More convenient way of reporting used and total

[bobrik]

Currently some metrics are reported like this:

# TYPE mesos_master_cpus gauge
mesos_master_cpus{type="free"} 1924.1
mesos_master_cpus{type="used"} 1579.9

and some like this:

mesos_slave_cpus{slave="slave(1)@1.2.3.4:5051"} 32
mesos_slave_cpus_used{slave="slave(1)@1.2.3.4:5051"} 2.4

I thin we should converge to the second model in a backward compatible fashion. WDYT?

[keis]

👍 for aligning the reporting style. The first seems to be the most common style currently (used by e.g mesos_slave_mem also) but the 2nd feels a bit cleaner imho.

[StephanErb]

AFAIK, sum and average over the label-dimensions of a metric should make sense. The second approach therefore seems to be the way to go.

Maybe @brian-brazil can enlighten us here :-)

[brian-brazil]

The 2nd way is preferred, as it's easier to work with.

[bobrik]

#8 is open for 2 weeks without attention, not sure if it's worth doing anything.