Source code for luigi.rpc

# -*- coding: utf-8 -*-
# Copyright 2012-2015 Spotify AB
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.
Implementation of the REST interface between the workers and the server. implements the client side of it, implements the server side.
See :doc:`/central_scheduler` for more info.
import os
import json
import logging
import socket
import time

from luigi.six.moves.urllib.parse import urljoin, urlencode, urlparse
from luigi.six.moves.urllib.request import urlopen
from luigi.six.moves.urllib.error import URLError

from luigi import configuration
from luigi.scheduler import RPC_METHODS


    import requests_unixsocket as requests
except ImportError:
        import requests
    except ImportError:
        HAS_REQUESTS = False

logger = logging.getLogger('luigi-interface')  # TODO: 'interface'?

def _urljoin(base, url):
    Join relative URLs to base URLs like urllib.parse.urljoin but support
    arbitrary URIs (esp. 'http+unix://').
    parsed = urlparse(base)
    scheme = parsed.scheme
    return urlparse(
        urljoin(parsed._replace(scheme='http').geturl(), url)

[docs]class RPCError(Exception): def __init__(self, message, sub_exception=None): super(RPCError, self).__init__(message) self.sub_exception = sub_exception
[docs]class URLLibFetcher(object): raises = (URLError, socket.timeout)
[docs] def fetch(self, full_url, body, timeout): body = urlencode(body).encode('utf-8') return urlopen(full_url, body, timeout).read().decode('utf-8')
[docs]class RequestsFetcher(object): def __init__(self, session): from requests import exceptions as requests_exceptions self.raises = requests_exceptions.RequestException self.session = session self.process_id = os.getpid()
[docs] def check_pid(self): # if the process id change changed from when the session was created # a new session needs to be setup since requests isn't multiprocessing safe. if os.getpid() != self.process_id: self.session = requests.Session() self.process_id = os.getpid()
[docs] def fetch(self, full_url, body, timeout): self.check_pid() resp = self.session.get(full_url, data=body, timeout=timeout) resp.raise_for_status() return resp.text
[docs]class RemoteScheduler(object): """ Scheduler proxy object. Talks to a RemoteSchedulerResponder. """ def __init__(self, url='http://localhost:8082/', connect_timeout=None): assert not url.startswith('http+unix://') or HAS_UNIX_SOCKET, ( 'You need to install requests-unixsocket for Unix socket support.' ) self._url = url.rstrip('/') config = configuration.get_config() if connect_timeout is None: connect_timeout = config.getfloat('core', 'rpc-connect-timeout', 10.0) self._connect_timeout = connect_timeout self._rpc_retry_attempts = config.getint('core', 'rpc-retry-attempts', 3) self._rpc_retry_wait = config.getint('core', 'rpc-retry-wait', 30) self._rpc_log_retries = config.getboolean('core', 'rpc-log-retries', True) if HAS_REQUESTS: self._fetcher = RequestsFetcher(requests.Session()) else: self._fetcher = URLLibFetcher() def _wait(self): if self._rpc_log_retries:"Wait for %d seconds" % self._rpc_retry_wait) time.sleep(self._rpc_retry_wait) def _fetch(self, url_suffix, body): full_url = _urljoin(self._url, url_suffix) last_exception = None attempt = 0 while attempt < self._rpc_retry_attempts: attempt += 1 if last_exception: if self._rpc_log_retries:"Retrying attempt %r of %r (max)" % (attempt, self._rpc_retry_attempts)) self._wait() # wait for a bit and retry try: response = self._fetcher.fetch(full_url, body, self._connect_timeout) break except self._fetcher.raises as e: last_exception = e if self._rpc_log_retries: logger.warning("Failed connecting to remote scheduler %r", self._url, exc_info=True) continue else: raise RPCError( "Errors (%d attempts) when connecting to remote scheduler %r" % (self._rpc_retry_attempts, self._url), last_exception ) return response def _request(self, url, data, attempts=3, allow_null=True): body = {'data': json.dumps(data)} for _ in range(attempts): page = self._fetch(url, body) response = json.loads(page)["response"] if allow_null or response is not None: return response raise RPCError("Received null response from remote scheduler %r" % self._url)
for method_name, method in RPC_METHODS.items(): setattr(RemoteScheduler, method_name, method)