Calculate power consumption (KWH) from prometheus wattage data

[arcety]

I’m currently trying to display the daily and monthly power consumption using the “estimated input power” metric retrieved from my PSU (hx1200i). I was able to get this metric into Grafana using the liquidctl-exporter. Now I’m trying to translate this into the total consumed power (kwh) per hour/day and month.

I found this forum post post in which a user tried to do this. However, I used Prometheus, and so I can not use the syntax used there.

To get a plot of the KWH per hour, I now use the following syntax:

avg_over_time(liquidctl_hidraw4_estimated_input_power[1h])

When I try to take the sum of this to get the total power consumption, I cannot find the correct syntax.

sum_over_time(avg_over_time(liquidctl_hidraw4_estimated_input_power[1h]))

Gives me a bad_data: 1:15: parse error: expected type range vector in call to function "sum_over_time", got instant vector error which makes sense. I therefore tried to convert my previous result into a range vector using:

sum_over_time(avg_over_time(liquidctl_hidraw4_estimated_input_power[1h])[1d])

But, I, however, get a bad_data: 1:73: parse error: ranges only allowed for vector selectors.

Therefore, does anybody know how to display the total power consumption on the Grafana dashboard based on Prometheus Power sensor data?

[arcety]

The solution turned out to be quite straight fowards:

sum_over_time(liquidctl_hidraw1_estimated_input_power[24h]) / (count_over_time(scrape_samples_scraped{job="liquidctl"}[1m]) * 60) / 1000 

In this (count_over_time(scrape_samples_scraped{job="liquidctl"}[1m]) retrieves the scrape frequency.

[gyulahalmos]

@arcety I used your suggestion and it worked great, until I had an outage in the data. Then it messed up my calculation:

So basically instead of the sum growing after the outage it wen back to 0 and started from there, even though I increased the range. Any idea how to fix this ?

Thanks in advance.