frontend: Improve Map performance

[sniok]

This fixes a stack overflow crash #2534 and improves grouping and filtering performance

There's still a bunch of stuff that can be optimized so this is only first step

Testing done:

• Tested on KWOK cluster with 20,000 pods

After fixes:
Map loading time: 5s
Filter by error: <1s
Group by node: <1s

Before:
Map loading time: ~3.5min
Filter by error: 50s
Group by node: 9s

[skoeva]

Huge perf improvement, thanks for this!

[illume]

Very nice.

Did you look into how it reacts with a fairly high frequency of change? I mean with pods/events being added, rather than with a static set of resources.
(If not, probably something to look at in another PR.)

[illume]

👍

[sniok]

Did you look into how it reacts with a fairly high frequency of change? I mean with pods/events being added, rather than with a static set of resources. (If not, probably something to look at in another PR.)

Yeah it struggles a bit with frequent changes, but that is another issue

[illume]

Doh. I think I broke the test plugins CI job with the release of the new headlamp-k8s/eslint-config. It is trying to upgrade to the "latest" headlamp-k8s/eslint-config, where it should be getting the version related to that version of headlamp-plugin.

So you can ignore that test plugins job failure in this PR.

[sniok]

Doh. I think I broke the test plugins CI job with the release of the new headlamp-k8s/eslint-config. It is trying to upgrade to the "latest" headlamp-k8s/eslint-config, where it should be getting the version related to that version of headlamp-plugin.

So you can ignore that test plugins job failure in this PR.

Got it, thanks for the explanation

[illume]

👍