gatus/watchdog/watchdog.go

94 lines
2.8 KiB
Go
Raw Normal View History

2019-09-05 01:37:13 +02:00
package watchdog
import (
"encoding/json"
"fmt"
"log"
"sync"
"time"
"github.com/TwinProduction/gatus/config"
"github.com/TwinProduction/gatus/core"
"github.com/TwinProduction/gatus/metric"
)
var (
2020-11-27 00:09:01 +01:00
serviceStatuses = make(map[string]*core.ServiceStatus)
2020-11-27 00:09:01 +01:00
// serviceStatusesMutex is used to prevent concurrent map access
serviceStatusesMutex 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
)
2020-11-27 00:09:01 +01:00
// GetJSONEncodedServiceStatuses returns a list of core.ServiceStatus 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.
2020-11-27 00:09:01 +01:00
func GetJSONEncodedServiceStatuses() ([]byte, error) {
serviceStatusesMutex.RLock()
data, err := json.Marshal(serviceStatuses)
serviceStatusesMutex.RUnlock()
return data, err
}
2020-08-18 02:25:29 +02:00
// 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)
}
}
2020-08-18 02:25:29 +02:00
// monitor monitors a single service in a loop
func monitor(service *core.Service) {
cfg := config.Get()
for {
if !cfg.DisableMonitoringLock {
// 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.EvaluateHealth()
metric.PublishMetricsForService(service, result)
2020-11-27 00:09:01 +01:00
UpdateServiceStatuses(service, result)
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,
)
HandleAlerting(service, result)
if cfg.Debug {
log.Printf("[watchdog][monitor] Waiting for interval=%s before monitoring serviceName=%s again", service.Interval, service.Name)
}
if !cfg.DisableMonitoringLock {
monitoringMutex.Unlock()
}
time.Sleep(service.Interval)
}
}
2020-11-27 00:09:01 +01:00
// UpdateServiceStatuses updates the slice of service statuses
func UpdateServiceStatuses(service *core.Service, result *core.Result) {
key := fmt.Sprintf("%s_%s", service.Group, service.Name)
2020-11-27 00:09:01 +01:00
serviceStatusesMutex.Lock()
serviceStatus, exists := serviceStatuses[key]
2020-11-27 00:09:01 +01:00
if !exists {
serviceStatus = core.NewServiceStatus(service)
serviceStatuses[key] = serviceStatus
2020-11-27 00:09:01 +01:00
}
serviceStatus.AddResult(result)
serviceStatusesMutex.Unlock()
}