Prometheus push gateway only returns the last value instead of all

Question

I use the push gateway from prometheus (https://prometheus.io/docs/practices/pushing/). And I push some metrics from a React UI very often like this:

# HELP prom_react_app_availability_total Application availability counter
# TYPE prom_react_app_availability_total counter
prom_react_app_availability_total{owner="test", status="started", timestamp="2022-09-13T06:48:19.280Z"} 1

and sometimes I have a "failed" status after "started".

# HELP prom_react_app_availability_total Application availability counter
# TYPE prom_react_app_availability_total counter
prom_react_app_availability_total{owner="test", status="failed", timestamp="2022-09-13T06:52:19.280Z"} 1 

But when I take a look on the gateway's /metrics endpoint, then I only see the last value instead of all the submitted values over time. My Prometheus server collects data from this endpoint every 30s and so I lost a lot of data and maybe also the "failed", that occurred because the next push was a status="started". I thought, I see all events or a summary with increased values.

What is wrong?

Answer 1

Prometheus Pushgateway remembers only the last pushed sample per each time series by design. See these docs. If you want counting events when your app was available/unavailable, then take a look at statsd_exporter.

Side suggestion: it isn't recommended storing timestamps in metric labels, since every unique timestamp results in a new time series. This may slow down Prometheus and increase its memory usage significantly. Such a condition is named high churn rate and/or high cardinality. See also these docs.