package watchdog import ( "encoding/json" "fmt" "github.com/TwinProduction/gatus/alerting" "github.com/TwinProduction/gatus/config" "github.com/TwinProduction/gatus/core" "github.com/TwinProduction/gatus/metric" "log" "sync" "time" ) var ( serviceResults = make(map[string][]*core.Result) // serviceResultsMutex is used to prevent concurrent map access serviceResultsMutex sync.RWMutex // monitoringMutex is used to prevent multiple services from being evaluated at the same time. // Without this, conditions using response time may become inaccurate. monitoringMutex sync.Mutex ) // GetJsonEncodedServiceResults returns a list of the last 20 results for each services encoded using json.Marshal. // The reason why the encoding is done here is because we use a mutex to prevent concurrent map access. func GetJsonEncodedServiceResults() ([]byte, error) { serviceResultsMutex.RLock() data, err := json.Marshal(serviceResults) serviceResultsMutex.RUnlock() return data, err } // Monitor loops over each services and starts a goroutine to monitor each services separately func Monitor(cfg *config.Config) { for _, service := range cfg.Services { go monitor(service) // To prevent multiple requests from running at the same time time.Sleep(1111 * time.Millisecond) } } // monitor monitors a single service in a loop func monitor(service *core.Service) { cfg := config.Get() for { // By placing the lock here, we prevent multiple services from being monitored at the exact same time, which // could cause performance issues and return inaccurate results monitoringMutex.Lock() if cfg.Debug { log.Printf("[watchdog][monitor] Monitoring serviceName=%s", service.Name) } result := service.EvaluateConditions() metric.PublishMetricsForService(service, result) serviceResultsMutex.Lock() serviceResults[service.Name] = append(serviceResults[service.Name], result) if len(serviceResults[service.Name]) > 20 { serviceResults[service.Name] = serviceResults[service.Name][1:] } serviceResultsMutex.Unlock() var extra string if !result.Success { extra = fmt.Sprintf("responseBody=%s", result.Body) } log.Printf( "[watchdog][monitor] Monitored serviceName=%s; success=%v; errors=%d; requestDuration=%s; %s", service.Name, result.Success, len(result.Errors), result.Duration.Round(time.Millisecond), extra, ) alerting.Handle(service, result) if cfg.Debug { log.Printf("[watchdog][monitor] Waiting for interval=%s before monitoring serviceName=%s again", service.Interval, service.Name) } monitoringMutex.Unlock() time.Sleep(service.Interval) } }