[SOLVED] Graph of disk and net from influxdb/grafana

[mbaldini]

Hello.

Following the guide at https://pve.proxmox.com/wiki/External_Metric_Server I configured my proxmox servers to send status data to an influxdb server, it's working and I can see nodes measurements and VM/LXC measurements, graphing them in grafana with influxdb data source. But I can't understand what is the format of fields diskread/diskwrite/netin/netout from the table system.

I can make a graph with memory stats (fields: maxmem/mem), cpu stats (fields: cpu) but for disk and net activity I cannot understand in what format are the values that proxmox send.

See an example of graphs for a Windows 2012 VM:

Spoiler

As you can see in the two graphs in the bottom row, I am receiving the right values, in fact memory and cpu graphs are working right.
In the graphs in the top row I used the fields netin, netout, diskread, diskwrite but those values have no sense to me, always growing while I think I should have activity values (bytes/sec)

This is an example of the query I used for those graphs:

Spoiler

If you can tell me how to analyze the values in those fields, it could be very useful to me.

Thanks

 

[dcsapak]

they are in bytes written/read/sent/received

if you want a graph for a rate you have to add a transformation -> non negative derivative

 

[mbaldini]

Thanks for your reply.

I just added the transformation as you said and now the graph is working as it should, the last problem is that values I see in the Grafana graph are different from the values I see in the Proxmox GUI in the same time range.

Look at examples:

This is the graph of disk usage from Proxmox GUI, at time 15:03 diskwrite is 9.67 MB/s

Spoiler

This is in Grafana, at time 15:03 diskwrite is 13.08 MB/s and it's not the graph peak as it is in the Proxmox GUI

Spoiler

In the graph peak as seen in Grafana, I get a value of 286.9 MB/s, this value seems too high to me, in fact in Proxmox GUI I don't have such a value

Spoiler

I suppose I'm still having some mistake in the query, but I can't understand what, I am not practical with the non_negative_derivative transformation so I don't know how to use that well.

Thanks

 

[dcsapak]

you have different timespans

since the source data is absolute and cumulative, there are no values for single points in time, but only for a timespan. in the gui the smallest timespan you can get is 1 minute, in your grafana example you have 10s

so if i write with 100 mbyte/s for 1 second and then 59 second nothing, the value for a 10 second interval will be 10mbyte/s for the first 10s and 0 for the rest
if my interval is 60 seconds, my value will be 1.6666 mbyte/s

 

[mbaldini]

Ok, thanks.

The interval of 60 seconds is ok for me, so I'll change the timespan in the graphs.
Thank you very much