How can we make chat app example alive for all time?

[pawanl1x]

The peers in the chat example are getting removed from the network after being inactive for some duration. Is there any way to make it alive for all time?

I've tried setting mDNS's TTL as Max but the new() of mDNS uses the default value of TTL in the discoverd_nodes config item.

[thomaseizinger]

You can increase the idle connection timeout on the SwarmBuilder. See here in the ping-example:

rust-libp2p/examples/ping/src/main.rs

Line 47 in 6369666

.idle_connection_timeout(Duration::from_secs(60)) // For illustrative purposes, keep idle connections alive for a minute so we can observe a few pings.

See #4306 for the tracking issue around this topic.

[thomaseizinger]

Do we have any other approach to make our connection alive?

Use them! As long as there is activity on the connections, they won't be closed automatically. Can you share your usecase of why you need connections to be kept alive beyond that? What would be the issue with re-establishing them if you need to send a message to a specific peer?

I mean related to gossipsub::publish so that every node makes itself active in the network?

I don't understand what you mean here, can you elaborate?

---

The overflow panic you are experiencing shouldn't be too hard to fix if you are interested in contributing. I comment on the issue you opened with an idea on how to!

[pawanl1x]

I mean, do we have any other option to make peers alive apart from idle_connection_timeout?
BTW, I saw your comment, we just need to wait for the release now.

Regarding the overflow issue, sure, I will take a look.

[thomaseizinger]

I mean, do we have any other option to make peers alive apart from idle_connection_timeout?

We don't have another (recommended) mechanism to keep idle connections alive. There is KeepAlive::Until but we are fading that out: #3844

Can you share your usecase why you would like to keep idle connections alive? In other words, connections that have no messages exchanged?

[pawanl1x]

My use-case is just gossip data into the network, but the issue right now is after 6 minutes peers are getting dropped. I just don't want to miss any data even if I get data after 1 hour or more than that.

[thomaseizinger]

Is this within a trusted network? i.e. are your peers definitely not malicious? Gossipsub doesn't guarantee that you are not going to miss any data. If a node crashes, it will miss all messages that are being gossiped while it reboots. If every node MUST witness all messages, you'd probably need some kind of "sync" protocol anyway that you can use to retrieve historic messages.

Also, how big is your network and how strong are your machines? Depending on the size of the network, you might be able to maintain full connectivity and don't need to gossip data but can just request it directly from other peers.