nimbus-eth1/fluffy/scripts/test_portal_testnet.nim

353 lines
13 KiB
Nim

# Fluffy
# Copyright (c) 2021-2024 Status Research & Development GmbH
# Licensed and distributed under either of
# * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
# * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
# at your option. This file may not be copied, modified, or distributed except according to those terms.
import
os,
std/sequtils,
unittest2,
testutils,
confutils,
chronos,
stew/byteutils,
eth/p2p/discoveryv5/random2,
eth/common/keys,
../common/common_types,
../rpc/portal_rpc_client,
../rpc/eth_rpc_client,
../eth_data/[history_data_seeding, history_data_json_store, history_data_ssz_e2s],
../network/history/[history_content, validation/historical_hashes_accumulator],
../tests/history_network_tests/test_history_util
type
FutureCallback[A] = proc(): Future[A] {.gcsafe, raises: [].}
CheckCallback[A] = proc(a: A): bool {.gcsafe, raises: [].}
PortalTestnetConf* = object
nodeCount* {.defaultValue: 17, desc: "Number of nodes to test", name: "node-count".}:
int
rpcAddress* {.
desc: "Listening address of the JSON-RPC service for all nodes",
defaultValue: "127.0.0.1",
name: "rpc-address"
.}: string
baseRpcPort* {.
defaultValue: 10000,
desc: "Port of the JSON-RPC service of the bootstrap (first) node",
name: "base-rpc-port"
.}: uint16
proc connectToRpcServers(config: PortalTestnetConf): Future[seq[RpcClient]] {.async.} =
var clients: seq[RpcClient]
for i in 0 ..< config.nodeCount:
let client = newRpcHttpClient()
await client.connect(config.rpcAddress, Port(config.baseRpcPort + uint16(i)), false)
clients.add(client)
return clients
proc withRetries[A](
f: FutureCallback[A],
check: CheckCallback[A],
numRetries: int,
initialWait: Duration,
checkFailMessage: string,
nodeIdx: int,
): Future[A] {.async.} =
## Retries given future callback until either:
## it returns successfuly and given check is true
## or
## function reaches max specified retries
var tries = 0
var currentDuration = initialWait
while true:
try:
let res = await f()
if check(res):
return res
else:
raise newException(ValueError, checkFailMessage)
except CatchableError as exc:
if tries > numRetries:
# if we reached max number of retries fail
let msg =
"Call failed with msg: " & exc.msg & ", for node with idx: " & $nodeIdx &
", after " & $tries & " tries."
raise newException(ValueError, msg)
inc tries
# wait before new retry
await sleepAsync(currentDuration)
currentDuration = currentDuration * 2
# Sometimes we need to wait till data will be propagated over the network.
# To avoid long sleeps, this combinator can be used to retry some calls until
# success or until some condition hold (or both)
proc retryUntil[A](
f: FutureCallback[A], c: CheckCallback[A], checkFailMessage: string, nodeIdx: int
): Future[A] =
# some reasonable limits, which will cause waits as: 1, 2, 4, 8, 16, 32 seconds
return withRetries(f, c, 3, seconds(1), checkFailMessage, nodeIdx)
