Blinking k6 metrics

Hi, I’m still a newbie in Grafana/k6 world.
I wrote a K6 script that requests for tokens then for licenses and creates it’s own metrics :

const token_req_duration   = new Trend('token_req_duration', true) ;
const license_req_duration = new Trend('license_req_duration', true) ;

Each VU gets a token and then for a license and sleeps 1sec.
Since I have several VUs, I should get more than 2 metrics per second.

The metrics are sent to a distant prometheus using “experimental-prometheus-rw”.

In order to check that my metrics are seen by prometheus, I run this command :

kubectl -n <namespace> exec name -- sh -c curl -sS -G <datasrc>/api/v1/label/__name__/values

and I look for my metrics in the result.

And that’s where trouble comes :
If run my command again and again, I can see that my metrics are presents at a time, and are missing the next time. Sometimes they are presents for a few minutes and disappear for an other few minutes.
I can’t see no pattern about their presence…

And when I “explore” them in Grafana, I get the same behavior in the query builder : if I enter “k6_” sometimes it autocomplete with my metrics, and sometimes not.

I am doing something wrong ? Is it nominal, and I should build a sophisticated query that smooths the metrics ?

Thanks in advance and sorry if I don’t post at the right place…

Hi @captainflam,

I would recommend graphing this two metrics and that will probably better show the possible problems.

Hopefully that will help us figure out what is happening.