From c50f63ad3570ec56b1e2870bae787be545f39211 Mon Sep 17 00:00:00 2001
From: Adel Haj Hassan <41540817+adel121@users.noreply.github.com>
Date: Wed, 8 Nov 2023 16:43:12 +0100
Subject: [PATCH] [pkg/util/kubernetes][CONTINT-477] add
 rate_limit_queries_remaining_min telemetry in cluster agent external metrics
 server (#20497)

* add rate_limit_queries_remaining_min telemetry in cluster agent external metrics server

* added release note

* add mutex lock to mrr struct

* improve unit test

* add a get() method to the minTracker struct

* add lock on the get method
---
 .../kubernetes/autoscalers/datadogexternal.go | 29 ++++++++++++
 .../autoscalers/datadogexternal_util.go       | 47 +++++++++++++++++++
 .../autoscalers/datadogexternal_util_test.go  | 47 +++++++++++++++++++
 ...aining_min_telemetry-233fcfdbc0fe3822.yaml | 11 +++++
 4 files changed, 134 insertions(+)
 create mode 100644 pkg/util/kubernetes/autoscalers/datadogexternal_util.go
 create mode 100644 pkg/util/kubernetes/autoscalers/datadogexternal_util_test.go
 create mode 100644 releasenotes-dca/notes/add-rate_limit_queries_remaining_min_telemetry-233fcfdbc0fe3822.yaml

diff --git a/pkg/util/kubernetes/autoscalers/datadogexternal.go b/pkg/util/kubernetes/autoscalers/datadogexternal.go
index d2c313dc01143..c00e8d59ad79e 100644
--- a/pkg/util/kubernetes/autoscalers/datadogexternal.go
+++ b/pkg/util/kubernetes/autoscalers/datadogexternal.go
@@ -13,11 +13,13 @@ import (
 	"math"
 	"strconv"
 	"strings"
+	"sync"
 	"time"
 
 	"gopkg.in/zorkian/go-datadog-api.v2"
 	utilserror "k8s.io/apimachinery/pkg/util/errors"
 
+	"github.com/DataDog/datadog-agent/pkg/config"
 	"github.com/DataDog/datadog-agent/pkg/telemetry"
 	le "github.com/DataDog/datadog-agent/pkg/util/kubernetes/apiserver/leaderelection/metrics"
 	"github.com/DataDog/datadog-agent/pkg/util/log"
@@ -36,6 +38,9 @@ var (
 	rateLimitsRemaining = telemetry.NewGaugeWithOpts("", "rate_limit_queries_remaining",
 		[]string{"endpoint", le.JoinLeaderLabel}, "number of queries remaining before next reset",
 		telemetry.Options{NoDoubleUnderscoreSep: true})
+	rateLimitsRemainingMin = telemetry.NewGaugeWithOpts("", "rate_limit_queries_remaining_min",
+		[]string{"endpoint", le.JoinLeaderLabel}, "minimum number of queries remaining before next reset observed during an expiration interval of 2*refresh period",
+		telemetry.Options{NoDoubleUnderscoreSep: true})
 	rateLimitsReset = telemetry.NewGaugeWithOpts("", "rate_limit_queries_reset",
 		[]string{"endpoint", le.JoinLeaderLabel}, "number of seconds before next reset",
 		telemetry.Options{NoDoubleUnderscoreSep: true})
@@ -61,6 +66,21 @@ const (
 	queryEndpoint = "/api/v1/query"
 )
 
+var (
+	minRemainingRequestsTracker *minTracker
+	once                        sync.Once
+)
+
+func getMinRemainingRequestsTracker() *minTracker {
+	once.Do(func() {
+		refreshPeriod := config.Datadog.GetInt("external_metrics_provider.refresh_period")
+		expiryDuration := 2 * refreshPeriod
+		minRemainingRequestsTracker = newMinTracker(time.Duration(time.Duration(expiryDuration) * time.Second))
+	})
+
+	return minRemainingRequestsTracker
+}
+
 // isRateLimitError is a helper function that checks if the received error is a rate limit error
 func isRateLimitError(err error) bool {
 	if err == nil {
@@ -171,6 +191,15 @@ func (p *Processor) queryDatadogExternal(ddQueries []string, timeWindow time.Dur
 		}
 	}
 
