We've implement our own accounting for XCP, so here some insight:
1) Most of the counters are overflowing. More crucial, they going to
sub-zero values.
2) When VM migrate it appear as new domain with new counters.
3) CPU time is better to read via xc.domain_getinfo().
PS No, XCP do not have internal reliable counters for resources.
There is some rrd-based statistic, but it not provides exact
numbers.
On 28.07.2012 18:38, Thomas Goirand wrote:
Hi there, Mike, Jon and John!
As there's now the new Ceilometer project in openstack, I was wondering
few things about VM usage monitoring XCP.
In a normal dom0 running Xen, my python daemon dtc-xen has a thread that
monitor usages the following way:
- Number of CPU used by each VM using "xm list"
- Network usage of each VM using the kernel bytes counter in /proc/net/dev
- Number of sector read/write for each VM using /sys/block/dm-%s/stat
Now, we would like to have the same information using XCP's REST. Is
there a way that is already implemented in XCP? If not, could this be a
new feature to add to XCP?
Cheers,
Thomas Goirand
_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api
|
