return err
-	}
-	if _, err := s.db.ExecContext(ctx, s.q("DELETE FROM alerts WHERE id=?"), id); err != nil {
-		return err
-	}
-	s.dialect.ResetSequenceOnEmpty(s.db, "alerts")
-	return nil
-}
-
 func (s *SQLStore) GetAllUsers(ctx context.Context) ([]models.User, error) {
 	rows, err := s.db.QueryContext(ctx, "SELECT id, username, public_key, role FROM users")
 	if err != nil {
@@ -389,85 +271,6 @@ func (s *SQLStore) DeleteUser(ctx context.Context, id int) error {
 	return err
 }

-func (s *SQLStore) SaveStateChange(ctx context.Context, siteID int, fromStatus, toStatus, errorReason string) error {
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO state_changes (site_id, from_status, to_status, error_reason) VALUES (?, ?, ?, ?)"),
-		siteID, fromStatus, toStatus, errorReason)
-	return err
-}
-
-func (s *SQLStore) GetStateChanges(ctx context.Context, siteID int, limit int) ([]models.StateChange, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? ORDER BY changed_at DESC LIMIT ?"), siteID, limit)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var changes []models.StateChange
-	for rows.Next() {
-		var sc models.StateChange
-		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
-			return changes, err
-		}
-		changes = append(changes, sc)
-	}
-	return changes, rows.Err()
-}
-
-func (s *SQLStore) GetStateChangesSince(ctx context.Context, siteID int, since time.Time) ([]models.StateChange, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? AND changed_at >= ? ORDER BY changed_at DESC"), siteID, since)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var changes []models.StateChange
-	for rows.Next() {
-		var sc models.StateChange
-		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
-			return changes, err
-		}
-		changes = append(changes, sc)
-	}
-	return changes, rows.Err()
-}
-
-func (s *SQLStore) SaveCheck(ctx context.Context, siteID int, latencyNs int64, isUp bool) error {
-	return s.SaveCheckFromNode(ctx, siteID, "", latencyNs, isUp)
-}
-
-// SaveCheckFromNode inserts a single check row. Retention is handled out of
-// band by PruneCheckHistory on a timer, not per-insert, to keep the write hot
-// path a plain INSERT.
-func (s *SQLStore) SaveCheckFromNode(ctx context.Context, siteID int, nodeID string, latencyNs int64, isUp bool) error {
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO check_history (site_id, node_id, latency_ns, is_up) VALUES (?, ?, ?, ?)"), siteID, nodeID, latencyNs, isUp)
-	return err
-}
-
-// PruneCheckHistory trims check_history to the newest maxCheckHistory rows per
-// site, across all sites, in one pass. Intended to run periodically.
-func (s *SQLStore) PruneCheckHistory(ctx context.Context) error {
-	q := fmt.Sprintf(`DELETE FROM check_history WHERE id IN (
-		SELECT id FROM (
-			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC, id DESC) AS rn
-			FROM check_history
-		) ranked WHERE rn > %d
-	)`, maxCheckHistory)
-	_, err := s.db.ExecContext(ctx, s.q(q))
-	return err
-}
-
-// PruneStateChanges trims state_changes to the newest maxStateChangesPerSite
-// rows per site. Generous so realistic SLA windows are unaffected; bounds the
-// otherwise unbounded growth of a flapping monitor's history.
-func (s *SQLStore) PruneStateChanges(ctx context.Context) error {
-	q := fmt.Sprintf(`DELETE FROM state_changes WHERE id IN (
-		SELECT id FROM (
-			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY changed_at DESC, id DESC) AS rn
-			FROM state_changes
-		) ranked WHERE rn > %d
-	)`, maxStateChangesPerSite)
-	_, err := s.db.ExecContext(ctx, s.q(q))
-	return err
-}
-
 func (s *SQLStore) RegisterNode(ctx context.Context, node models.ProbeNode) error {
 	_, err := s.db.ExecContext(ctx, s.dialect.UpsertNodeSQL(), node.ID, node.Name, node.Region, node.Version)
 	return err
@@ -578,172 +381,6 @@ func (s *SQLStore) LoadLogs(ctx context.Context, limit int) ([]models.LogEntry,
 	return entries, rows.Err()
 }

-func (s *SQLStore) LoadAllHistory(ctx context.Context, limit int) (map[int][]models.CheckRecord, error) {
-	result := make(map[int][]models.CheckRecord)
-	rows, err := s.db.QueryContext(ctx, s.q(`
-		SELECT site_id, latency_ns, is_up FROM (
-			SELECT site_id, latency_ns, is_up,
-				ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC) AS rn
-			FROM check_history
-		) sub WHERE rn <= ?`), limit)
-	if err != nil {
-		return result, err
-	}
-	defer rows.Close()
-	for rows.Next() {
-		var r models.CheckRecord
-		if err := rows.Scan(&r.SiteID, &r.LatencyNs, &r.IsUp); err != nil {
-			return result, err
-		}
-		result[r.SiteID] = append(result[r.SiteID], r)
-	}
-	for id, records := range result {
-		for i, j := 0, len(records)-1; i < j; i, j = i+1, j-1 {
-			records[i], records[j] = records[j], records[i]
-		}
-		result[id] = records
-	}
-	return result, rows.Err()
