change(network): Broadcast block submissions to all peers in the peer set

[arya2]

Motivation

It would be nice to broadcast accepted block submissions to all ready peers.

Solution

• Adds a AdvertiseBlockToAll variant to the zebra-network Request type
• Handles the new request variant in the peer set by sending it to all ready peers
  • Factors logic for sending a request to multiple peers out of route_multiple()
  • Adds a broadcast_all() method which sends a request to every ready peer
• Updates the gossip_best_tip_block_hashes() task to call the peer set with a AdvertiseBlockToAll request instead of AdvertiseBlock if the block hash/height being advertised were received from the mined block channel.

Related changes:

• Corrects a doc comment on AdvertiseBlock (it routes to a third of ready peers, not half)
• Updates error message from the submitblock RPC when the mined block channel is closed

PR Checklist

• The PR name is suitable for the release notes.
• The PR follows the contribution guidelines.
• The library crate changelogs are up to date.
• The solution is tested.
• The documentation is up to date.

[oxarbitrage]

Nice, this is part of #9721

[oxarbitrage]

We might want to test this by removing:

https://github.com/ZcashFoundation/zebra/blob/main/zebra-rpc/qa/rpc-tests/addnode.py#L73-L79

[arya2]

We might want to test this by removing:

https://github.com/ZcashFoundation/zebra/blob/main/zebra-rpc/qa/rpc-tests/addnode.py#L73-L79

I replaced that with a single call to generate(10), it should only broadcast most of those to 1 peer as the rest will be unready and the peerset only sends the last AdvertiseBlockToAll request to unready peers once they become ready, but the 1 peer that receives each advertised block will then advertise to all of its peers.

We could make queued_broadcast_all_request a vector and always send every block submission to every peer, but that doesn't seem useful on Mainnet.

[arya2]

@gustavovalverde any idea why the deploy nodes job is being skipped?

[gustavovalverde]

@gustavovalverde any idea why the deploy nodes job is being skipped?

@arya2 We recently decided to move it to the main branch only. I also recently merged a fix to avoid the success job from failing. If you'd like to run the stateful tests in this PR you can add the run-stateful-tests label to it.

[oxarbitrage]

Looking good. This will also close #9721 if merged as the other task was done in #9917

[oxarbitrage]

thanks!