Overzealous getUncles KvtNotFound(Kvt, ctx=get, error=GetNotFound) warning on GetBlockBodies request

[tersec]

This isn't anything going wrong, and should not be a warning.

from status-im/infra-nimbus#193 (comment)

WRN 2024-09-17 09:05:17.889+00:00 getUncles()                                topics="core_db" ommersHash=3EF3400295E8F4DFE6AED323009018E85DADBB1A4D960FC5CEE0709D484F6E97 error="KvtNotFound(Kvt, ctx=get, error=GetNotFound)"

with many more like this in the logs, even before any CLs had started to use the nimbus-eth1 engine API, i.e. it was only based on connected eth68 peers. Based on this, the most likely reason is that the not completely synced nimbus-eth1 node

nimbus-eth1/nimbus/sync/protocol/eth68.nim

Lines 210 to 231 in ea74e03

	requestResponse:
	# User message 0x05: GetBlockBodies.
	proc getBlockBodies(peer: Peer, hashes: openArray[Hash256]) =
	trace trEthRecvReceived & "GetBlockBodies (0x05)", peer,
	hashes=hashes.len
	if hashes.len > maxBodiesFetch:
	debug "GetBlockBodies (0x05) requested too many bodies",
	peer, requested=hashes.len, max=maxBodiesFetch
	await peer.disconnect(BreachOfProtocol)
	return
	let ctx = peer.networkState()
	let bodies = ctx.getBlockBodies(hashes)
	handleHandlerError(bodies)
	if bodies.get.len > 0:
	trace trEthSendReplying & "with BlockBodies (0x06)", peer,
	sent=bodies.get.len, requested=hashes.len
	else:
	trace trEthSendReplying & "EMPTY BlockBodies (0x06)", peer,
	sent=0, requested=hashes.len
	await response.send(bodies.get)

calling

nimbus-eth1/nimbus/sync/handlers/eth.nim

Lines 354 to 367 in ea74e03

	method getBlockBodies*(ctx: EthWireRef,
	hashes: openArray[Hash256]):
	Result[seq[BlockBody], string]
	{.gcsafe.} =
	let db = ctx.db
	var body: BlockBody
	var list: seq[BlockBody]
	for blockHash in hashes:
	if db.getBlockBody(blockHash, body):
	list.add body
	else:
	list.add BlockBody()
	trace "handlers.getBlockBodies: blockBody not found", blockHash
	return ok(list)

calls (eventually)

nimbus-eth1/nimbus/db/core_db/core_apps.nim

Lines 649 to 662 in ea74e03

	proc getBlockBody*(
	db: CoreDbRef;
	header: BlockHeader;
	output: var BlockBody;
	): bool =
	try:
	output.transactions = db.getTransactions(header.txRoot)
	output.uncles = db.getUncles(header.ommersHash)
	if header.withdrawalsRoot.isSome:
	output.withdrawals = Opt.some(db.getWithdrawals(header.withdrawalsRoot.get))
	true
	except RlpError:
	false

calls

nimbus-eth1/nimbus/db/core_db/core_apps.nim

Lines 596 to 609 in ea74e03

	proc getUncles*(
	db: CoreDbRef;
	ommersHash: Hash256;
	): seq[BlockHeader]
	{.gcsafe, raises: [RlpError].} =
	const info = "getUncles()"
	if ommersHash != EMPTY_UNCLE_HASH:
	let encodedUncles = block:
	let key = genericHashKey(ommersHash)
	db.ctx.getKvt().get(key.toOpenArray).valueOr:
	if error.error == KvtNotFound:
	warn info, ommersHash, error=($$error)
	return @[]
	return rlp.decode(encodedUncles, seq[BlockHeader])

when the header is present but not the body is not.

For example, this can occur after having imported era1 files -- persistBlocks etc do populate the headers in the database, but not (because the blocks already live in the era1 files) the blocks themselves. And they are pre-merge, so they have non-empty uncles/ommers.