feat: Origin of agent webpack chunks now changeable

[cwli24]

Allow the agent to operate through a proxy. The agent needs to be able to fetch its source code and send analytics to a different domain, and this solves the former. NPM is not affected.

Overview

In support of NR-152264 / Proxy Support, a configuration has been exposed that will change where the agent chunk is loaded from. In context, our default domain js-agent.newrelic.com is on many ad blocker list today which prevent us from collecting data from end users. The use case here is to circumvent that block by using a proxy to fetch agent assets. (The beacon can also be changed so that analytics data flow throw another--or the same--proxy server.)

init: {
  proxy: {
    beacon,
    assets
  }
}

Note that this applies to agent pulled from the CDN (installation methods: copy+paste or APM-injected). Agent installed via NPM should not be affected by this change, as it's irrelevant.

Related Issue(s)

NR-149705

Testing

Existing CDN / NPM agent tests ensure new code does not break default functionality. New e2e has been added that tests the chunk source switching behavior.

UPDATE: this has been merged with #673 .

[github-actions]

Asset Size Report

Merging this pull request will result in the following asset size changes:

Agent	Asset	Previous Size	New Size	Diff
lite	loader	28.35 kB / 9.82 kB (gzip)	28.54 kB / 9.89 kB (gzip)	0.66% / 0.75% (gzip)
lite	async-chunk	45.4 kB / 15.03 kB (gzip)	45.53 kB / 15.1 kB (gzip)	0.28% / 0.43% (gzip)
pro	loader	45.76 kB / 15.31 kB (gzip)	45.95 kB / 15.38 kB (gzip)	0.41% / 0.46% (gzip)
pro	async-chunk	166.12 kB / 52.32 kB (gzip)	166.12 kB / 52.32 kB (gzip)	0% / 0% (gzip)
spa	loader	52.18 kB / 17.27 kB (gzip)	52.36 kB / 17.35 kB (gzip)	0.36% / 0.46% (gzip)
spa	async-chunk	166.12 kB / 52.32 kB (gzip)	166.12 kB / 52.32 kB (gzip)	0% / 0% (gzip)
lite-polyfills	loader	98.81 kB / 31.46 kB (gzip)	99.04 kB / 31.55 kB (gzip)	0.23% / 0.29% (gzip)
lite-polyfills	async-chunk	57.86 kB / 17.24 kB (gzip)	57.95 kB / 17.32 kB (gzip)	0.16% / 0.45% (gzip)
pro-polyfills	loader	118.79 kB / 37.44 kB (gzip)	119.02 kB / 37.53 kB (gzip)	0.19% / 0.26% (gzip)
pro-polyfills	async-chunk	98.94 kB / 26.68 kB (gzip)	99.08 kB / 26.77 kB (gzip)	0.14% / 0.33% (gzip)
spa-polyfills	loader	126.77 kB / 39.57 kB (gzip)	126.99 kB / 39.67 kB (gzip)	0.18% / 0.25% (gzip)
spa-polyfills	async-chunk	113.82 kB / 31.1 kB (gzip)	113.96 kB / 31.16 kB (gzip)	0.12% / 0.21% (gzip)
worker	loader	40.46 kB / 13.69 kB (gzip)	40.65 kB / 13.78 kB (gzip)	0.47% / 0.64% (gzip)
worker	async-chunk	45.36 kB / 15.37 kB (gzip)	45.51 kB / 15.44 kB (gzip)	0.33% / 0.42% (gzip)

[codecov]

Codecov Report

Merging #659 (6da45ab) into main (2004f5f) will increase coverage by 0.13%.
The diff coverage is 82.35%.

@@            Coverage Diff             @@
##             main     #659      +/-   ##
==========================================
+ Coverage   86.63%   86.76%   +0.13%     
==========================================
  Files         135      136       +1     
  Lines        5251     5273      +22     
  Branches      720      724       +4     
==========================================
+ Hits         4549     4575      +26     
+ Misses        641      635       -6     
- Partials       61       63       +2     

Flag	Coverage Δ
unit-tests	79.05% <70.83%> (-0.09%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
src/features/page_view_event/aggregate/index.js	87.75% <ø> (+1.75%)	⬆️
src/loaders/agent.js	57.89% <ø> (ø)
src/loaders/api/apiAsync.js	36.53% <0.00%> (-6.32%)	⬇️
src/loaders/configure/configure.js	86.95% <85.71%> (+1.24%)	⬆️
src/common/config/state/init.js	92.98% <100.00%> (-0.13%)	⬇️
src/common/constants/runtime.js	97.36% <100.00%> (ø)
src/common/harvest/harvest.js	100.00% <100.00%> (ø)
src/features/ajax/aggregate/index.js	92.45% <100.00%> (+0.14%)	⬆️
src/features/metrics/aggregate/index.js	89.06% <100.00%> (+0.53%)	⬆️
src/loaders/configure/public-path.js	100.00% <100.00%> (ø)

... and 5 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[github-actions]

Last ran on September 12, 2023 14:12:55 CDT
Checking merge of (029cdfd) into main (2004f5f)

[metal-messiah]

I think we need to regroup on the config values and implementation of this. We will likely actually need a new property for both values (something like init.proxy.beacon, init.proxy.assets) instead of overwriting beacon/errorBeacon. This will be necessary for compatibility/sanity with customers reverting back to the standard beacon values in the DB (once those have been changed to a custom value, theres no easy way to get the initial value back for the customer).

I also can't help but feel in combination with #673 that we may be over-engineering all these url checks. The combo of those 2 PRs is growing the lite build by more than 5%, for something that should be relatively benign (just adding a property and overwriting the beacon and asset values if set) and unused by the large majority of customers. Validating the url could likely be accomplished by good docs to follow and less by hard-coding checks and parsing all over the agent so we should circle back on this as well.

[cwli24]

I think we need to regroup on the config values and implementation of this. We will likely actually need a new property for both values (something like init.proxy.beacon, init.proxy.assets) instead of overwriting beacon/errorBeacon. This will be necessary for compatibility/sanity with customers reverting back to the standard beacon values in the DB (once those have been changed to a custom value, theres no easy way to get the initial value back for the customer).

I also can't help but feel in combination with #673 that we may be over-engineering all these url checks. The combo of those 2 PRs is growing the lite build by more than 5%, for something that should be relatively benign (just adding a property and overwriting the beacon and asset values if set) and unused by the large majority of customers. Validating the url could likely be accomplished by good docs to follow and less by hard-coding checks and parsing all over the agent so we should circle back on this as well.

Reworked: validations removed and the input moved to a new object in the init config.