vCPU overcommitment stats
I'm trying to get some metrics and info about CPU overcommitment and if there are hosts in my cluster which are fairly unbalanced (not in a configuration kind of way, but where vms fighting to get CPU time).
I've read that other Hypervisors show this kind of data as "CPU ready" counter and that the Xen equivalent should be "runstate_full_contention" (Fraction of time that all vCPUs are runnable (that is, waiting for CPU)). Sadly this counter does not seem to move in my xen center and the netdata plugin does not track the value.
Does anyone have more info around that topic?
I think you can use the
stealmetric from netdata.
In my practice I use zabbix and track the
@jmara said in vCPU overcommitment stats:
This should be the right data
@olivierlambert yeah i thought so too is there any way to visualize this in XOA/netdata?
I don't think we fetch this one, and Netdata doesn't parse XAPI RRD (the source of XCP metrics used in XenCenter/XO).
To me, that might be a bug. Are you sure about generating enough over-commitment to trigger that metric?
@olivierlambert I'll check later this afternoon, as the loads keeps increasing through the day and let you know if the counter moves.
Should I file a feature request for XOA as support ticket to get the counter into the stats section? Might be an interesting counter for others as well.
XO 6 will provide an UI to choose which metric to display, so I don't think we'll add a deep change like this in the existing XO 5 design.
However, you can create an issue explaining you'd like to get more stats visible, yes
@olivierlambert ah good to know thats fine I'll take a look at the metrics maybe I'll write a small rrd stats pusher script as an interim solution for my usecase until XO6 is available.
Thanks for letting me know