feat(alerting): adds support for retries

pkg/models/alerts.go

@@ -113,6 +113,7 @@ type AlertJob struct {
 	Offset  int64
 	Delay   bool
 	Running bool
+	Retry   int
 	Rule    AlertRule
 }
 

pkg/models/alerts_state.go

@@ -27,6 +27,7 @@ var (
 	AlertStateCritical     = "CRITICAL"
 	AlertStateAcknowledged = "ACKNOWLEDGED"
 	AlertStateMaintenance  = "MAINTENANCE"
+	AlertStatePending      = "PENDING"
 )
 
 func (this *UpdateAlertStateCommand) IsValidState() bool {

pkg/services/alerting/alerting.go

@@ -1,6 +1,7 @@
 package alerting
 
 import (
+	"fmt"
 	"time"
 
 	"github.com/grafana/grafana/pkg/bus"
@@ -9,6 +10,10 @@ import (
 	"github.com/grafana/grafana/pkg/setting"
 )
 
+var (
+	MaxRetries = 3
+)
+
 func Init() {
 	if !setting.AlertingEnabled {
 		return
@@ -70,6 +75,7 @@ func (scheduler *Scheduler) updateJobs(alertRuleFn func() []m.AlertRule) {
 		} else {
 			job = &m.AlertJob{
 				Running: false,
+				Retry:   0,
 			}
 		}
 
@@ -104,18 +110,32 @@ func (scheduler *Scheduler) executor(executor Executor) {
 
 func (scheduler *Scheduler) handleResponses() {
 	for response := range scheduler.responseQueue {
-		log.Info("Response: alert(%d) status(%s) actual(%v) running(%v)", response.Id, response.State, response.ActualValue, response.AlertJob.Running)
+		log.Info("Response: alert(%d) status(%s) actual(%v) retry(%d) running(%v)", response.Id, response.State, response.ActualValue, response.AlertJob.Retry, response.AlertJob.Running)
 		response.AlertJob.Running = false
 
-		cmd := &m.UpdateAlertStateCommand{
+		if response.State == m.AlertStatePending {
-			AlertId:  response.Id,
+			response.AlertJob.Retry++
-			NewState: response.State,
+			if response.AlertJob.Retry > MaxRetries {
-			Info:     response.Description,
+				response.State = m.AlertStateCritical
+				response.Description = fmt.Sprintf("Failed to run check after %d retires", MaxRetries)
+				scheduler.saveState(response)
+			}
+		} else {
+			response.AlertJob.Retry = 0
+			scheduler.saveState(response)
 		}
+	}
+}
 
-		if err := bus.Dispatch(cmd); err != nil {
+func (scheduler *Scheduler) saveState(response *m.AlertResult) {
-			log.Error(2, "failed to save state %v", err)
+	cmd := &m.UpdateAlertStateCommand{
-		}
+		AlertId:  response.Id,
+		NewState: response.State,
+		Info:     response.Description,
+	}
+
+	if err := bus.Dispatch(cmd); err != nil {
+		log.Error(2, "failed to save state %v", err)
 	}
 }
 
@@ -129,7 +149,7 @@ func (scheduler *Scheduler) measureAndExecute(exec Executor, job *m.AlertJob) {
 	case <-time.After(time.Second * 5):
 		scheduler.responseQueue <- &m.AlertResult{
 			Id:       job.Rule.Id,
-			State:    "timed out",
+			State:    m.AlertStatePending,
 			Duration: float64(time.Since(now).Nanoseconds()) / float64(1000000),
 			AlertJob: job,
 		}

pkg/services/alerting/executor.go

@@ -81,7 +81,7 @@ func (this *ExecutorImpl) Execute(job *m.AlertJob, responseQueue chan *m.AlertRe
 	response, err := b.GetSeries(job)
 
 	if err != nil {
-		responseQueue <- &m.AlertResult{State: "PENDING", Id: job.Rule.Id, AlertJob: job}
+		responseQueue <- &m.AlertResult{State: m.AlertStatePending, Id: job.Rule.Id, AlertJob: job}
 	}
 
 	result := this.validateRule(job.Rule, response)