nimbus-eth2/beacon_chain/nimbus_binary_common.nim

457 lines
16 KiB
Nim

# beacon_chain
# Copyright (c) 2018-2024 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.
{.push raises: [].}
# Common routines for a BeaconNode and a ValidatorClient
import
# Standard library
std/[tables, strutils, terminal, typetraits],
# Nimble packages
chronos, confutils, presto, toml_serialization, metrics,
chronicles, chronicles/helpers as chroniclesHelpers, chronicles/topics_registry,
stew/io2,
# Local modules
./spec/[helpers, keystore],
./spec/datatypes/base,
"."/[beacon_clock, beacon_node_status, conf, conf_common, version]
when defined(posix):
import termios
declareGauge versionGauge, "Nimbus version info (as metric labels)", ["version", "commit"], name = "version"
versionGauge.set(1, labelValues=[fullVersionStr, gitRevision])
declareGauge nimVersionGauge, "Nim version info", ["version", "nim_commit"], name = "nim_version"
nimVersionGauge.set(1, labelValues=[NimVersion, getNimGitHash()])
export
confutils, toml_serialization, beacon_clock, beacon_node_status, conf,
conf_common
type
SlotStartProc*[T] = proc(node: T, wallTime: BeaconTime,
lastSlot: Slot): Future[bool] {.gcsafe,
raises: [].}
# silly chronicles, colors is a compile-time property
when defaultChroniclesStream.outputs.type.arity == 2:
func stripAnsi(v: string): string =
var
res = newStringOfCap(v.len)
i: int
while i < v.len:
let c = v[i]
if c == '\x1b':
var
x = i + 1
found = false
while x < v.len: # look for [..m
let c2 = v[x]
if x == i + 1:
if c2 != '[':
break
else:
if c2 in {'0'..'9'} + {';'}:
discard # keep looking
elif c2 == 'm':
i = x + 1
found = true
break
else:
break
inc x
if found: # skip adding c
continue
res.add c
inc i
res
proc updateLogLevel*(logLevel: string) {.raises: [ValueError].} =
# Updates log levels (without clearing old ones)
let directives = logLevel.split(";")
try:
setLogLevel(parseEnum[LogLevel](directives[0].capitalizeAscii()))
except ValueError:
raise (ref ValueError)(msg: "Please specify one of TRACE, DEBUG, INFO, NOTICE, WARN, ERROR or FATAL")
if directives.len > 1:
for topicName, settings in parseTopicDirectives(directives[1..^1]):
if not setTopicState(topicName, settings.state, settings.logLevel):
warn "Unrecognized logging topic", topic = topicName
proc detectTTY*(stdoutKind: StdoutLogKind): StdoutLogKind =
if stdoutKind == StdoutLogKind.Auto:
if isatty(stdout):
# On a TTY, let's be fancy
StdoutLogKind.Colors
else:
# When there's no TTY, we output no colors because this matches what
# released binaries were doing before auto-detection was around and
# looks decent in systemd-captured journals.
StdoutLogKind.NoColors
else:
stdoutKind
when defaultChroniclesStream.outputs.type.arity == 2:
from std/os import splitFile
from "."/filepath import secureCreatePath
proc setupFileLimits*() =
when not defined(windows):
# In addition to databases and sockets, we need a file descriptor for every
# validator - setting it to 16k should provide sufficient margin
let
limit = getMaxOpenFiles2().valueOr(16384)
if limit < 16384:
setMaxOpenFiles2(16384).isOkOr:
warn "Cannot increase open file limit", err = osErrorMsg(error)
