Metric Server data documentation

U

Undergrid

Well-Known Member
Jul 15, 2018
31
3
48
49
Is there documentation about what the metrics stored by the metrics server mean and the units they are in? Things like diskread, diskwrite, netin and netout etc
 
Last edited:
no it's sadly not really documented, but most of them should be in the most basic unit available, e.g. bytes or bytes/s
is there anything particular that is unclear?
 
dcsapak said:
no it's sadly not really documented, but most of them should be in the most basic unit available, e.g. bytes or bytes/s
is there anything particular that is unclear?
Click to expand...

After a whole bunch of testing, I'm pretty sure that the receive and trasmit values in the nics measurement set are recorded in bits/s,
otherwise I'm able to do 10Gb/s speed over a 1Gb/s network interface.

The diskread and diskwrite I'm not so sure of. Even in bits/s it seems to be over reporting speeds. A dd reporting 330MB/s on a LXC reports speeds of 4Gb/s or 500MB/s but I guess there could be some buffering going on there. If it's Bytes/s, it's reporting speeds faster than the physical hardware can actually perform.
 
Regarding diskread and diskwrite, the Summary tab in web front shows Mb/s, but the metric I receive from metrics server (influxdb) are in e+12. even if I do calculations I am ending up with Tb/s. Looks like these values are aggregated? (running proxmox 7.4-3)
 
yes, these data points are aggregated, so in your monitoring/graphing you must transform the data (e.g. using 'positive derivative' IIRC ) to get a 'per seconds' graph
 
You must log in or register to reply here.
Top Bottom