Hey Brian, Thanks for the reply. I do agree that I'm not using Prometheus as intended. The rationale for my question is that I'm interested in working "around" what I would consider "limitations" of Grafana. I want to coordinate the visualization of events alongside the time series data I am collecting. By allowing Prometheus to scrape one set of numbers at specific instances and then posting zeros for others, I am creating a fake "time series" that really hosts events. Grafana will just ingest that data as any other and then present a beautiful synchronized visualization.
I got it working by getting a callback from the webserver. Once Prometheus has scraped the page, the webserver initiates the callback, I zero out the result. Best regards, - Steve P.S. I already send event data to Grafana using annotations. This gives the nice visualization, but doesn't allow the end user to easily download the event data for offline processing. On Friday, June 26, 2020 at 1:30:49 AM UTC-5, Brian Candler wrote: > > Prometheus is not an event database. If you want to show the results of > *individual* tests (or the duration of individual tests) then this isn't > the tool for you. Use a normal SQL database, or a noSQL like > ElasticSearch, InfluxDB etc. > > Prometheus is a metric database, so if you want to show the *number* or > *rate* of tests (or test failures/successes) then it will do a good job. > To do that, you should expose a counter of the number of test > runs/passes/failures, or an accumulated total of the test run times, or a > histogram of test run times. > > Then it doesn't matter how often it's scraped: the data is always > correct. The graph can show how many tests ran in a certain amount of time > by looking at how the counter has changed over that time. If you have a > histogram you can also answer questions like "what percentage of tests > completed in under 5 seconds". But you are still thinking about tests in > the aggregate, not as individual events. > > Resetting the data on a scrape is not a good idea, because it breaks as > soon as two prometheus servers scrape the same endpoint - e.g. a > high-availability pair, or a laptop scraping the data to test a new > collector. > > You mention pushgateway. That's rarely the right solution for any > application. You could use it here, if the only thing you care about is > the result and/or timestamp of the *last* test that was run. To accumulate > counters or totals between test runs, have a look at statsd_exporter > instead. > -- 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/ebd7e582-c03f-41b6-97dd-3ee2bc27d2cbo%40googlegroups.com.
