Hang when given parallel request to engine API

[jangko]

One of hive test involving engine_getPayloadBodiesByRangeV1 will make the test timeout.
The hive is sending parallel request to client.
It is not clear who is responsible for this, either in json-rpc server or in nimbus itself.

[advaita-saha]

DBG 2025-06-13 09:24:09.100+00:00 Processing JSON-RPC request                id=317 name=engine_getPayloadBodiesByHashV1
DBG 2025-06-13 09:24:38.565+00:00 Returning JSON-RPC response                id=317 name=engine_getPayloadBodiesByHashV1 len=291296035
DBG 2025-06-13 09:24:39.813+00:00 Peer disconnected                          topics="eth p2p rlpx" remote=Node[164.92.67.202:43676] clientId= reason=TcpError
DBG 2025-06-13 09:24:39.813+00:00 Internal error while processing JSON-RPC call topics="JSONRPC-HTTP-SERVER" msg="Unable to send response data, reason: Stream finished or remote side dropped connection"
DBG 2025-06-13 09:24:39.817+00:00 Accped handshake disconnection             topics="eth p2p rlpx" err="Data incomplete!" reason=TcpError remote=Node[174.138.13.118:42896]
DBG 2025-06-13 09:24:39.817+00:00 Accped handshake disconnection             topics="eth p2p rlpx" err="Data incomplete!" reason=TcpError remote=Node[35.233.210.124:41188]

Makes us not compatible with prysm

[cheatfate]

Have you noted size? 291,296,035 is a lot of bytes, this could cause problems if you do not send response chunk by chunk.

[jangko]

We set limit to engine_getPayloadBodiesByHashV1 max 32 bodies. We also have chunked transfer. But the size is indeed big. it means each block body is around 9MB! is that even possible?

[advaita-saha]

block body is around 9MB! is that even possible?

Capped at 10MB, but with the current gas limit 9MB is not possible
( Would it be a possibility that the seralization is messing things up? @jangko )

[jangko]

If I can have a block number and network name, I will happily run a test to get the bottom of this.