dht.provide fails if any closest peer is blacklisted

[kumavis]

reproduction / error flow

• get peer blacklisted somehow
• call dht.provide()
• blacklisted peer in getClosestPeers
• async/each over peers and call this.network.sendMessage(peer, msg, cb)
• blacklist error causes whole async/each to end with error

js-libp2p-kad-dht/src/index.js

Lines 530 to 542 in 9db17eb

	waterfall([
	(cb) => this.providers.addProvider(key, this.peerInfo.id, cb),
	(cb) => this.getClosestPeers(key.buffer, cb),
	(peers, cb) => {
	const msg = new Message(Message.TYPES.ADD_PROVIDER, key.buffer, 0)
	msg.providerPeers = peers.map((p) => new PeerInfo(p))
	each(peers, (peer, cb) => {
	this._log('putProvider %s to %s', key.toBaseEncodedString(), peer.toB58String())
	this.network.sendMessage(peer, msg, cb)
	}, cb)
	}
	], (err) => callback(err))

suggestions:

• ignore all errors?
• filter peers by not-blacklisted before async/each => sendMessage

[kumavis]

for reference my peers are usually blacklisted for "no transport for peer"

[jacobheun]

Since the network is still fairly unstable having peers in the dht that are no longer dialable is going to be an issue we need to address. The blacklisting is likely due to this.

The provide shouldn't ignore the errors, I think it should continue trying further peers until it hits kValue (20) peers. Also, if we fail to connect to the peer they should probably get removed from the table or at least penalized up to some threshold before removal.

If we don't provide to enough peers we risk findProviders having a higher failure rate for other peers.

Right now getClosestProviders returns only kValue peers. It would be great to be able to continue to pull from it until we've messaged enough peers.

[kumavis]

was also encountering this here

js-libp2p-kad-dht/src/private.js

Line 559 in ddf80fe

dht.network.sendRequest(peer, msg, callback)