> This is a question that I have asked in the past in other channels, but > I never found a solution, so I am trying again in case somebody knows of > some tool or hack, I cannot be the only one with this need! > > I have a NodeJS application that I have developed for which I really > need to add instrumentation. It is a one-shot application, not a daemon, > but it runs for a considerable amount of time, sometimes up to an hour, > and it is launched on demand by a job scheduler (separate Python app). > > There is no fixed number of jobs running in parallel, so I cannot define > pseudo-instances for each job without exploding the metric cardinality.
I believe that what you may want is the statsd exporter. The statsd exporter can be used to accumulate running metrics over time from whatever source feeds into it, with whatever labels you want to stick on, and sending metrics updates to it is pretty simple. If you want to feed in ongoing metrics while the NodeJS application is running, you'd presumably reset any in-application counters every time you sent them to the statsd exporter. This does have the drawback that you can't use the Prometheus client libraries as-is, but you may be able to find statsd libraries instead (and such libraries may already handle the counter reset issue). You want libraries that support the extended statsd format with tags. A few years ago I wrote a web page on this, with the focus on one-shot scripts: https://utcc.utoronto.ca/~cks/space/blog/sysadmin/PrometheusStatsdForMetricsUpdates - cks -- 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/20230908184505.B59AC60286%40apps0.cs.toronto.edu.
