cic-internal-integration/apps/cic-eth/cic_eth/runnable/daemons/retry.py

280 lines
8.9 KiB
Python

# standard imports
import os
import sys
import logging
import argparse
import re
import datetime
# external imports
import confini
import celery
from cic_eth_registry import CICRegistry
from chainlib.chain import ChainSpec
from chainlib.eth.tx import unpack
from chainlib.connection import RPCConnection
from chainlib.eth.block import (
block_latest,
block_by_number,
Block,
)
from chainsyncer.driver import HeadSyncer
from chainsyncer.backend import MemBackend
from chainsyncer.error import NoBlockForYou
# local imports
from cic_eth.db import dsn_from_config
from cic_eth.db import SessionBase
from cic_eth.queue.tx import (
get_status_tx,
get_tx,
# get_upcoming_tx,
)
from cic_eth.admin.ctrl import lock_send
from cic_eth.db.enum import (
StatusEnum,
StatusBits,
LockEnum,
)
logging.basicConfig(level=logging.WARNING)
logg = logging.getLogger()
config_dir = os.path.join('/usr/local/etc/cic-eth')
argparser = argparse.ArgumentParser(description='daemon that monitors transactions in new blocks')
argparser.add_argument('-p', '--provider', dest='p', type=str, help='rpc provider')
argparser.add_argument('-c', type=str, default=config_dir, help='config root to use')
argparser.add_argument('-i', '--chain-spec', dest='i', type=str, help='chain spec')
argparser.add_argument('--batch-size', dest='batch_size', type=int, default=50, help='max amount of txs to resend per iteration')
argparser.add_argument('--retry-delay', dest='retry_delay', type=int, help='seconds to wait for retrying a transaction that is marked as sent')
argparser.add_argument('--env-prefix', default=os.environ.get('CONFINI_ENV_PREFIX'), dest='env_prefix', type=str, help='environment prefix for variables to overwrite configuration')
argparser.add_argument('-q', type=str, default='cic-eth', help='celery queue to submit transaction tasks to')
argparser.add_argument('-v', help='be verbose', action='store_true')
argparser.add_argument('-vv', help='be more verbose', action='store_true')
args = argparser.parse_args(sys.argv[1:])
if args.v == True:
logging.getLogger().setLevel(logging.INFO)
elif args.vv == True:
logging.getLogger().setLevel(logging.DEBUG)
config_dir = os.path.join(args.c)
os.makedirs(config_dir, 0o777, True)
config = confini.Config(config_dir, args.env_prefix)
config.process()
# override args
args_override = {
'ETH_PROVIDER': getattr(args, 'p'),
'CIC_CHAIN_SPEC': getattr(args, 'i'),
'CIC_TX_RETRY_DELAY': getattr(args, 'retry_delay'),
}
config.dict_override(args_override, 'cli flag')
config.censor('PASSWORD', 'DATABASE')
config.censor('PASSWORD', 'SSL')
logg.debug('config loaded from {}:\n{}'.format(config_dir, config))
config.add(args.batch_size, '_BATCH_SIZE', True)
app = celery.Celery(backend=config.get('CELERY_RESULT_URL'), broker=config.get('CELERY_BROKER_URL'))
queue = args.q
chain_spec = ChainSpec.from_chain_str(config.get('CIC_CHAIN_SPEC'))
RPCConnection.register_location(config.get('ETH_PROVIDER'), chain_spec, tag='default')
dsn = dsn_from_config(config)
SessionBase.connect(dsn, debug=config.true('DATABASE_DEBUG'))
straggler_delay = int(config.get('CIC_TX_RETRY_DELAY'))
# TODO: we already have the signed raw tx in get, so its a waste of cycles to get_tx here
def sendfail_filter(w3, tx_hash, rcpt, chain_spec):
tx_dict = get_tx(tx_hash)
tx = unpack(tx_dict['signed_tx'], chain_spec)
logg.debug('submitting tx {} for retry'.format(tx_hash))
s_check = celery.signature(
'cic_eth.admin.ctrl.check_lock',
[
tx_hash,
chain_str,
LockEnum.QUEUE,
tx['from'],
],
queue=queue,
)
# s_resume = celery.signature(
# 'cic_eth.eth.tx.resume_tx',
# [
# chain_str,
# ],
# queue=queue,
# )
# s_retry_status = celery.signature(
# 'cic_eth.queue.state.set_ready',
# [],
# queue=queue,
# )
s_resend = celery.signature(
'cic_eth.eth.gas.resend_with_higher_gas',
[
chain_str,
],
queue=queue,
)
#s_resume.link(s_retry_status)
#s_check.link(s_resume)
s_check.link(s_resend)
s_check.apply_async()