proc setupLogging*(
logLevel: string, stdoutKind: StdoutLogKind, logFile: Option[OutFile]) =
# In the cfg file for nimbus, we create two formats: textlines and json.
# Here, we either write those logs to an output, or not, depending on the
# given configuration.
# Arguably, if we don't use a format, chronicles should not create it.
when defaultChroniclesStream.outputs.type.arity != 2:
warn "Logging configuration options not enabled in the current build"
else:
# Naive approach where chronicles will form a string and we will discard
# it, even if it could have skipped the formatting phase
proc noOutput(logLevel: LogLevel, msg: LogOutputStr) = discard
proc writeAndFlush(f: File, msg: LogOutputStr) =
try:
f.write(msg)
f.flushFile()
except IOError as err:
logLoggingFailure(cstring(msg), err)
proc stdoutFlush(logLevel: LogLevel, msg: LogOutputStr) =
writeAndFlush(stdout, msg)
proc noColorsFlush(logLevel: LogLevel, msg: LogOutputStr) =
writeAndFlush(stdout, stripAnsi(msg))
let fileWriter =
if logFile.isSome():
let
logFile = logFile.get.string
logFileDir = splitFile(logFile).dir
lres = secureCreatePath(logFileDir)
if lres.isOk():
try:
let
f = open(logFile, fmAppend)
x = proc(logLevel: LogLevel, msg: LogOutputStr) =
writeAndFlush(f, msg) # will close when program terminates
x
except CatchableError as exc:
error "Failed to create log file", logFile, msg = exc.msg
noOutput
else:
error "Failed to create directory for log file",
path = logFileDir, err = ioErrorMsg(lres.error)
noOutput
else:
noOutput
defaultChroniclesStream.outputs[1].writer = fileWriter
let tmp = detectTTY(stdoutKind)
case tmp
of StdoutLogKind.Auto: raiseAssert "checked above"
of StdoutLogKind.Colors:
defaultChroniclesStream.outputs[0].writer = stdoutFlush
of StdoutLogKind.NoColors:
defaultChroniclesStream.outputs[0].writer = noColorsFlush
of StdoutLogKind.Json:
defaultChroniclesStream.outputs[0].writer = noOutput
let prevWriter = defaultChroniclesStream.outputs[1].writer
defaultChroniclesStream.outputs[1].writer =
proc(logLevel: LogLevel, msg: LogOutputStr) =
stdoutFlush(logLevel, msg)
prevWriter(logLevel, msg)
of StdoutLogKind.None:
defaultChroniclesStream.outputs[0].writer = noOutput
if logFile.isSome():
warn "The --log-file option is deprecated. Consider redirecting the standard output to a file instead"
try:
updateLogLevel(logLevel)
except ValueError as err:
try:
stderr.write "Invalid value for --log-level. " & err.msg
except IOError:
echo "Invalid value for --log-level. " & err.msg
quit 1
template makeBannerAndConfig*(clientId: string, ConfType: type): untyped =
let
version = clientId & "\p" & copyrights & "\p\p" &
"eth2 specification v" & SPEC_VERSION & "\p\p" &
nimBanner
# TODO for some reason, copyrights are printed when doing `--help`
{.push warning[ProveInit]: off.}
let config = try:
ConfType.load(
version = version, # but a short version string makes more sense...
copyrightBanner = clientId,
secondarySources = proc (
config: ConfType, sources: ref SecondarySources
) {.raises: [ConfigurationError].} =
if config.configFile.isSome:
sources.addConfigFile(Toml, config.configFile.get)
)
except CatchableError as err:
# We need to log to stderr here, because logging hasn't been configured yet
try:
stderr.write "Failure while loading the configuration:\n"
stderr.write err.msg
stderr.write "\n"
if err[] of ConfigurationError and
err.parent != nil and
err.parent[] of TomlFieldReadingError:
let fieldName = ((ref TomlFieldReadingError)(err.parent)).field
if fieldName in ["web3-url", "bootstrap-node",
"direct-peer", "validator-monitor-pubkey"]:
stderr.write "Since the '" & fieldName & "' option is allowed to " &
"have more than one value, please make sure to supply " &
"a properly formatted TOML array\n"
except IOError:
discard
quit 1
{.pop.}
config
proc checkIfShouldStopAtEpoch*(scheduledSlot: Slot,
stopAtEpoch: uint64): bool =
# Offset backwards slightly to allow this epoch's finalization check to occur
if scheduledSlot > 3 and stopAtEpoch > 0'u64 and
(scheduledSlot - 3).epoch() >= stopAtEpoch:
info "Stopping at pre-chosen epoch",
chosenEpoch = stopAtEpoch,
epoch = scheduledSlot.epoch(),
slot = scheduledSlot
true
else:
false
proc resetStdin*() =
when defined(posix):
# restore echoing, in case it was disabled by a password prompt
let fd = stdin.getFileHandle()
var attrs: Termios
discard fd.tcGetAttr(attrs.addr)
attrs.c_lflag = attrs.c_lflag or Cflag(ECHO)
discard fd.tcSetAttr(TCSANOW, attrs.addr)
proc runKeystoreCachePruningLoop*(cache: KeystoreCacheRef) {.async.} =
while true:
let exitLoop =
try:
await sleepAsync(60.seconds)
false
except CatchableError:
cache.clear()
true
if exitLoop: break
cache.pruneExpiredKeys()
proc sleepAsync*(t: TimeDiff): Future[void] =
sleepAsync(nanoseconds(
if t.nanoseconds < 0: 0'i64 else: t.nanoseconds))
proc runSlotLoop*[T](node: T, startTime: BeaconTime,
slotProc: SlotStartProc[T]) {.async.} =
var
curSlot = startTime.slotOrZero()
nextSlot = curSlot + 1 # No earlier than GENESIS_SLOT + 1
timeToNextSlot = nextSlot.start_beacon_time() - startTime
info "Scheduling first slot action",
startTime = shortLog(startTime),
nextSlot = shortLog(nextSlot),
timeToNextSlot = shortLog(timeToNextSlot)
while true:
# Start by waiting for the time when the slot starts. Sleeping relinquishes
# control to other tasks which may or may not finish within the alotted
# time, so below, we need to be wary that the ship might have sailed
# already.
await sleepAsync(timeToNextSlot)
let
wallTime = node.beaconClock.now()
wallSlot = wallTime.slotOrZero() # Always > GENESIS!
