So if I'm understanding you correctly. You have multiple pods exporting a gauge based on a remote request, and because it is load balanced the request could go to one of many pods, and you are trying to figure out which one is the most recent? This sounds like a perfect use case for pushgateway as it will have the latest value, and the metric will survive pod death.
That said, if you are married to the current pattern you have set up, the answer would be to export a corresponding timestamp metric so you can select the most recently updated pod's gauge value. I hope this helps. --Matt On Mon, Sep 4, 2023 at 3:23 PM Stuart Clark <[email protected]> wrote: > > On 25/08/2023 04:37, sunil sahu wrote: > > Hello All, > > I need help in my below use case. > > We have a centralized python application which continuously monitors state > for other applications by running tests. We creates GUAGE metrics with a > value (Ex. 100) and set to different value when state changes (Ex. 200 or > 300) using gauge's set function (python prometheus_client instrumentation). > This run frequently (1min, 5m etc) depending on tests. > > Our central application runs with multiples pods in load balancing. So any > pod can serve the request and set the metric value to new/existing one. > > We are scraping individual pods every minute using k8s discovery (role: > pods), now here any of the pod/instance can have latest/recent updated value. > Now my problem is how to get the metric with latest value out of all > instances' metrics. > > Aggregations on 'instance' label like sum/min or max won't be a ideal choice > here, may be `last` (which I don't find. I am aware of last_over_time but > that is kind of to fill the data-point gaps with previous values) > > Currently I am living with `last_over_time(max without(instance) > (metric_name))` but as I explained it is giving wrong results for flaky tests. > > > I tried to explain as best as possible, still let me know if any missing gaps > to understand the scenario. > > So are you saying that the multiple pods which are all getting scraped might > give totally different values for this metric, but only one of those is > correct? > > If that is the case then you'd need to ensure the correct value is returned > by every pod, or you store that value elsewhere and scrape that other system. > > -- > Stuart Clark > > -- > You received this message because you are subscribed to the Google Groups > "Prometheus Developers" 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-developers/37460f3a-1ef0-69b5-d3ad-02cf2f8e4154%40Jahingo.com. -- --Matt -- You received this message because you are subscribed to the Google Groups "Prometheus Developers" 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-developers/CAGyBzcxTa6-UT7KHBKdoTnPbP9nDR%2BfUArdtGqtvftE9v9DMmQ%40mail.gmail.com.