# TODO: can we merely use the dispatcher instead?
def dispatch(conn, chain_spec):
txs = get_status_tx(StatusEnum.RETRY, before=datetime.datetime.utcnow())
if len(txs) == 0:
logg.debug('no retry state txs found')
return
#signed_txs = list(txs.values())
#logg.debug('signed txs {} chain {}'.format(signed_txs, chain_str))
#for tx in signed_txs:
for k in txs.keys():
#tx_cache = get_tx_cache(k)
tx_raw = txs[k]
tx = unpack(tx_raw, chain_spec)
s_check = celery.signature(
'cic_eth.admin.ctrl.check_lock',
[
[tx_raw],
chain_str,
LockEnum.QUEUE,
tx['from'],
],
queue=queue,
)
s_send = celery.signature(
'cic_eth.eth.tx.send',
[
chain_str,
],
queue=queue,
)
s_check.link(s_send)
t = s_check.apply_async()
# try:
# r = t.get()
# logg.debug('submitted as {} result {} with queue task {}'.format(t, r, t.children[0].get()))
# except PermanentTxError as e:
# logg.error('tx {} permanently failed: {}'.format(tx, e))
# except TemporaryTxError as e:
# logg.error('tx {} temporarily failed: {}'.format(tx, e))
#
#
#def straggler_filter(w3, tx, rcpt, chain_str):
# before = datetime.datetime.utcnow() - datetime.timedelta(seconds=straggler_delay)
# txs = get_status_tx(StatusEnum.SENT, before)
# if len(txs) == 0:
# logg.debug('no straggler txs found')
# return
# txs = list(txs.keys())
# logg.debug('straggler txs {} chain {}'.format(signed_txs, chain_str))
# s_send = celery.signature(
# 'cic_eth.eth.gas.resend_with_higher_gas',
# [
# txs,
# chain_str,
# ],
# queue=queue,
# )
# s_send.apply_async()
class StragglerFilter:
def __init__(self, chain_spec, queue='cic-eth'):
self.chain_spec = chain_spec
self.queue = queue
def filter(self, conn, block, tx, db_session=None):
logg.debug('tx {}'.format(tx))
s_send = celery.signature(
'cic_eth.eth.gas.resend_with_higher_gas',
[
tx,
self.chain_spec.asdict(),
],
queue=self.queue,
)
return s_send.apply_async()
#return s_send
def __str__(self):
return 'stragglerfilter'
class RetrySyncer(HeadSyncer):
def __init__(self, conn, chain_spec, stalled_grace_seconds, batch_size=50, failed_grace_seconds=None):
backend = MemBackend(chain_spec, None)
super(RetrySyncer, self).__init__(backend)
self.chain_spec = chain_spec
if failed_grace_seconds == None:
failed_grace_seconds = stalled_grace_seconds
self.stalled_grace_seconds = stalled_grace_seconds
self.failed_grace_seconds = failed_grace_seconds
self.batch_size = batch_size
self.conn = conn
def get(self, conn):
o = block_latest()
r = conn.do(o)
(pair, flags) = self.backend.get()
n = int(r, 16)
if n == pair[0]:
raise NoBlockForYou('block {} already checked'.format(n))
o = block_by_number(n)
r = conn.do(o)
b = Block(r)
return b
def process(self, conn, block):
before = datetime.datetime.utcnow() - datetime.timedelta(seconds=self.stalled_grace_seconds)
stalled_txs = get_status_tx(
StatusBits.IN_NETWORK.value,
not_status=StatusBits.FINAL | StatusBits.MANUAL | StatusBits.OBSOLETE,
before=before,
limit=self.batch_size,
)
# stalled_txs = get_upcoming_tx(
# status=StatusBits.IN_NETWORK.value,
# not_status=StatusBits.FINAL | StatusBits.MANUAL | StatusBits.OBSOLETE,
# before=before,
# limit=self.batch_size,
# )
for tx in stalled_txs:
self.filter.apply(self.conn, block, tx)
self.backend.set(block.number, 0)
def main():
#o = block_latest()
conn = RPCConnection.connect(chain_spec, 'default')
#block = conn.do(o)
syncer = RetrySyncer(conn, chain_spec, straggler_delay, batch_size=config.get('_BATCH_SIZE'))
syncer.backend.set(0, 0)
syncer.add_filter(StragglerFilter(chain_spec, queue=queue))
syncer.loop(float(straggler_delay), conn)
if __name__ == '__main__':
main()