feat(metrics) upscale error ratio metrics from 1h to 6h

This MR continues building up the infrastructure we need to upscale metrics from 1h to 6h for certain high cardinality SLIs.

This MR makes the following changes.

  1. We no longer maintain operation rate and error rate metrics for 6h periods. Only the error ratio for 6h. This is not a problem since we do not consume either of these metrics in any dashboards or alerts.
  2. We update the error ratio to be composed of intermediate/source metrics at 6h (for metrics that we do not use upscaling) or upscaled 1h metrics in the case that we do upscale. This is exactly the same approach that we used for apdex upscaling.
  3. The only service using upscaling continues to be camoproxy. This is because it is a relatively low risk service. Once we have confirmed that this is working as expected, we will roll it out across higher-cardinality problematic services, such as patroni.
Edited by Andrew Newdigate

Merge request reports

Loading