bug: waku_total_unique_peers metric wrong value

[AlbertoSoutullo]

Problem

Metric waku_total_unique_peers does not report correct information.

Impact

This leads to inconsistencies when trying to analyze Waku's behavior.

To reproduce

1. Create a bootstrap node without relay
2. Create 3 relay nodes and connect them through discv5 to the bootstrap node. Set them in the same shard.
3. Check metrics

Expected behavior

Depending on the implementation, but waku_total_unique_peers should never report 5, as there are only 4 nodes in the network.

Screenshots/logs

If I enter to one node and do curl -X GET http://localhost:8645/admin/v1/peers I obtain:

[{"multiaddr":"/ip4/10.42.61.17/tcp/60000/p2p/16Uiu2HAmAHhKnrcw3kdP5VddyxLvHd331uepHshzJBSDGYMrqZp8","protocols":[{"protocol":"/vac/waku/relay/2.0.0","connected":true}],"origin":"Discv5"},
{"multiaddr":"/ip4/10.42.82.42/tcp/60000/p2p/16Uiu2HAmJsu1bPtcg4r1YpX28S3EKze5Tk23PP4jQhWtZt5Gdg8W","protocols":[{"protocol":"/vac/waku/relay/2.0.0","connected":true}],"origin":"Discv5"}]

But waku_total_unique_peers is reporting a value of 5.

nwaku version/commit hash

/usr/bin/wakunode --version
version / git commit hash: v0.34.0-rc.1

[NagyZoltanPeter]

Hm, pretty strange as the metric exclude self and already managed peers.
The only problem I found that it does not follow disconnects, thus adding a peer, disconnet and re-connect will count one peer twice - my unvalidated assumption.

[AlbertoSoutullo]

Hm, pretty strange as the metric exclude self and already managed peers. The only problem I found that it does not follow disconnects, thus adding a peer, disconnet and re-connect will count one peer twice - my unvalidated assumption.

In my case, there were no disconnections.

[NagyZoltanPeter]

Hm, pretty strange as the metric exclude self and already managed peers. The only problem I found that it does not follow disconnects, thus adding a peer, disconnet and re-connect will count one peer twice - my unvalidated assumption.

In my case, there were no disconnections.

Hm, which dashboard do you use exactly, or how is your exact config?
It would be interesting to see the panel code that accumulates that particular metric.

[AlbertoSoutullo]

Hm, pretty strange as the metric exclude self and already managed peers. The only problem I found that it does not follow disconnects, thus adding a peer, disconnet and re-connect will count one peer twice - my unvalidated assumption.

In my case, there were no disconnections.

Hm, which dashboard do you use exactly, or how is your exact config? It would be interesting to see the panel code that accumulates that particular metric.

I did it locally. I am just quering that metric for a single node. It's not like I am aggregating or summing or anything. The value is directly reported by the waku node.

[NagyZoltanPeter]

Hm, pretty strange as the metric exclude self and already managed peers. The only problem I found that it does not follow disconnects, thus adding a peer, disconnet and re-connect will count one peer twice - my unvalidated assumption.

In my case, there were no disconnections.

Hm, which dashboard do you use exactly, or how is your exact config? It would be interesting to see the panel code that accumulates that particular metric.

I did it locally. I am just quering that metric for a single node. It's not like I am aggregating or summing or anything. The value is directly reported by the waku node.

Thank you.
It is even interesting there, due without disconnect it can only show the connected peers without self.