path: root/internal/monitor/scheduler.go

package monitor

import (
	"context"
	"log/slog"
	"sync"
	"time"

	"github.com/Fuwn/kaze/internal/config"
	"github.com/Fuwn/kaze/internal/storage"
)

type OnCheckCallback func()

type Scheduler struct {
	monitors    []Monitor
	monitorCfg  map[string]config.MonitorConfig
	configPath  string
	storage     *storage.Storage
	logger      *slog.Logger
	wg          sync.WaitGroup
	ctx         context.Context
	cancel      context.CancelFunc
	onCheckDone OnCheckCallback
}

// NewScheduler creates a new monitor scheduler
func NewScheduler(cfg *config.Config, store *storage.Storage, logger *slog.Logger, configPath string) (*Scheduler, error) {
	ctx, cancel := context.WithCancel(context.Background())

	s := &Scheduler{
		monitorCfg: make(map[string]config.MonitorConfig),
		configPath: configPath,
		storage:    store,
		logger:     logger,
		ctx:        ctx,
		cancel:     cancel,
	}

	// Create monitors from configuration
	for _, group := range cfg.Groups {
		for _, monCfg := range group.Monitors {
			mon, err := New(monCfg)
			if err != nil {
				cancel()
				return nil, err
			}
			s.monitors = append(s.monitors, mon)
			s.monitorCfg[mon.ID()] = monCfg // Store config by ID (group/name) for reset flag checks
			logger.Info("registered monitor",
				"id", mon.ID(),
				"group", mon.Group(),
				"name", mon.Name(),
				"type", mon.Type(),
				"target", mon.Target(),
				"interval", mon.Interval())
		}
	}

	return s, nil
}

// Start begins running all monitors
func (s *Scheduler) Start() {
	s.logger.Info("starting scheduler", "monitors", len(s.monitors))

	for _, mon := range s.monitors {
		s.wg.Add(1)
		go s.runMonitor(mon)
	}

	s.wg.Add(1)
	go s.runCleanup()

	s.wg.Add(1)
	go s.runMaintenance()
}

// Stop gracefully stops all monitors
func (s *Scheduler) Stop() {
	s.logger.Info("stopping scheduler")
	s.cancel()
	s.wg.Wait()
	s.logger.Info("scheduler stopped")
}

// runMonitor runs a single monitor in a loop
func (s *Scheduler) runMonitor(mon Monitor) {
	defer s.wg.Done()

	// Run immediately on start
	s.executeCheck(mon)

	ticker := time.NewTicker(mon.Interval())
	defer ticker.Stop()

	for {
		select {
		case <-s.ctx.Done():
			s.logger.Info("monitor stopped", "id", mon.ID())
			return
		case <-ticker.C:
			s.executeCheck(mon)
		}
	}
}

// executeCheck performs a single check and saves the result
func (s *Scheduler) executeCheck(mon Monitor) {
	// Check if reset flag is set for this monitor
	if monCfg, exists := s.monitorCfg[mon.ID()]; exists && monCfg.ResetOnNextCheck {
		s.logger.Info("resetting monitor data", "id", mon.ID())

		// Delete all historical data for this monitor (using ID as the key)
		if err := s.storage.ResetMonitorData(s.ctx, mon.ID()); err != nil {
			s.logger.Error("failed to reset monitor data",
				"id", mon.ID(),
				"error", err)
		} else {
			s.logger.Info("monitor data reset complete", "id", mon.ID())

			// Flip the reset flag to false in the config file
			// Note: UpdateResetFlag uses monitor name for YAML lookup, group for context
			if err := config.UpdateResetFlag(s.configPath, monCfg.Group, monCfg.Name, false); err != nil {
				s.logger.Error("failed to update reset flag in config",
					"id", mon.ID(),
					"error", err)
			} else {
				// Update in-memory config
				monCfg.ResetOnNextCheck = false
				s.monitorCfg[mon.ID()] = monCfg
				s.logger.Info("reset flag cleared in config", "id", mon.ID())
			}
		}
	}

