I believe the following is true as well: - if prometheus does a scrape and a metric is missing (in the exporter output) which was present in a previous scrape, it's immediately marked as "stale", i.e. a staleness marker is inserted into the timeseries.
- however in your case, you're turning off the prometheus server, so there's no scraping taking place. You just get a point in the timeseries at the time of the last scrape before prometheus shut down, and then at the first scrape after prometheus starts up. There is no indication within the timeseries data itself that anything is missing; therefore, PromQL queries will look back up to 5 minutes. On Wednesday, 19 January 2022 at 07:59:57 UTC Brian Candler wrote: > Yes. Google "Prometheus Staleness". > > In short: the value of a metric at time T is the most recent value > recorded *on or before time T*, and prometheus will look backwards in > time up to --query.lookback-delta (default 5 minutes) to find this value. > > On Tuesday, 18 January 2022 at 23:54:07 UTC [email protected] wrote: > >> Hi, >> >> I just run Prometheus (2.32.0-beta.0) without changing any CLI flags. If >> I shut it down for <5 Mins, I don't see a gap in my graphs. If I shut it >> down for > 5Mins, I see a gap. Is this valid? >> >> Thanks, >> Teja >> [image: ppk.PNG] > > -- You received this message because you are subscribed to the Google Groups "Prometheus Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/f5c32907-d425-4369-a705-b21043f0b2a6n%40googlegroups.com.
