[m3nsch] support generating new unique metrics

[schallert]

m3nsch supports generating load for a defined set of metrics, but does
not currently support generating new uniques over time. This allows
specifying a percentage of generated metrics that will be unique.

[codecov]

Codecov Report

Merging #1397 into master will decrease coverage by <.1%.
The diff coverage is 66.6%.

@@           Coverage Diff            @@
##           master   #1397     +/-   ##
========================================
- Coverage    70.7%   70.7%   -0.1%     
========================================
  Files         827     827             
  Lines       71238   71233      -5     
========================================
- Hits        50411   50386     -25     
- Misses      17521   17541     +20     
  Partials     3306    3306

Flag	Coverage Δ
#aggregator	82.3% <ø> (-0.1%)	⬇️
#cluster	85.8% <ø> (-0.1%)	⬇️
#collector	63.7% <ø> (ø)	⬆️
#dbnode	80.7% <ø> (-0.1%)	⬇️
#m3em	73.2% <ø> (ø)	⬆️
#m3ninx	74.2% <ø> (-0.1%)	⬇️
#m3nsch	51.3% <66.6%> (+0.1%)	⬆️
#metrics	17.6% <ø> (ø)	⬆️
#msg	74.9% <ø> (ø)	⬆️
#query	65.2% <ø> (-0.1%)	⬇️
#x	76% <ø> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 667f864...9abb2e9. Read the comment docs.

[codecov]

Codecov Report

Merging #1397 into master will increase coverage by <.1%.
The diff coverage is 50%.

@@           Coverage Diff            @@
##           master   #1397     +/-   ##
========================================
+ Coverage    70.8%   70.8%   +<.1%     
========================================
  Files         832     832             
  Lines       71400   71404      +4     
========================================
+ Hits        50577   50596     +19     
+ Misses      17516   17505     -11     
+ Partials     3307    3303      -4

Flag	Coverage Δ
#aggregator	82.3% <ø> (ø)	⬆️
#cluster	85.8% <ø> (-0.1%)	⬇️
#collector	63.7% <ø> (ø)	⬆️
#dbnode	80.8% <ø> (ø)	⬆️
#m3em	73.2% <ø> (ø)	⬆️
#m3ninx	74.2% <ø> (ø)	⬆️
#m3nsch	51.1% <50%> (-0.1%)	⬇️
#metrics	17.6% <ø> (ø)	⬆️
#msg	74.9% <ø> (ø)	⬆️
#query	65.5% <ø> (-0.1%)	⬇️
#x	76% <ø> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 985eb5f...bab55e8. Read the comment docs.

[prateek]

lol why?

[schallert]

It was causing lint errors or some similar build error when I was testing and I wanted to ignore it, I'll loop back and try to figure out what that was

[prateek]

LGTM save the one question

[schallert]

Looks like something's off with my approach here. The below screenshot is from runs with u equal to 0.5, 1.0, and 0.2 respectively. The general goal is accomplished, in that uniques scale with u, but it's not the 1:1 behavior I was hoping for. I would have expected to see 3k unique/s when u=1.0. Not sure if I'm misunderstanding the write behavior or if my numbers are bad.

Admittedly I don't want to sink a lot of time into this if the general spirit is accomplished, so may merge as is after a bit more testing.

[prateek]

@schallert yep the behaviour you're seeing is expected. The agent has a bunch of "workers" (goroutines) which are assigned a set of metrics to generate traffic for every time the workload changes - 1, 2. Subsequently, in each worker's event loop they just use the metrics they're assigned to create writes - 3.

Your change is only going thru the code-path at 1/2. i.e. You're only creating new uniques when the workload is updated. You'll need to update the logic to be triggered during 3 if you want the uniques to be consistently generated.