PrometheusPushGateway cuts off current valid metrics

[vkorchik]

Hi,
I am using Kamon version 2.1.9

I have found it very strange but PrometheusPushGatewayReporter is created with a hardcoded stale period of 365 days, and this forces all my current metrics to be cleaned here. Here we have from debug:

stalePeriod = PT8760H // which is 365 days
currentInstant = 2024-12-18T00:00:00Z
cutoff = 2023-12-19T00:00:00Z

So here i will give you my config file, and the way i emit metrics.

// application.conf

kamon {
  prometheus {
    start-embedded-http-server = no
    pushgateway.api-url = "http://localhost:9091/metrics/job/example"
  }

  modules {
    prometheus-reporter.enabled = false
    pushgateway-reporter.enabled = true
  }
}


// Main.scala

Kamon.init()
Kamon.counter("example-counter-1").withoutTags().increment(1)
Kamon.counter("example-counter-1").withoutTags().increment(1)

PS:
The usual PrometheusReporter works well, since stale period there is 5 years as well as period, so cutoff date is basically the now date.

[muej]

I faced this issue too and have tracked it down to be an issue in the PeriodSnapshot.Accumulator and I guess this change was not applied here PrometheusPushgatewayReporter.scala#L34.

[muej]

If it is of any help, my configuration can be found on gitter

[muej]

#822 seems to be related.

[vkorchik]

@muej, yes you are right, the problem is exactly in instantiating the snapshot accumulator.
But PR is there for more than 6 months, hope it will be merged anytime soon.

[SimunKaracic]

Fixed in version 2.1.14!