On Thu, Feb 27, 2014 at 02:14:32PM +0100, Johan Kooijman wrote:
Dan,
How about storing the rx_byte per 5 minutes in the engine DB? That way a
reset of the counters has a minimal impact and analytics as "traffic for VM
x in month Y" could be made.
Another approach could be to have iptables keep the count?
I do not see how iptables can help here. On the host, Vdsm reports how
much traffic did VM x consume. However, when the VM is migrated to
another host, the accounting at the destination are reset.
If Engine simply copied the value, you can start month Y with 10GiB of
traffic, and end it with 7KiB.
This should be solved by Engine, probably by "banking" the Vdsm-reported
values on certain occasions (vm shutdown, migration, counter reset), and
exposing only the accumulated result.
Dan.