-}
-
-func (s *SQLStore) scanMaintenanceWindow(rows *sql.Rows) (models.MaintenanceWindow, error) {
-	var mw models.MaintenanceWindow
-	var endTime sql.NullTime
-	if err := rows.Scan(&mw.ID, &mw.MonitorID, &mw.Title, &mw.Description, &mw.Type, &mw.StartTime, &endTime, &mw.CreatedBy, &mw.CreatedAt); err != nil {
-		return mw, err
-	}
-	if endTime.Valid {
-		mw.EndTime = endTime.Time
-	}
-	return mw, nil
-}
-
-func (s *SQLStore) GetActiveMaintenanceWindows(ctx context.Context) ([]models.MaintenanceWindow, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows WHERE start_time <= CURRENT_TIMESTAMP AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP) ORDER BY start_time DESC"))
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var windows []models.MaintenanceWindow
-	for rows.Next() {
-		mw, err := s.scanMaintenanceWindow(rows)
-		if err != nil {
-			return windows, err
-		}
-		windows = append(windows, mw)
-	}
-	return windows, rows.Err()
-}
-
-func (s *SQLStore) GetAllMaintenanceWindows(ctx context.Context, limit int) ([]models.MaintenanceWindow, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows ORDER BY created_at DESC LIMIT ?"), limit)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var windows []models.MaintenanceWindow
-	for rows.Next() {
-		mw, err := s.scanMaintenanceWindow(rows)
-		if err != nil {
-			return windows, err
-		}
-		windows = append(windows, mw)
-	}
-	return windows, rows.Err()
-}
-
-func (s *SQLStore) GetOverlappingMaintenanceWindows(ctx context.Context, monitorID int, startTime, endTime time.Time) ([]models.MaintenanceWindow, error) {
-	var timeClause string
-	var args []interface{}
-
-	if endTime.IsZero() {
-		timeClause = "(end_time IS NULL OR end_time > ?)"
-		args = append(args, startTime)
-	} else {
-		timeClause = "(end_time IS NULL OR end_time > ?) AND start_time < ?"
-		args = append(args, startTime, endTime)
-	}
-
-	var scopeClause string
-	if monitorID == 0 {
-		scopeClause = "1=1"
-	} else {
-		scopeClause = "(monitor_id = ? OR monitor_id = 0 OR monitor_id IN (SELECT parent_id FROM sites WHERE id = ? AND parent_id > 0))"
-		args = append(args, monitorID, monitorID)
-	}
-
-	query := fmt.Sprintf(
-		"SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows WHERE %s AND %s ORDER BY start_time",
-		timeClause, scopeClause,
-	)
-
-	rows, err := s.db.QueryContext(ctx, s.q(query), args...)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-
-	var windows []models.MaintenanceWindow
-	for rows.Next() {
-		mw, err := s.scanMaintenanceWindow(rows)
-		if err != nil {
-			return nil, err
-		}
-		windows = append(windows, mw)
-	}
-	return windows, rows.Err()
-}
-
-func (s *SQLStore) AddMaintenanceWindow(ctx context.Context, mw models.MaintenanceWindow) error {
-	if mw.StartTime.IsZero() {
-		mw.StartTime = time.Now()
-	}
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO maintenance_windows (monitor_id, title, description, type, start_time, end_time, created_by) VALUES (?, ?, ?, ?, ?, ?, ?)"),
-		mw.MonitorID, mw.Title, mw.Description, mw.Type, mw.StartTime, sql.NullTime{Time: mw.EndTime, Valid: !mw.EndTime.IsZero()}, mw.CreatedBy)
-	return err
-}
-
-func (s *SQLStore) EndMaintenanceWindow(ctx context.Context, id int) error {
-	_, err := s.db.ExecContext(ctx, s.q("UPDATE maintenance_windows SET end_time = CURRENT_TIMESTAMP WHERE id = ?"), id)
-	return err
-}
-
-func (s *SQLStore) DeleteMaintenanceWindow(ctx context.Context, id int) error {
-	_, err := s.db.ExecContext(ctx, s.q("DELETE FROM maintenance_windows WHERE id = ?"), id)
-	if err != nil {
-		return err
-	}
-	s.dialect.ResetSequenceOnEmpty(s.db, "maintenance_windows")
-	return nil
-}
-
-func (s *SQLStore) PruneExpiredMaintenanceWindows(ctx context.Context, retention time.Duration) (int64, error) {
-	cutoff := time.Now().Add(-retention)
-	result, err := s.db.ExecContext(ctx,
-		s.q("DELETE FROM maintenance_windows WHERE end_time IS NOT NULL AND end_time < ?"),
-		cutoff,
-	)
-	if err != nil {
-		return 0, err
-	}
-	return result.RowsAffected()
-}
-
-func (s *SQLStore) IsMonitorInMaintenance(ctx context.Context, monitorID int) (bool, error) {
-	var count int
-	err := s.db.QueryRowContext(ctx, s.q(`SELECT COUNT(*) FROM maintenance_windows
-		WHERE type = 'maintenance'
-		AND start_time <= CURRENT_TIMESTAMP
-		AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP)
-		AND (monitor_id = 0 OR monitor_id = ?
-			OR monitor_id IN (SELECT parent_id FROM sites WHERE id = ? AND parent_id > 0))`),
-		monitorID, monitorID).Scan(&count)
-	if err != nil {
-		return false, err
-	}
-	return count > 0, nil
-}
-
 func (s *SQLStore) GetPreference(ctx context.Context, key string) (string, error) {
 	var value string
 	err := s.db.QueryRowContext(ctx, s.q("SELECT value FROM preferences WHERE key = ?"), key).Scan(&value)