Now that we can compute metrics discretely, we should consider monitoring results to make sure there aren't any metrics missing.
This is a follow up to #3122, where we discovered that ratio metrics were simply being skipped because they are configured differently and were thus being filtered out.
Note that it might not be worth pursuing this because it's sort of a paradoxical situation:
- if we know that we've missed a metric, we probably already logged an error about it
- if we skipped a metric and don't know about it, we probably wouldn't catch it with other monitoring that would presumably use the same mechanisms as the original analysis
But filing this anyway in case we can approach the monitoring in a different way than analysis.
┆Issue is synchronized with this Jira Task
Now that we can compute metrics discretely, we should consider monitoring results to make sure there aren't any metrics missing.
This is a follow up to #3122, where we discovered that ratio metrics were simply being skipped because they are configured differently and were thus being filtered out.
Note that it might not be worth pursuing this because it's sort of a paradoxical situation:
But filing this anyway in case we can approach the monitoring in a different way than analysis.
┆Issue is synchronized with this Jira Task