[QUESTION] Yace return metrics with zero value for stopped and terminated EC2 instances and stopped RDS about yet-another-cloudwatch-exporter HOT 3 CLOSED

A little addition to the situation. If we put
addCloudwatchTimestamp: true
to the metric config, e.g. "- name: CPUCreditBalance" - the exporter continues to export metric on the terminated instance with negative timestamp...
aws_ec2_cpucredit_balance_average{dimension_instance_id="i-0013b...",region="eu-west-1"} 0 -62135596800000
So the prometheus can't scrap such a metric with "expected timestamp or new record, got "INVALID"" error.

from yet-another-cloudwatch-exporter.

Try setting

# Passes down the flag `--recently-active PT3H` to the CloudWatch API. This will only return metrics that have been active in the last 3 hours.
# This is useful for reducing the number of metrics returned by CloudWatch, which can be very large for some services. See AWS Cloudwatch API docs for [ListMetrics](https://docs.aws.amazon.com/AmazonCloudWatch/latest/APIReference/API_ListMetrics.html) for more details.
[ recentlyActiveOnly: <boolean> ]

in your config which should limit this to ~3 hours. This is a really unfortunate CloudWatch behavior.

Regarding the addCloudwatchTimestamp behavior, you should upgrade to v0.57 which contains a bugfix for this, https://github.com/nerdswords/yet-another-cloudwatch-exporter/releases/tag/v0.57.0. The bugfix is to drop this data because it is invalid.

from yet-another-cloudwatch-exporter.

Apparently, the problem was solved after the version update to 0.59

from yet-another-cloudwatch-exporter.