	// Create a context with timeout for this check
	checkCtx, cancel := context.WithTimeout(s.ctx, mon.Interval())
	defer cancel()

	var result *Result
	retries := mon.Retries()

	// Try the check, with retries if configured
	for attempt := 0; attempt <= retries; attempt++ {
		result = mon.Check(checkCtx)

		// If check succeeded (up or degraded), no need to retry
		if result.Status == StatusUp || result.Status == StatusDegraded {
			break
		}

		// If this wasn't the last attempt, log and retry
		if attempt < retries {
			s.logger.Debug("check failed, retrying",
				"name", mon.Name(),
				"attempt", attempt+1,
				"max_retries", retries,
				"error", result.Error)

			// Small delay before retry (500ms)
			select {
			case <-checkCtx.Done():
				// Context cancelled, abort retries
				break
			case <-time.After(500 * time.Millisecond):
				// Continue to next retry
			}
		}
	}

	// Log the result
	logAttrs := []any{
		"id", mon.ID(),
		"status", result.Status,
		"response_time", result.ResponseTime,
	}
	if result.StatusCode > 0 {
		logAttrs = append(logAttrs, "status_code", result.StatusCode)
	}
	if result.SSLDaysLeft > 0 {
		logAttrs = append(logAttrs, "ssl_days_left", result.SSLDaysLeft)
	}
	if result.Error != nil {
		logAttrs = append(logAttrs, "error", result.Error)
	}

	if result.Status == StatusUp {
		s.logger.Debug("check completed", logAttrs...)
	} else {
		s.logger.Warn("check completed", logAttrs...)
	}

	// Save to storage
	if err := s.storage.SaveCheckResult(s.ctx, result.ToCheckResult()); err != nil {
		s.logger.Error("failed to save check result",
			"id", mon.ID(),
			"error", err)
	}

	s.storage.RecordCheck()

	if s.onCheckDone != nil {
		s.onCheckDone()
	}
}

// runCleanup periodically cleans up old data
func (s *Scheduler) runCleanup() {
	defer s.wg.Done()

	ticker := time.NewTicker(24 * time.Hour)
	defer ticker.Stop()

	for {
		select {
		case <-s.ctx.Done():
			return
		case <-ticker.C:
			s.logger.Info("running database cleanup")
			if err := s.storage.Cleanup(s.ctx); err != nil {
				s.logger.Error("cleanup failed", "error", err)
			} else {
				s.logger.Info("cleanup completed")
			}
		}
	}
}

func (s *Scheduler) runMaintenance() {
	defer s.wg.Done()

	if s.storage.GetMaintenanceMode() == "never" {
		return
	}

	ticker := time.NewTicker(1 * time.Minute)
	defer ticker.Stop()

	for {
		select {
		case <-s.ctx.Done():
			return
		case <-ticker.C:
			ran, err := s.storage.CheckMaintenance()
			if err != nil {
				s.logger.Error("maintenance check failed", "error", err)
			} else if ran {
				s.logger.Info("database maintenance completed",
					"mode", s.storage.GetMaintenanceMode())
			}
		}
	}
}

func (s *Scheduler) GetMonitors() []Monitor {
	return s.monitors
}

func (s *Scheduler) SetOnCheckCallback(cb OnCheckCallback) {
	s.onCheckDone = cb
}

// RunCheck manually triggers a check for a specific monitor by ID (group/name format)
func (s *Scheduler) RunCheck(id string) *Result {
	for _, mon := range s.monitors {
		if mon.ID() == id {
			ctx, cancel := context.WithTimeout(context.Background(), mon.Interval())
			defer cancel()
			result := mon.Check(ctx)

			// Save the result
			if err := s.storage.SaveCheckResult(context.Background(), result.ToCheckResult()); err != nil {
				s.logger.Error("failed to save manual check result",
					"id", mon.ID(),
					"error", err)
			}

			return result
		}
	}
	return nil
}