if wallSlot < nextSlot:
# While we were sleeping, the system clock changed and time moved
# backwards!
if wallSlot + 1 < nextSlot:
# This is a critical condition where it's hard to reason about what
# to do next - we'll call the attention of the user here by shutting
# down.
fatal "System time adjusted backwards significantly - clock may be inaccurate - shutting down",
nextSlot = shortLog(nextSlot),
wallSlot = shortLog(wallSlot)
bnStatus = BeaconNodeStatus.Stopping
return
# Time moved back by a single slot - this could be a minor adjustment,
# for example when NTP does its thing after not working for a while
warn "System time adjusted backwards, rescheduling slot actions",
wallTime = shortLog(wallTime),
nextSlot = shortLog(nextSlot),
wallSlot = shortLog(wallSlot)
# cur & next slot remain the same
timeToNextSlot = nextSlot.start_beacon_time() - wallTime
continue
if wallSlot > nextSlot + SLOTS_PER_EPOCH:
# Time moved forwards by more than an epoch - either the clock was reset
# or we've been stuck in processing for a long time - either way, we will
# skip ahead so that we only process the events of the last
# SLOTS_PER_EPOCH slots
warn "Time moved forwards by more than an epoch, skipping ahead",
curSlot = shortLog(curSlot),
nextSlot = shortLog(nextSlot),
wallSlot = shortLog(wallSlot)
curSlot = wallSlot - SLOTS_PER_EPOCH
elif wallSlot > nextSlot:
notice "Missed expected slot start, catching up",
delay = shortLog(wallTime - nextSlot.start_beacon_time()),
curSlot = shortLog(curSlot),
nextSlot = shortLog(curSlot)
let breakLoop = await slotProc(node, wallTime, curSlot)
if breakLoop:
break
curSlot = wallSlot
nextSlot = wallSlot + 1
timeToNextSlot = nextSlot.start_beacon_time() - node.beaconClock.now()
proc init*(T: type RestServerRef,
ip: IpAddress,
port: Port,
allowedOrigin: Option[string],
validateFn: PatternCallback,
ident: string,
config: AnyConf): T =
let
address = initTAddress(ip, port)
serverFlags = {HttpServerFlags.QueryCommaSeparatedArray,
HttpServerFlags.NotifyDisconnect}
# We increase default timeout to help validator clients who poll our server
# at least once per slot (12.seconds).
let
headersTimeout =
if config.restRequestTimeout == 0:
chronos.InfiniteDuration
else:
seconds(int64(config.restRequestTimeout))
maxHeadersSize = config.restMaxRequestHeadersSize * 1024
maxRequestBodySize = config.restMaxRequestBodySize * 1024
let res = RestServerRef.new(RestRouter.init(validateFn, allowedOrigin),
address, serverFlags = serverFlags,
serverIdent = ident,
httpHeadersTimeout = headersTimeout,
maxHeadersSize = maxHeadersSize,
maxRequestBodySize = maxRequestBodySize,
errorType = string)
if res.isErr():
notice "REST HTTP server could not be started", address = $address,
reason = res.error()
nil
else:
let server = res.get()
notice "Starting REST HTTP server", url = "http://" & $server.localAddress()
server
type
KeymanagerInitResult* = object
server*: RestServerRef
token*: string
proc initKeymanagerServer*(
config: AnyConf,
existingRestServer: RestServerRef = nil): KeymanagerInitResult
{.raises: [].} =
var token: string
let keymanagerServer = if config.keymanagerEnabled:
if config.keymanagerTokenFile.isNone:
echo "To enable the Keymanager API, you must also specify " &
"the --keymanager-token-file option."
quit 1
let
tokenFilePath = config.keymanagerTokenFile.get.string
tokenFileReadRes = readAllChars(tokenFilePath)
if tokenFileReadRes.isErr:
fatal "Failed to read the keymanager token file",
error = $tokenFileReadRes.error
quit 1
token = tokenFileReadRes.value.strip
if token.len == 0:
fatal "The keymanager token should not be empty", tokenFilePath
quit 1
when config is BeaconNodeConf:
if existingRestServer != nil and
config.restAddress == config.keymanagerAddress and
config.restPort == config.keymanagerPort:
existingRestServer
else:
RestServerRef.init(config.keymanagerAddress, config.keymanagerPort,
config.keymanagerAllowedOrigin,
validateKeymanagerApiQueries,
nimbusAgentStr,
config)
else:
RestServerRef.init(config.keymanagerAddress, config.keymanagerPort,
config.keymanagerAllowedOrigin,
validateKeymanagerApiQueries,
nimbusAgentStr,
config)
else:
nil
KeymanagerInitResult(server: keymanagerServer, token: token)
proc quitDoppelganger*() =
# Avoid colliding with
# https://www.freedesktop.org/software/systemd/man/systemd.exec.html#Process%20Exit%20Codes
# This error code is used to permanently shut down validators
fatal "Doppelganger detection triggered! It appears a validator loaded into " &
"this process is already live on the network - the validator is at high " &
"risk of being slashed due to the same keys being used in two setups. " &
"See https://nimbus.guide/doppelganger-detection.html for more information!"
const QuitDoppelganger = 129
quit QuitDoppelganger