Unsyncable Blocks

[SurfingNerd]

During a Stress Test Session, the 25 Nodes validator set created blocks, that were not able to sync for Nodes outside the validator set.
A investigation trail leads to the suspicion that we did hit some boundaries in the block sync protocol. may it be byte based or time based how long a operation is allowed to take.

[SurfingNerd]

Further investigation proved that theory to be true.

Nodes request a list of blocks from other nodes, by requesting the blocks via a list of block hashes they want to get, without knowing their size.
If there are large blocks in a longer sequence, it becomes to much either for the receiver or the sender to process, resulting in timouts.
The solution was to increase this maximum limits of send/received bytes, increase the timeout and only request one block at a time.
Unfortunately this change to only request one block at a time might also affect the sync speed by a little.
Further P2P Protocol improvements could be that the responder of the block body request only responds with a subset of blocks, and the requester still sees that responder as a honest node, even he did not respond all blocks requested.

[SurfingNerd]

solved as a hotfix solution. with further investment we could make the improvements that does not decrease the number of blocks propagated in scenarios were the blocksize is small.