External metrics server and missing data

[Terje Solend Nomeland]

I'm currently setting up external metric and have some issue with data that I was expecting but there is no data. I'm using Grafana with influxdb and use this wiki page for guidence

https://pve.proxmox.com/wiki/External_Metric_Server

influxdb:
        server monitor.mylocaldomain.no
        port 8089

This work perfect for VMs and LXCs but not for the Host (Proxmox)



SQL I currently using in grafana

SELECT derivative(mean("netin"), 1s) AS "$host" FROM "nics" WHERE ("object" = 'qemu') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

If i try this with nodes (proxmox host) there is no data.

SELECT derivative(mean("netin"), 1s) AS "$host" FROM "nics" WHERE ("object" = 'nodes') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

or

SELECT derivative(mean("netin"), 1s) AS "$host" FROM "system" WHERE ("object" = 'nodes') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

The same issue with disk IOPS and data read/write. If I have to do modification to get this working is there possible to include data from ceph and temprature sensors? I know that using telegraf I can fix this, but feels this is kinda uncessary when proxmox have this function inluded

 

[dcsapak]

the fields have a different name for hosts
this works for example on my personal grafana dashboard:

SELECT non_negative_derivative(last("transmit"), 1s) *8 AS "out", non_negative_derivative(last("receive"), 1s) *8 AS "in" FROM "nics" WHERE ("host" = 'HOSTNAME' AND "instance" = 'bond0') AND $timeFilter GROUP BY time(1s) fill(none)

 

[Terje Solend Nomeland]

the fields have a different name for hosts
this works for example on my personal grafana dashboard:

SELECT non_negative_derivative(last("transmit"), 1s) *8 AS "out", non_negative_derivative(last("receive"), 1s) *8 AS "in" FROM "nics" WHERE ("host" = 'HOSTNAME' AND "instance" = 'bond0') AND $timeFilter GROUP BY time(1s) fill(none)

Thanks! I didn't even think about using the transmit/receive to calculate it. Had to remove * 8 to get correct value but it works!

Network transfer (in)

SELECT non_negative_derivative(last("receive"), 1s) AS "in" FROM "nics" WHERE ("object" = 'nodes' AND "instance" = 'bond0') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

Network transfer (out)

SELECT non_negative_derivative(last("transmit"), 1s)  AS "out" FROM "nics" WHERE ("object" = 'nodes' AND "instance" = 'bond0') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

If you use this SQL remember to use correct instance for your setup


If somebody have quick fix for getting read/write/IOPS from proxmox nodes I would love it.

 

[marigo]

write IOPS

SELECT derivative(mean("wr_operations"), 1s) AS "$host" FROM "blockstat" WHERE ("object" = 'qemu') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

read IOPS

SELECT derivative(mean("rd_operations"), 1s) AS "$host" FROM "blockstat" WHERE ("object" = 'qemu') AND $timeFilter GROUP BY time($__interval), "host" fill(null)

 

[oversite]

is there any way to get ceph metrics this way? I don't see any relevant fields currently, and i don't know how to add anything to the metric server.
> show measurements
name: measurements
name
----
ballooninfo
blockstat
cpustat
memory
nics
system