GitLab

* fix: fixLowPeers now runs periodically
Co-authored-by: 刘林欣 <llx@ipfsmain.cn>

removes duplicate comment

Revert "Peer Diversity in the Routing Table (#658)"

This reverts commit df4f28c3.

Ongoing work is being done in feat/dht-hardening-0.7 and this change isn't quite ready.

Fixed problem with refresh logging

feat: protect all peers in low buckets, tag everyone else with 5

Anything else is easy to game. This way, we remain connected to the core peers
and slightly prefer all others.

Pruning doesn't really matter as we'll just reconnect. The only concern is that
this will lock in 40 connections. However, that's nothing for libp2p anyways.
and the DHT will create that many connections _just_ for a single query.

* Peer Diversity in the DHT

fix: lookup context cancellation race condition

fix: protect useful peers in low buckets

* method for getting bootstrap peers and double the usefulness interval

And tag useful peers in higher buckets

That way, we never disconnect from the core of the network.

We don't protect our closest peers as:

* We'll reconnect to them frequently when we query for ourself.
* We use the core nodes most frequently.

enhancement/remove-unused-variable

Put back TestSelfWalkOnAddressChange

* rt refresh refactor

* Bootstrap when RT is empty and optimize allocations.
Co-authored-by: Steven Allen <steven@stebalien.com>

fix all flaky tests

Just wait for peers to fill up the routing table.

Also, fail if the routing tables aren't well-formed.

It turns out there were a couple of issues with this test:

1. It used too many file descriptors.
2. It tried to query the DHT without bootstrapping any nodes. That's never going
to work.