Improve gossip block receive time

[twoeths]

Describe the bug

On one node, the block receive time is 3.62s

2023-11-11 04:24:38.676 | Nov-11 01:24:38.676[network]       verbose: Received gossip block slot=7736821, root=0x06c9…4c44, curentSlot=7736821, peerId=16Uiu2HAmS7eWw4dWQ78FQdYqfoZg4K3YbvZ6rZkF6adoAYrk8d5M, delaySec=3.628000020980835, recvToVal=0.04800009727478027

it's alsmost 0.45s delay compared to other node

2023-11-11 04:24:38.222 | Nov-11 01:24:38.221[network]       verbose: Received gossip block slot=7736821, root=0x06c9…4c44, curentSlot=7736821, peerId=16Uiu2HAmTwzuybJco3ABCj7UpSrotK3HkCgf8vSjR9fayZJkLtRd, delaySec=3.174999952316284, recvToVal=0.046000003814697266

Note that we have a missed attestations with this block slot, a 200ms improvement could help validators vote for the correct head.

Expected behavior

Note that we have a missed attestations with this block slot, a 200ms improvement could help validators vote for the correct head.

• Right now UnknownBlock sync only based on forkchoice check to search for a block while gossip block already comes, when gossip block comes we should let forkchoice know there's incoming block so that we don't need to search for that same block again https://github.com/ChainSafe/lodestar/blob/v1.12.0/packages/beacon-node/src/network/processor/index.ts#L270
  This is because downloaded unknown block comes with the cost of decrypting at different layers at least, that blocks the main thread I/O when validating gossip block at execution engine
• Also consider Consider using snappy along with snappyjs #4170

Steps to reproduce

No response

Additional context

No response

Operating system

Linux

Lodestar version or commit hash

v1.12.0

[wemeetagain]

@twoeths is this stale?

[twoeths]

yes closing it