Open
Description
Summary
We use in our project pyrra
with a 4w window to generate burnrate alerts. However, we are not interested in the increase
recording rules. The involuntary generation of those rules causes our HTTP traffic to spike. In case of our mimir-ruler/querier
instance to fetch the data from a bucket outside our kubernetes cluster. Presumably fetching 4w weeks of data seems to overflow our cache.
Also we it seems the increase
rules are not used anywhere. At least in our project. However, we do not use Dashboard nor UI.
Describe alternatives you've considered
We considered two alternatives:
- Increasing cache size: we'll try this as a workaround but it would be best not have this rule in the first place.
- Shorten window: Some of our services are rather low traffic and we'd like to avoid shortening the windows.
Metadata
Metadata
Assignees
Labels
No labels