Specify storage resolution

README.md

@@ -72,6 +72,7 @@ It requires a `provided.al2` environment and respects the following env vars:
 - `BUILDKITE_CLOUDWATCH_DIMENSIONS` : A comma separated list in the form of
    `Key=Value,Other=Value` containing the Cloudwatch dimensions to index metrics
    under.
+ - `BUILDKITE_CLOUDWATCH_HIGH_RESOLUTION` : Whether to enable [High-Resolution Metrics](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/publishingMetrics.html#high-resolution-metrics) which incurs additional charges.
 
 To adjust timeouts, and connection pooling in the HTTP client use the following env vars:
 
@@ -160,7 +161,9 @@ Usage of buildkite-agent-metrics:
   -endpoint string
         A custom Buildkite Agent API endpoint (default "https://agent.buildkite.com/v3")
   -interval duration
-        Update metrics every interval, rather than once
+    	  Update metrics every interval, rather than once
+  -cloudwatch-high-resolution
+        If `-interval` is less than 60 seconds send metrics to CloudWatch as [High-Resolution Metrics](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/publishingMetrics.html#high-resolution-metrics) which incurs additional charges.
   -max-idle-conns int
         Maximum number of idle (keep-alive) HTTP connections for Buildkite Agent API. Zero means no limit, -1 disables connection reuse. (default 100)
   -newrelic-app-name string

backend/cloudwatch.go

@@ -42,13 +42,17 @@ func ParseCloudWatchDimensions(ds string) ([]CloudWatchDimension, error) {
 type CloudWatchBackend struct {
 	region     string
 	dimensions []CloudWatchDimension
+	interval int64
+	enableHighResolution bool
 }
 
 // NewCloudWatchBackend returns a new CloudWatchBackend with optional dimensions
-func NewCloudWatchBackend(region string, dimensions []CloudWatchDimension) *CloudWatchBackend {
+func NewCloudWatchBackend(region string, dimensions []CloudWatchDimension, interval int64, enableHighResolution bool) *CloudWatchBackend {
 	return &CloudWatchBackend{
 		region:     region,
 		dimensions: dimensions,
+		interval: interval,
+		enableHighResolution: enableHighResolution,
 	}
 }
 
@@ -89,7 +93,7 @@ func (cb *CloudWatchBackend) Collect(r *collector.Result) error {
 	}
 
 	// Add total metrics
-	metrics = append(metrics, cloudwatchMetrics(r.Totals, nil)...)
+	metrics = append(metrics, cb.cloudwatchMetrics(r.Totals, nil)...)
 
 	for name, c := range r.Queues {
 		queueDimensions := dimensions
@@ -100,7 +104,7 @@ func (cb *CloudWatchBackend) Collect(r *collector.Result) error {
 		)
 
 		// Add per-queue metrics
-		metrics = append(metrics, cloudwatchMetrics(c, queueDimensions)...)
+		metrics = append(metrics, cb.cloudwatchMetrics(c, queueDimensions)...)
 	}
 
 	log.Printf("Extracted %d cloudwatch metrics from results", len(metrics))
@@ -120,15 +124,25 @@ func (cb *CloudWatchBackend) Collect(r *collector.Result) error {
 	return nil
 }
 
-func cloudwatchMetrics(counts map[string]int, dimensions []*cloudwatch.Dimension) []*cloudwatch.MetricDatum {
+func (cb *CloudWatchBackend) cloudwatchMetrics(counts map[string]int, dimensions []*cloudwatch.Dimension) []*cloudwatch.MetricDatum {
 	m := []*cloudwatch.MetricDatum{}
 
+	var duration int64
+	if cb.interval < 60 && cb.enableHighResolution {
+		// PutMetricData supports either normal (60s) or high frequency (1s)
+		// metrics - other values result in an error.
+		duration = 1
+	} else {
+		duration = 60
+	}
+
 	for k, v := range counts {
 		m = append(m, &cloudwatch.MetricDatum{
-			MetricName: aws.String(k),
-			Dimensions: dimensions,
-			Value:      aws.Float64(float64(v)),
-			Unit:       aws.String("Count"),
+			MetricName:        aws.String(k),
+			Dimensions:        dimensions,
+			Value:             aws.Float64(float64(v)),
+			Unit:              aws.String("Count"),
+			StorageResolution: &duration,
 		})
 	}
 

lambda/main.go

@@ -32,6 +32,7 @@ const (
 
 var (
 	nextPollTime time.Time
+	lastPollTime time.Time
 )
 
 func main() {
@@ -55,6 +56,8 @@ func Handler(ctx context.Context, evt json.RawMessage) (string, error) {
 	clwDimensions := os.Getenv("BUILDKITE_CLOUDWATCH_DIMENSIONS")
 	quietString := os.Getenv("BUILDKITE_QUIET")
 	quiet := quietString == "1" || quietString == "true"
+	enableHighResolutionString := os.Getenv("BUILDKITE_CLOUDWATCH_HIGH_RESOLUTION")
+	enableHighResolution := enableHighResolutionString == "1" || enableHighResolutionString == "true"
 	timeout := os.Getenv("BUILDKITE_AGENT_METRICS_TIMEOUT")
 	maxIdleConns := os.Getenv("BUILDKITE_AGENT_METRICS_MAX_IDLE_CONNS")
 
@@ -146,7 +149,7 @@ func Handler(ctx context.Context, evt json.RawMessage) (string, error) {
 		if err != nil {
 			return "", err
 		}
-		metricsBackend = backend.NewCloudWatchBackend(awsRegion, dimensions)
+		metricsBackend = backend.NewCloudWatchBackend(awsRegion, dimensions, int64(time.Since(lastPollTime).Seconds()), enableHighResolution)
 	}
 
 	// minimum res.PollDuration across collectors
@@ -177,6 +180,7 @@ func Handler(ctx context.Context, evt json.RawMessage) (string, error) {
 		}
 	}
 
+	lastPollTime = time.Now()
 	log.Printf("Finished in %s", time.Since(startTime))
 
 	// Store the next acceptable poll time in global state

main.go

@@ -38,6 +38,7 @@ func main() {
 		prometheusPath = flag.String("prometheus-path", "/metrics", "Prometheus metrics transport path")
 		clwRegion      = flag.String("cloudwatch-region", "", "AWS Region to connect to, defaults to $AWS_REGION or us-east-1")
 		clwDimensions  = flag.String("cloudwatch-dimensions", "", "Cloudwatch dimensions to index metrics under, in the form of Key=Value, Other=Value")
+		clwHighResolution = flag.Bool("cloudwatch-high-resolution", false, "Send metrics at a high-resolution, which incurs extra costs")
 		gcpProjectID   = flag.String("stackdriver-projectid", "", "Specify Stackdriver Project ID")
 		nrAppName      = flag.String("newrelic-app-name", "", "New Relic application name for metric events")
 		nrLicenseKey   = flag.String("newrelic-license-key", "", "New Relic license key for publishing events")
@@ -86,8 +87,7 @@ func main() {
 			fmt.Println(err)
 			os.Exit(1)
 		}
-		metricsBackend = backend.NewCloudWatchBackend(region, dimensions)
-
+		metricsBackend = backend.NewCloudWatchBackend(region, dimensions, int64(interval.Seconds()), *clwHighResolution)
 	case "statsd":
 		metricsBackend, err = backend.NewStatsDBackend(*statsdHost, *statsdTags)
 		if err != nil {