# Note:
# When doing json-rpc requests following `RpcPostError` can occur:
# "Failed to send POST Request with JSON-RPC." when a `HttpClientRequestRef`
# POST request is send in the json-rpc http client.
# This error is raised when the httpclient hits error:
# "Could not send request headers", which in its turn is caused by the
# "Incomplete data sent or received" in `AsyncStream`, which is caused by
# `ECONNRESET` or `EPIPE` error (see `isConnResetError()`) on the TCP stream.
# This can occur when the server side closes the connection, which happens after
# a `httpHeadersTimeout` of default 10 seconds (set on `HttpServerRef.new()`).
# In order to avoid here hitting this timeout a `close()` is done after each
# json-rpc call. Because the first json-rpc call opens up the connection, and it
# remains open until a close() (or timeout). No need to do another connect
# before any new call as the proc `connectToRpcServers` doesn't actually connect
# to servers, as client.connect doesn't do that. It just sets the `httpAddress`.
# Yes, this client json rpc API couldn't be more confusing.
# Could also just retry each call on failure, which would set up a new
# connection.
# We are kind of abusing the unittest2 here to run json rpc tests against other
# processes. Needs to be compiled with `-d:unittest2DisableParamFiltering` or
# the confutils cli will not work.
procSuite "Portal testnet tests":
let config = PortalTestnetConf.load()
let rng = newRng()
asyncTest "Discv5 - Random node lookup from each node":
let clients = await connectToRpcServers(config)
var nodeInfos: seq[NodeInfo]
for client in clients:
let nodeInfo = await client.discv5_nodeInfo()
await client.close()
nodeInfos.add(nodeInfo)
# Kick off the network by trying to add all records to each node.
# These nodes are also set as seen, so they get passed along on findNode
# requests.
# Note: The amount of Records added here can be less but then the
# probability that all nodes will still be reached needs to be calculated.
# Note 2: One could also ping all nodes but that is much slower and more
# error prone
for client in clients:
discard await client.discv5_addEnrs(
nodeInfos.map(
proc(x: NodeInfo): Record =
x.enr
)
)
await client.close()
for client in clients:
let routingTableInfo = await client.discv5_routingTableInfo()
await client.close()
var start: seq[NodeId]
let nodes = foldl(routingTableInfo.buckets, a & b, start)
# A node will have at least the first bucket filled. One could increase
# this based on the probability that x amount of nodes fit in the buckets.
check nodes.len >= (min(config.nodeCount - 1, 16))
# grab a random node its `NodeInfo` and lookup that node from all nodes.
let randomNodeInfo = sample(rng[], nodeInfos)
for client in clients:
var enr: Record
enr = await client.discv5_lookupEnr(randomNodeInfo.nodeId)
check enr == randomNodeInfo.enr
await client.close()
asyncTest "Portal State - Random node lookup from each node":
let clients = await connectToRpcServers(config)
var nodeInfos: seq[NodeInfo]
for client in clients:
let nodeInfo = await client.portal_stateNodeInfo()
await client.close()
nodeInfos.add(nodeInfo)
for client in clients:
discard await client.portal_stateAddEnrs(
nodeInfos.map(
proc(x: NodeInfo): Record =
x.enr
)
)
await client.close()
for client in clients:
let routingTableInfo = await client.portal_stateRoutingTableInfo()
await client.close()
var start: seq[NodeId]
let nodes = foldl(routingTableInfo.buckets, a & b, start)
check nodes.len >= (min(config.nodeCount - 1, 16))
# grab a random node its `NodeInfo` and lookup that node from all nodes.
let randomNodeInfo = sample(rng[], nodeInfos)
for client in clients:
var enr: Record
try:
enr = await client.portal_stateLookupEnr(randomNodeInfo.nodeId)
except CatchableError as e:
echo e.msg
# TODO: For state network this occasionally fails. It might be because the
# distance function is not used in all locations, or perhaps it just
# doesn't converge to the target always with this distance function. To be
# further investigated.
skip()
# check enr == randomNodeInfo.enr
await client.close()
asyncTest "Portal History - Random node lookup from each node":
let clients = await connectToRpcServers(config)
var nodeInfos: seq[NodeInfo]
for client in clients:
let nodeInfo = await client.portal_historyNodeInfo()
await client.close()
nodeInfos.add(nodeInfo)
for client in clients:
discard await client.portal_historyAddEnrs(
nodeInfos.map(
proc(x: NodeInfo): Record =
x.enr
)
)
await client.close()
for client in clients:
let routingTableInfo = await client.portal_historyRoutingTableInfo()
await client.close()
var start: seq[NodeId]
let nodes = foldl(routingTableInfo.buckets, a & b, start)
check nodes.len >= (min(config.nodeCount - 1, 16))
