FWIW, Prometheus v2.47.0 (just released) includes an experimental OpenTelemetry (OTLP) Ingestion endpoint. https://github.com/prometheus/prometheus/pull/12643/files
On Sunday, 3 September 2023 at 09:13:27 UTC+1 Brian Candler wrote: > It sounds like you're having problems with a third-party software which > you refer to as "an OpenTelemetry collector" > > As a starting point, can you provide a link to that project and/or its > documentation, and show how you have configured it? > > Can you also show the exact error message you see? And what version of > Prometheus you are running? > > > my question is why doesn't Prometheus use the metric timestamp which has > nanosecond precision > > Prometheus *does* preserve these timestamps with remote write. So at this > point, it sounds like a problem with the third-party software. > > (Aside: Prometheus generally ignores the timestamps when doing standard > scraping over HTTP, but the timestamp is then the time of the scrape, so > they are always different) > > On Sunday, 3 September 2023 at 07:31:24 UTC+1 John Perrin wrote: > >> Hello >> We are exporting metrics from an OpenTelemetry collector using the >> prometheusremotewrite exporter to Prometheus started with >> the --enable-feature=remote-write-receiver >> The metrics are the results of k6 http requests eg response time which >> are passed into the Opentelemetry collector and then exported to Prometheus >> and we want to ingest all the data points into Prometheus. We then use >> Grafana to aggregate and visualise the metrics. >> It generally works fine except when a metric has the same name and value >> within the same time interval inside Prometheus. In this case Prometheus >> rejects the metric as a duplicate when Opentelemetry tries to export it and >> the metric is dropped. >> In this use case my question is why doesn't Prometheus use the metric >> timestamp which has nanosecond precision. If it did then it would be >> unlikely that 2 metrics would be duplicates. >> The other option we have had some success with is to add a random number >> tag to the metric but this doesn't always work and is not ideal. >> I'm not clear on how Prometheus creates it's internal timestamps in this >> case and wondered if there is some way to configure it to use the metric >> timestamp. >> >> -- 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/da92c106-2058-44b3-b879-88a70ccad95cn%40googlegroups.com.
