[BUG] YACE internal metrics does not reflect actual GetMetricData API usage costs

[donatasjocas]

Is there an existing issue for this?

• I have searched the existing issues

YACE version

0.62.1

Config file

apiVersion: v1alpha1
sts-region: eu-central-1
discovery:
exportedTagsOnMetrics:
AWS/EBS:
- Name
jobs:

• type: AWS/EBS
  regions:
  • eu-central-1
    period: 300
    length: 300
    delay: 300
    nilToZero: true
    addCloudwatchTimestamp: true
    recentlyActiveOnly: true
    roles:
  • roleArn: "arn:aws:iam::************:role/MetricsCollectorRole"
    statistics: [Minimum,Maximum,Average,Sum,SampleCount]
    metrics:
  • name: VolumeReadBytes
  • name: VolumeWriteBytes
  • name: VolumeReadOps
  • name: VolumeWriteOps
  • name: VolumeTotalReadTime
  • name: VolumeTotalWriteTime
  • name: VolumeIdleTime
  • name: VolumeQueueLength
  • name: VolumeThroughputPercentage
  • name: VolumeConsumedReadWriteOps
  • name: BurstBalance

Current Behavior

YACE internal metrics are inconsistent when it comes to actual GetMetricData API usage and associated costs.

It seems that right now internal metric yace_cloudwatch_getmetricdata_metrics_requested_total accounts for all metric/statistic combination on the export side while "metrics requested" in AWS terms should only include single metric requested (I assume up to 5 statistics as described in CloudWatch -> Metrics documentation)

Right now when using AWS Pricing Calculator for CloudWatch API costs with YACE internal metrics of yace_cloudwatch_requests_total and yace_cloudwatch_getmetricdata_metrics_requested_total results in completely off numbers that are incomparable with actual API costs.

I've build a test case in my AWS account proving this - when exporting only single statistic (Average), YACE metrics around API usage were in-line with the actual costs when put into Pricing Calculator. When I ran same exporter, but requesting 5 statistics metric yace_cloudwatch_getmetricdata_metrics_requested_total increased by a large number while the actual costs of GetMetricData API remained the same as if I were requesting 1 statistic for EBS metrics

Expected Behavior

YACE internal metric yace_cloudwatch_getmetricdata_metrics_requested_total should result with number that accounts for actual API usage costs according to pricing.

Steps To Reproduce

1. Run YACE discovery scrape requesting 1 statistic for each metric;
2. Check the internal metric yace_cloudwatch_getmetricdata_metrics_requested_total and actual costs of GetMetricData API
3. Run YACE discovery job requesting 5 statistics for each metric;
4. Check the internal metric yace_cloudwatch_getmetricdata_metrics_requested_total and actual costs of GetMetricData API
5. Compare results from step 2 and 4 to see that metric value does not stand correctly for how AWS bills the API usage

Anything else?

No response