The open and composable observability and data visualization platform. Visualize metrics, logs, and traces from multiple sources like Prometheus, Loki, Elasticsearch, InfluxDB, Postgres and many more.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
grafana/pkg/services/alerting/engine.go

111 lines
2.3 KiB

package alerting
import (
"time"
"github.com/benbjohnson/clock"
"github.com/grafana/grafana/pkg/log"
)
type Engine struct {
execQueue chan *Job
resultQueue chan *EvalContext
clock clock.Clock
ticker *Ticker
scheduler Scheduler
evalHandler EvalHandler
ruleReader RuleReader
log log.Logger
resultHandler ResultHandler
}
func NewEngine() *Engine {
e := &Engine{
ticker: NewTicker(time.Now(), time.Second*0, clock.New()),
execQueue: make(chan *Job, 1000),
resultQueue: make(chan *EvalContext, 1000),
scheduler: NewScheduler(),
evalHandler: NewEvalHandler(),
ruleReader: NewRuleReader(),
log: log.New("alerting.engine"),
resultHandler: NewResultHandler(),
}
return e
}
func (e *Engine) Start() {
e.log.Info("Starting Alerting Engine")
go e.alertingTicker()
go e.execDispatcher()
go e.resultDispatcher()
}
func (e *Engine) Stop() {
close(e.execQueue)
close(e.resultQueue)
}
func (e *Engine) alertingTicker() {
defer func() {
if err := recover(); err != nil {
e.log.Error("Scheduler Panic: stopping alertingTicker", "error", err, "stack", log.Stack(1))
}
}()
tickIndex := 0
for {
select {
case tick := <-e.ticker.C:
// TEMP SOLUTION update rules ever tenth tick
if tickIndex%10 == 0 {
e.scheduler.Update(e.ruleReader.Fetch())
}
e.scheduler.Tick(tick, e.execQueue)
tickIndex++
}
}
}
func (e *Engine) execDispatcher() {
for job := range e.execQueue {
e.log.Debug("Starting executing alert rule", "alert id", job.Rule.Id)
go e.executeJob(job)
}
}
func (e *Engine) executeJob(job *Job) {
defer func() {
if err := recover(); err != nil {
e.log.Error("Execute Alert Panic", "error", err, "stack", log.Stack(1))
}
}()
job.Running = true
context := NewEvalContext(job.Rule)
e.evalHandler.Eval(context)
job.Running = false
e.resultQueue <- context
}
func (e *Engine) resultDispatcher() {
for result := range e.resultQueue {
go e.handleResponse(result)
}
}
func (e *Engine) handleResponse(result *EvalContext) {
defer func() {
if err := recover(); err != nil {
e.log.Error("Panic in resultDispatcher", "error", err, "stack", log.Stack(1))
}
}()
e.log.Debug("Alert Rule Result", "ruleId", result.Rule.Id, "firing", result.Firing)
e.resultHandler.Handle(result)
}