Gauges get expired

[hryamzik]

I have a set of constants per host and would like to not to update them as often as other metrics. So I've tried to add timestamp to them:

# TYPE node_cpu_facts gauge
node_cpu_facts { fact="cores" } 8 1459466225000
node_cpu_facts { fact="count" } 2 1459466225000
node_cpu_facts { fact="threads_per_core" } 2 1459466225000

The problem is that they expire. It does look like a bug for me.

[brian-brazil]

This is working as expected, timestamps should be avoided for all bar expert users.

What you want here is to expose without timestamps via the textfile collector. I'd also suggest using a metrics per fact, rather than having one metric with many labels.

[hryamzik]

Indeed, had to move to:

# cat /var/lib/node_exporter/ansible_info.prom 
# Ansible managed: file modified on 2016-04-01 02:27:07 by hryamzik on hryamziks-mac
# TYPE node_cpu_facts gauge
node_cpu_cores 8
node_cpu_count 2
node_cpu_threads_per_core 2

Thanks for the fast replay, I think we can close this. I would say documentation update could be very helpful for this scenario.

[brian-brazil]

Can you point me at what docs lead you astray? I'm not aware of any official project docs in this area, only my own company blogs at http://www.robustperception.io/tag/textfile-collector/

[hryamzik]

I would expect to find these recommendations here

[brian-brazil]

It's covered more at https://prometheus.io/docs/practices/naming/, there's an existing issue around cleaning up this split.

[hryamzik]

Good, thank you very much for your detailed and fast answers!

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.