Drop expiration in the alert cache

The cache in `AlertGetter` might not be that useful because it expires mostly before the next call happens and it could lead to stale data. The cache is kept to preserve the presence of an alert to avoid breaking if an attempt to retrieve alerts hiccuped. We rely on the existing `minimumUpdateCheckInterval` to avoid bombarding the OpenShift monitoring.

We may come back to figure out a more efficient method to get the alerts instead of polling periodically if performance becomes an issue.

Author: hongkailiu

pkg/clusterconditions/promql/alerts.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"fmt"
 	"sync"
-	"time"
 
 	"github.com/prometheus/client_golang/api"
 	prometheusv1 "github.com/prometheus/client_golang/api/prometheus/v1"
@@ -26,23 +25,19 @@ func NewAlertGetter(promQLTarget clusterconditions.PromQLTarget) Getter {
 	if !ok {
 		panic("invalid condition type")
 	}
-	return &ocAlertGetter{promQL: v, expiration: 1 * time.Minute}
+	return &ocAlertGetter{promQL: v}
 }
 
 type ocAlertGetter struct {
 	promQL *PromQL
 
-	mutex       sync.Mutex
-	cached      prometheusv1.AlertsResult
-	expiration  time.Duration
-	lastRefresh time.Time
+	mutex  sync.Mutex
+	cached prometheusv1.AlertsResult
 }
 
 func (o *ocAlertGetter) Get(ctx context.Context) prometheusv1.AlertsResult {
-	if time.Now().After(o.lastRefresh.Add(o.expiration)) {
-		if err := o.refresh(ctx); err != nil {
-			klog.Errorf("Failed to refresh alerts, using stale cache instead: %v", err)
-		}
+	if err := o.refresh(ctx); err != nil {
+		klog.Errorf("Failed to refresh alerts, using stale cache instead: %v", err)
 	}
 	return o.cached
 }
@@ -89,7 +84,6 @@ func (o *ocAlertGetter) refresh(ctx context.Context) error {
 		return fmt.Errorf("failed to get alerts: %w", err)
 	}
 	o.cached = r
-	o.lastRefresh = time.Now()
 	klog.Infof("refreshed: %d alerts", len(o.cached.Alerts))
 	return nil
 }