fix: peer-exchange issue

[darshankabariya]

closes #2414

This implementation will enhance the reliability of the peers shared by the peer-exchange protocol with its clients. because now the cache updates every 10 min.

After the resolution of the peer_manager issue #2905, the reliability of peer exchanges has also been improved.

[github-actions]

You can find the image built from this PR at

quay.io/wakuorg/nwaku-pr:2889

Built from faee769

[darshankabariya]

After understanding the discussion in issue #2414, try to implement @alrevuelta's proposed solution. Please review the code.

Is it possible to test this complex scenario in a unit test?
I chose three hours. Is there any comment on reduce or increase that span ?

cc @gabrielmer @chaitanyaprem

[Ivansete-status]

LGTM! Thanks! 💯
I just added a naïve question :)

[Ivansete-status]

Are we sure the units are in seconds?

[darshankabariya]

Actually, the secs and hours functions cause compilation errors. will try to move hourly formula. regarding 3 hours I will check with alvaro before merging.

[chaitanyaprem]

is it just me or something got messed up during force push?

I literally don't see any diff at all. Here is what i am seeing

[darshankabariya]

is it just me or something got messed up during force push?

I literally don't see any diff at all. Here is what i am seeing

Currently, there's nothing in this PR. I implemented Alvaro's suggested proposal, but Hanno recommended a more feasible long-term solution. so i revert older version. I'm working on that now and will update once it's done. Actually discussion going on this thread, https://discord.com/channels/1110799176264056863/1260493155791405096

[jm-clius]

This is a good start. Perhaps we can evaluate if things improve? Afaics this is a very cheap refresh, so we may even lower interval further if needed.

[darshankabariya]

This is a good start. Perhaps we can evaluate if things improve? Afaics this is a very cheap refresh, so we may even lower interval further if needed.

Thank you for your suggestion. I've already created a draft PR to analyze peer exchange behavior: #2940. The plan is to investigate whether increasing the frequency of cache refreshes helps. If it doesn't, we would be unnecessarily loading the node. soon I will present matrics of analysis using these tools and then we will decide.

cc @jm-clius

[darshankabariya]

This is a good start. Perhaps we can evaluate if things improve. Afaics this is a very cheap refresh, so we may even lower interval further if needed.

As you suggest @jm-clius
I've been developing tools to analyze peer exchange availability for the waku network. You can find the code in PR #2940

Tool Overview
This tool evaluates the effectiveness of peer exchange responses. Typically, the peer list updates every 15 minutes. In my analysis, I requested 5 peers every 2 minutes, checking their availability by attempting to connect at the P2P level.

The process involved 60 iterations, accumulating 300 peer requests over 3 hours. I generated percentage charts to visualize success rates at different cache refresh intervals. The results are as follows:

15-minute interval: 265/300
10-minute interval: 278/300
5-minute interval: 277/300

Conclusion
Increasing the frequency by decreasing the cache refresh interval appears effective only for 10 minutes. At 5 minutes, similar to 10 minutes. I recommend updating the cache refresh rate from 15 to 10 minutes.

I’d welcome any suggestions or alternative approaches to this analysis

cc @gabrielmer @chaitanyaprem @Ivansete-status @NagyZoltanPeter

[Ivansete-status]

Thanks for the analysis in #2889 (comment) @darshankabariya . It seems reasonable to set it to 10min (considering very few test iterations but it is fine.)
Make sure to enhance the CI before merging :)