Add block stale-ness check

[rustaceanrob]

I noticed my home node got stuck on a block with the connection it had, but continued to exchange ping and pongs. This adds a very basic check to see if blocks are being withheld from us by disconnecting from a peer if they have not sent a block to us for 20 minutes.

[sedited]

I've seen this too a couple of times. It doesn't really make sense to me. Seems fine to be rotating peers every now and again, but I wonder if at this point we should just introduce multi-peer messages handling.

[rustaceanrob]

I think in that case we would still want to disconnect all connections if we haven't seen a block for a certain amount of time? Maybe it could be less aggressive in the first stale check (like 30 minutes), then move to something more aggressive?

[sedited]

Ran this for a bit and saw a few re-connects. The warning seems a bit harsh, since this will print even during normal block rate variations. Maybe relegate the log level to info?

[rustaceanrob]

Changed to info in 9e10d18

[yancyribbens]

Just curious, how does bitcoin-core node handle this problem? It doesn't seem like this is a case of re-inventing the wheel?

[sedited]

Just curious, how does bitcoin-core node handle this problem? It doesn't seem like this is a case of re-inventing the wheel?

We still only maintain a single connection at a time here. Running Core with many connections alleviates this, since there is some cycling between peers and the likelihood of finding a good peer is high. However even in Core, this is a problem during ibd, where blocks are fetched from a single peer at a time. To avoid stalling in that case, there are a bunch of timeouts (grep stalling in net_processing.cpp).