[POC] [MiqDefaults] Remove ActiveSupport dep

[NickLaMuro]

Note: Made this a "POC", since a) that is what it is, but b) I haven't validated it at all aside from the local tests on my work machine (shown below), so I think testing this in a container to see if we have an appreciable difference is worth considering.

This is a slow class to load, so just removing this class allow shaves off 4x of the slowness, for something that is just about as clean without it:

$ time ruby lib/workers/bin/heartbeat_check.rb

real    0m0.097s
user    0m0.070s
sys     0m0.021s
$ time ruby lib/workers/bin/heartbeat_check.rb

real    0m0.399s
user    0m0.270s
sys     0m0.124s

It is even faster if you run ruby with --disable-gems:

$ time ruby --disable-gems lib/workers/bin/heartbeat_check.rb

real    0m0.028s
user    0m0.016s
sys     0m0.008s

Links

• Addressing this comment: We only need the ruby environment, not full container_env manageiq-pods#687 (comment)
• PR that added this form of heartbeating: Add heartbeat_check script for file-based worker process heartbeating #15494

[miq-bot]

Checked commit NickLaMuro@bd2da27 with ruby 2.6.3, rubocop 0.82.0, haml-lint 0.35.0, and yamllint
1 file checked, 0 offenses detected
Everything looks fine. ⭐

[jrafanie]

@NickLaMuro wow, yeah, @Fryguy suggested similar things... it definitely helps... I'm trying a poc doing it in shell since it seems like we're really not using much of the options here and if it's much faster and can avoid the time parsing, it might be worth it.

[jrafanie]

Basically, dropping container_env saves around 200-300ms in (from about 1.2/1.3 baseline)
Dropping activesupport saves another 100ms.
using disable gems saves ~50 ms... it's really hard to measure small improvements in pods since performance is very inconsistent.

Dropping English is negligble.

[NickLaMuro]

@jrafanie do you have a baseline in said container you can give for the following?

(these times are from my local machine)

$ time python3 -c ""

real    0m0.031s
user    0m0.014s
sys     0m0.011s
$ time ruby -e ""

real    0m0.091s
user    0m0.064s
sys     0m0.020s
$ time ruby --disable-gems -e ""

real    0m0.019s
user    0m0.009s
sys     0m0.007s
$ time /bin/bash -c ""

real    0m0.006s
user    0m0.001s
sys     0m0.003s

I am curious if there is just a generic "container penalty" we are getting that is contributing to the times here. If /bin/bash for example executes at 500ms on your container, I would say we have a .5 second penalty we can never get around.