# grab a random node its `NodeInfo` and lookup that node from all nodes.
let randomNodeInfo = sample(rng[], nodeInfos)
for client in clients:
var enr: Record
enr = await client.portal_historyLookupEnr(randomNodeInfo.nodeId)
await client.close()
check enr == randomNodeInfo.enr
asyncTest "Portal History - Propagate blocks and do content lookups":
const
headerFile =
"./vendor/portal-spec-tests/tests/mainnet/history/headers/1000001-1000010.e2s"
accumulatorFile =
"./vendor/portal-spec-tests/tests/mainnet/history/accumulator/epoch-record-00122.ssz"
blockDataFile = "./fluffy/tests/blocks/mainnet_blocks_1000001_1000010.json"
let
blockHeaders = readBlockHeaders(headerFile).valueOr:
raiseAssert "Invalid header file: " & headerFile
epochRecord = readEpochRecordCached(accumulatorFile).valueOr:
raiseAssert "Invalid epoch accumulator file: " & accumulatorFile
blockHeadersWithProof = buildHeadersWithProof(blockHeaders, epochRecord).valueOr:
raiseAssert "Could not build headers with proof"
blockData = readJsonType(blockDataFile, BlockDataTable).valueOr:
raiseAssert "Invalid block data file" & blockDataFile
clients = await connectToRpcServers(config)
# Gossiping all block headers with proof first, as bodies and receipts
# require them for validation.
for (contentKey, contentValue) in blockHeadersWithProof:
discard (
await clients[0].portal_historyGossip(contentKey.toHex(), contentValue.toHex())
)
# TODO: Fix iteration order: Because the blockData gets parsed into a
# BlockDataTable, iterating over this result in gossiping the block bodies
# and receipts of block in a different order than the headers.
# Because of this, block bodies and receipts for block
# 0x6251d65b8a8668efabe2f89c96a5b6332d83b3bbe585089ea6b2ab9b6754f5e9
# come right after the headers with proof. This is likely to cause validation
# failures on the nodes, as the block bodies and receipts require the header
# to get validated.
await sleepAsync(seconds(1))
# Gossiping all block bodies and receipts.
for b in blocks(blockData, false):
for i, value in b:
if i == 0:
# Note: Skipping the headers, they are handled above already
continue
# Only sending non empty data, e.g. empty receipts are not send
# TODO: Could do a similar thing for a combination of empty
# txs and empty uncles, as then the serialization is always the same.
if value[1].len() > 0:
let
contentKey = history_content.encode(value[0]).asSeq().toHex()
contentValue = value[1].toHex()
discard (await clients[0].portal_historyGossip(contentKey, contentValue))
await clients[0].close()
for i, client in clients:
# Note: Once there is the Canonical Indices Network, we don't need to
# access this file anymore here for the block hashes.
for hash in blockData.blockHashes():
# Note: More flexible approach instead of generic retries could be to
# add a json-rpc debug proc that returns whether the offer queue is empty or
# not. And then poll every node until all nodes have an empty queue.
let content = await retryUntil(
proc(): Future[Opt[BlockObject]] {.async.} =
try:
let res = await client.eth_getBlockByHash(hash, true)
await client.close()
return res
except CatchableError as exc:
await client.close()
raise exc,
proc(mc: Opt[BlockObject]): bool =
return mc.isSome(),
"Did not receive expected Block with hash " & hash.data.toHex(),
i,
)
check content.isSome()
let blockObj = content.get()
check blockObj.hash == hash
for tx in blockObj.transactions:
doAssert(tx.kind == tohTx)
check tx.tx.blockHash.get == hash
let filterOptions = FilterOptions(blockHash: Opt.some(hash))
let logs = await retryUntil(
proc(): Future[seq[LogObject]] {.async.} =
try:
let res = await client.eth_getLogs(filterOptions)
await client.close()
return res
except CatchableError as exc:
await client.close()
raise exc,
proc(mc: seq[LogObject]): bool =
return true,
"",
i,
)
for l in logs:
check:
l.blockHash == Opt.some(hash)
# TODO: Check ommersHash, need the headers and not just the hashes
# for uncle in blockObj.uncles:
# discard
await client.close()