Bryan Boreham

Pretty good since the last change, which I excerpted as #13427. (Sorting of series is slower when it has to indirect through the symbol-table, but prombench never actually needs to...

/prombench cancel

/prombench main

I don't think it's enough to assert the code "should be different". When I read the code, I see the point of the check is to check if the data...

I believe this is related: https://github.com/prometheus/prometheus/pull/12527#issuecomment-1829677659

I don't see why this wouldn't work, given some mechanism to pick a different time on different machines. This could be random, e.g. like here: https://github.com/grafana/mimir/blob/255ddaa9fb3019522a5cafca3f58f052a5cc42ca/pkg/ingester/ingester.go#L2592-L2599 It's perhaps not a...

More prior art, this time in Prometheus itself: https://github.com/prometheus/prometheus/blob/ac8abdaacda605bdcd358ec4b6d30e634a0a569a/scrape/target.go#L156-L158

I believe the current implementation aims for consistent results within a single query but not within a rule group. All targets scheduled to scrape at the same time will get...

Without having tried it: Number of samples: `count_over_time({__name__=".+"}[my_retention_period])` Number of series: `count(max_over_time({__name__=".+"}[my_retention_period]))` (replace `my_retention_period` with `14d` or however your Prometheus is configured). May be resource-intensive to execute these queries. Write...

Thanks for the feedback. 1. This is an oversight. Although point 4 relates. 2. "behave unexpectedly" - how did you expect them to behave? 3. Agreed this is a trade-off....