chore: pruning excess in relay connections

[gabrielmer]

Description

Pruning excess relay connections to respect our relay connection targets

Issue

closes #2929

[github-actions]

You can find the image built from this PR at

quay.io/wakuorg/nwaku-pr:2965

Built from 5be623b

[Ivansete-status]

LGTM! Thanks for it! 💯
As mentioned somewhere, we would need to check with DST and maybe manually deploy into waku.test or waku.sandbox and see how behaves the node.
The original purpose of the issue was to prevent relay occupying the whole libp2p available connections.

One possible option could be having 1/3 for in, 1/3 for out, and 1/3 for "others"

[gabrielmer]

One possible option could be having 1/3 for in, 1/3 for out, and 1/3 for "others"

So we have max-relay-peers and max-connections parameters.

I understand that if we configure max-connections to be a bigger number than max-relay-peers then it's enough. Not sure if we need to change something? I think that it's just a matter on the config parameters that we use to run a node.

As mentioned somewhere, we would need to check with DST and maybe manually deploy into waku.test or waku.sandbox and see how behaves the node

I did test it in a normal node. What I was thinking is to merge it if tests pass, it will go through our regular regression testing and we will figure out if there's an issue. And if so, we open a specific issue for it and not leave it commented without any info. Although I think it shouldn't bring issues.

We can ask to go through explicit testing by DST before merging, but I think that in case this brings an issue, it will be thoroughly tested by all parties and we will know it before this actually gets released. Not sure if we should add this extra step or not.

LMK how you prefer to proceed :)

[gabrielmer]

Oh nice, tests failed! So will start looking at that first :)

Turning the PR into draft

[gabrielmer]

Well the CI was a false alarm, it failed on a non-related RLN test and then was never able to reproduce it. So it's ready for review :)

@AlbertoSoutullo ran a simulation with the image and there's no connectivity issues