+	// Update rateLimitsRemainingMin metric
+	updateMap := p.datadogClient.GetRateLimitStats()
+	queryLimits := updateMap[queryEndpoint]
+	newVal, err := strconv.Atoi(queryLimits.Remaining)
+	if err == nil {
+		getMinRemainingRequestsTracker().update(newVal)
+		rateLimitsRemainingMin.Set(float64(minRemainingRequestsTracker.get()), queryEndpoint, le.JoinLeaderLabel)
+	}
+
 	return processedMetrics, nil
 }
 
diff --git a/pkg/util/kubernetes/autoscalers/datadogexternal_util.go b/pkg/util/kubernetes/autoscalers/datadogexternal_util.go
new file mode 100644
index 0000000000000..b888a386c6432
--- /dev/null
+++ b/pkg/util/kubernetes/autoscalers/datadogexternal_util.go
@@ -0,0 +1,47 @@
+// Unless explicitly stated otherwise all files in this repository are licensed
+// under the Apache License Version 2.0.
+// This product includes software developed at Datadog (https://www.datadoghq.com/).
+// Copyright 2017-present Datadog, Inc.
+
+//go:build kubeapiserver
+
+package autoscalers
+
+import (
+	"sync"
+	"time"
+)
+
+type minTracker struct {
+	sync.Mutex
+	val            int
+	timestamp      time.Time
+	expiryDuration time.Duration
+}
+
+func newMinTracker(expiryDuration time.Duration) *minTracker {
+	return &minTracker{
+		val:            -1,
+		timestamp:      time.Now(),
+		expiryDuration: expiryDuration,
+	}
+}
+
+func (mt *minTracker) update(newVal int) {
+	mt.Lock()
+	defer mt.Unlock()
+
+	isSet := mt.val >= 0
+	hasExpired := time.Since(mt.timestamp) > mt.expiryDuration
+
+	if newVal <= mt.val || !isSet || hasExpired {
+		mt.val = newVal
+		mt.timestamp = time.Now()
+	}
+}
+
+func (mt *minTracker) get() int {
+	mt.Lock()
+	defer mt.Unlock()
+	return mt.val
+}
diff --git a/pkg/util/kubernetes/autoscalers/datadogexternal_util_test.go b/pkg/util/kubernetes/autoscalers/datadogexternal_util_test.go
new file mode 100644
index 0000000000000..e9781496bf39a
--- /dev/null
+++ b/pkg/util/kubernetes/autoscalers/datadogexternal_util_test.go
@@ -0,0 +1,47 @@
+// Unless explicitly stated otherwise all files in this repository are licensed
+// under the Apache License Version 2.0.
+// This product includes software developed at Datadog (https://www.datadoghq.com/).
+// Copyright 2017-present Datadog, Inc.
+
+//go:build kubeapiserver
+
+package autoscalers
+
+import (
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestUpdateMinTracker(t *testing.T) {
+	expiryDuration := 60 * time.Second
+
+	mt := newMinTracker(expiryDuration)
+
+	// Should update
+	mt.update(10)
+	assert.Equal(t, mt.get(), 10)
+
+	// Should not update, since value didn't expire yet
+	mt.update(11)
+	assert.Equal(t, mt.get(), 10)
+
+	// simulate waiting half the expirationDuration
+	mt.timestamp = time.Now().Add(-expiryDuration / 2)
+
+	// Should not update
+	mt.update(199)
+	assert.Equal(t, mt.get(), 10)
+
+	// Shoud update, even if value didn't expire because new value is lower
+	mt.update(5)
+	assert.Equal(t, mt.get(), 5)
+
+	// Change timestamp to simulate expiration
+	mt.timestamp = time.Now().Add(-2 * expiryDuration)
+
+	// Shoud update because current value has expired
+	mt.update(100)
+	assert.Equal(t, mt.get(), 100)
+}
diff --git a/releasenotes-dca/notes/add-rate_limit_queries_remaining_min_telemetry-233fcfdbc0fe3822.yaml b/releasenotes-dca/notes/add-rate_limit_queries_remaining_min_telemetry-233fcfdbc0fe3822.yaml
new file mode 100644
index 0000000000000..d76c197d2e94d
--- /dev/null
+++ b/releasenotes-dca/notes/add-rate_limit_queries_remaining_min_telemetry-233fcfdbc0fe3822.yaml
@@ -0,0 +1,11 @@
+# Each section from every release note are combined when the
+# CHANGELOG-DCA.rst is rendered. So the text needs to be worded so that
+# it does not depend on any information only available in another
+# section. This may mean repeating some details, but each section
+# must be readable independently of the other.
+#
+# Each section note must be formatted as reStructuredText.
+---
+features:
+  - |
+    Report `rate_limit_queries_remaining_min` telemetry from `external-metrics` server.