Thank you. Brian. I'm worried about the blackbox_exporter is expensive or not.
I also have other jobs, other jobs are with global 30s scrape intervals and 30s evaluation invervals, when I set the special job with scrape interval 2m, will it cause any problem? " If you only want to *check* a value every 1 day for alerting, then it's perfectly fine to have an alerting rule group with a 1 day evaluation interval - even if the underlying metric is collected more frequently. The alerting rule will use the most recent value at evaluation time. " regarding this point, is it possible to set alerting rule group for a special rule or a job? On Friday, August 12, 2022 at 3:40:23 PM UTC+8 Brian Candler wrote: > Perhaps the most import question to ask is: *why* do you only want to > scrape once per day? > > If you're worrying about storage space consumption: don't. Prometheus > uses differential encoding. If you record the same metric repeatedly with > the same value, i.e. the difference between the adjacent values is zero, > then the amount of storage space used is miniscule. Even if it's not the > same value, it's extremely highly compressed. And disk space is cheap. > > If you only want to *check* a value every 1 day for alerting, then it's > perfectly fine to have an alerting rule group with a 1 day evaluation > interval - even if the underlying metric is collected more frequently. The > alerting rule will use the most recent value at evaluation time. > > If you're worried that the exporter itself is expensive to run on each > scrape, then you should do one of the other things that Stuart suggested > (i.e. using textfile collector or push gateway). On the target node use a > cron job to update the metric at whatever interval makes sense - but let > prometheus scrape it as often as it likes. But you should still scrape the > cached value every 1-2 minutes. > > If it's blackbox_exporter that you're talking about, those scrapes are > cheap. So you may as well scrape every 2 minutes. > > If you're worried about transient errors causing spurious alarms, then > adjust your alerting rules (e.g. in the simplest case, add "for: 30m" which > means that you'll only get a notification if the alert has been active for > 30m *continuously*). > > On Friday, 12 August 2022 at 08:31:09 UTC+1 Stuart Clark wrote: > >> On 12/08/2022 08:29, nina guo wrote: >> > So if with push gateway or textfile collector, we need to also to >> > customize our metrics, am I right? >> >> What do you mean? >> >> -- >> Stuart Clark >> >> -- 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/0a952d82-c1d7-4ad1-9978-20b84985b698n%40googlegroups.com.
