gettting metrics from ALL guest instead of the selected instance
tibz7 opened this issue · 7 comments
I'm having the exporter installed on prometheus host.
I using the prometheus config provided in the doc. I can collect the metrics and render them on the dashboard provided as a link.
In the target I put the IP of ONE proxmox host I want to get the metrics from ( 1 out of 24)
But for some magic reason, I'm getting ALL guests metrics, including the ones that are NOT on that proxmox host!
And filtering does not work of course
Thanks for the report.
What you are observing is the intended behavior. Prometheus PVE exporter collects metrics by scraping resources via the PVE cluster API.
Ok thank you!
So it is not possible to the guest by their host?
The node
label is on the pve_guest_info
metric. In order to make that available on any other metric, it is necessary to join it to pve_guest_info
. E.g.:
pve_up * on(id) group_left(node) pve_guest_info
More examples in #54 and detailed explanation of joins in the robust perception blog.
So the solution was for me to make this
pve_cpu_usage_ratio /pve_cpu_usage_limit and on(id) pve_node_info{name="HOST"}
by the way I did not find a way to get the cpu usage ratio for the whole cluster... I tried to sum all the pve_cpu_usage_ratio
but that does not really work in the gauge, it gives a gigantic percentage
by the way I did not find a way to get the cpu usage ratio for the whole cluster... I tried to sum all the
pve_cpu_usage_ratio
but that does not really work in the gauge, it gives a gigantic percentage
For whole cluster, maybe you can try to using this
avg(pve_cpu_usage_ratio * on (id) group_left (name) pve_node_info)