Global counters for all users?

[SibertJabbla]

Hi,

I'm experimenting with opentelemetry and prometheus. Say I would like to know how many times the application has been started today for all users aggregated (e.g. if user 1 runs it twice, and user 2 four times, I want to see "The application has been run 6 times today"). I have got counter working, but the value stays '1' in the prometheus dashboard, no matter how many times I run the code.

Am I understanding the use of the telemetry wrong or am I missing something?

[malkia]

IMHO, metrics won't be reliable here.

Traces might work, but you need to keep them,

Best, for such situations is to simply log an event - "Started", and if you can "Stopped" (or "Crashed", if you have crash handler to do that for you).

--

But let's give it a try with metrics.

You can emit a metric up - e.g. my app is up. But you need to identify it (e.g. proper resource identification), so you may send the app name, computer name, maybe user, and now you are thinking should I send the pid - because what if two processes overlap each other?

And here is the catch, sending something like pid creates high cardinality - it might be okay for whatever system you are using to collect your metrics, but it might be harmful. There are "solutions" here but still..

But even then, all you'll get is watermarks - e.g. how many concurrent apps were running (per machine, overall, etc.)

You can try and find gaps - "Oh here, no metric was emitted, so must've been stopped"... But no - due to sampling effects, this is not something you can rely on.

All in all, metrics is not the best tool for the job - but must be something event based - like logs, or if you can trace