# Copyright (c) 2010-2011 OpenStack Foundation # # 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 # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or # implied. # See the License for the specific language governing permissions and # limitations under the License. """ Logging middleware for the Swift proxy. This serves as both the default logging implementation and an example of how to plug in your own logging format/method. The logging format implemented below is as follows: client_ip remote_addr datetime request_method request_path protocol status_int referer user_agent auth_token bytes_recvd bytes_sent client_etag transaction_id headers request_time source log_info request_start_time request_end_time These values are space-separated, and each is url-encoded, so that they can be separated with a simple .split() * remote_addr is the contents of the REMOTE_ADDR environment variable, while client_ip is swift's best guess at the end-user IP, extracted variously from the X-Forwarded-For header, X-Cluster-Ip header, or the REMOTE_ADDR environment variable. * source (swift.source in the WSGI environment) indicates the code that generated the request, such as most middleware. (See below for more detail.) * log_info (swift.log_info in the WSGI environment) is for additional information that could prove quite useful, such as any x-delete-at value or other "behind the scenes" activity that might not otherwise be detectable from the plain log information. Code that wishes to add additional log information should use code like ``env.setdefault('swift.log_info', []).append(your_info)`` so as to not disturb others' log information. * Values that are missing (e.g. due to a header not being present) or zero are generally represented by a single hyphen ('-'). The proxy-logging can be used twice in the proxy server's pipeline when there is middleware installed that can return custom responses that don't follow the standard pipeline to the proxy server. For example, with staticweb, the middleware might intercept a request to /v1/AUTH_acc/cont/, make a subrequest to the proxy to retrieve /v1/AUTH_acc/cont/index.html and, in effect, respond to the client's original request using the 2nd request's body. In this instance the subrequest will be logged by the rightmost middleware (with a swift.source set) and the outgoing request (with body overridden) will be logged by leftmost middleware. Requests that follow the normal pipeline (use the same wsgi environment throughout) will not be double logged because an environment variable (swift.proxy_access_log_made) is checked/set when a log is made. All middleware making subrequests should take care to set swift.source when needed. With the doubled proxy logs, any consumer/processor of swift's proxy logs should look at the swift.source field, the rightmost log value, to decide if this is a middleware subrequest or not. A log processor calculating bandwidth usage will want to only sum up logs with no swift.source. """ import sys import time import six from six.moves.urllib.parse import quote, unquote from swift.common.swob import Request from swift.common.utils import (get_logger, get_remote_client, get_valid_utf8_str, config_true_value, InputProxy, list_from_csv, get_policy_index) from swift.common.storage_policy import POLICIES QUOTE_SAFE = '/:' class ProxyLoggingMiddleware(object): """ Middleware that logs Swift proxy requests in the swift log format. """ def __init__(self, app, conf, logger=None): self.app = app self.log_hdrs = config_true_value(conf.get( 'access_log_headers', conf.get('log_headers', 'no'))) log_hdrs_only = list_from_csv(conf.get( 'access_log_headers_only', '')) self.log_hdrs_only = [x.title() for x in log_hdrs_only] # The leading access_* check is in case someone assumes that # log_statsd_valid_http_methods behaves like the other log_statsd_* # settings. self.valid_methods = conf.get( 'access_log_statsd_valid_http_methods', conf.get('log_statsd_valid_http_methods', 'GET,HEAD,POST,PUT,DELETE,COPY,OPTIONS')) self.valid_methods = [m.strip().upper() for m in self.valid_methods.split(',') if m.strip()] access_log_conf = {} for key in ('log_facility', 'log_name', 'log_level', 'log_udp_host', 'log_udp_port', 'log_statsd_host', 'log_statsd_port', 'log_statsd_default_sample_rate', 'log_statsd_sample_rate_factor', 'log_statsd_metric_prefix'): value = conf.get('access_' + key, conf.get(key, None)) if value: access_log_conf[key] = value self.access_logger = logger or get_logger(access_log_conf, log_route='proxy-access') self.access_logger.set_statsd_prefix('proxy-server') self.reveal_sensitive_prefix = int( conf.get('reveal_sensitive_prefix', 16)) def method_from_req(self, req): return req.environ.get('swift.orig_req_method', req.method) def req_already_logged(self, env): return env.get('swift.proxy_access_log_made') def mark_req_logged(self, env): env['swift.proxy_access_log_made'] = True def obscure_sensitive(self, value): if value and len(value) > self.reveal_sensitive_prefix: return value[:self.reveal_sensitive_prefix] + '...' return value def log_request(self, req, status_int, bytes_received, bytes_sent, start_time, end_time, resp_headers=None): """ Log a request. :param req: swob.Request object for the request :param status_int: integer code for the response status :param bytes_received: bytes successfully read from the request body :param bytes_sent: bytes yielded to the WSGI server :param start_time: timestamp request started :param end_time: timestamp request completed :param resp_headers: dict of the response headers """ resp_headers = resp_headers or {} req_path = get_valid_utf8_str(req.path) the_request = quote(unquote(req_path), QUOTE_SAFE) if req.query_string: the_request = the_request + '?' + req.query_string logged_headers = None if self.log_hdrs: if self.log_hdrs_only: logged_headers = '\n'.join('%s: %s' % (k, v) for k, v in req.headers.items() if k in self.log_hdrs_only) else: logged_headers = '\n'.join('%s: %s' % (k, v) for k, v in req.headers.items()) method = self.method_from_req(req) end_gmtime_str = time.strftime('%d/%b/%Y/%H/%M/%S', time.gmtime(end_time)) duration_time_str = "%.4f" % (end_time - start_time) start_time_str = "%.9f" % start_time end_time_str = "%.9f" % end_time policy_index = get_policy_index(req.headers, resp_headers) self.access_logger.info(' '.join( quote(str(x) if x else '-', QUOTE_SAFE) for x in ( get_remote_client(req), req.remote_addr, end_gmtime_str, method, the_request, req.environ.get('SERVER_PROTOCOL'), status_int, req.referer, req.user_agent, self.obscure_sensitive(req.headers.get('x-auth-token')), bytes_received, bytes_sent, req.headers.get('etag', None), req.environ.get('swift.trans_id'), logged_headers, duration_time_str, req.environ.get('swift.source'), ','.join(req.environ.get('swift.log_info') or ''), start_time_str, end_time_str, policy_index ))) # Log timing and bytes-transferred data to StatsD metric_name = self.statsd_metric_name(req, status_int, method) metric_name_policy = self.statsd_metric_name_policy(req, status_int, method, policy_index) # Only log data for valid controllers (or SOS) to keep the metric count # down (egregious errors will get logged by the proxy server itself). if metric_name: self.access_logger.timing(metric_name + '.timing', (end_time - start_time) * 1000) self.access_logger.update_stats(metric_name + '.xfer', bytes_received + bytes_sent) if metric_name_policy: self.access_logger.timing(metric_name_policy + '.timing', (end_time - start_time) * 1000) self.access_logger.update_stats(metric_name_policy + '.xfer', bytes_received + bytes_sent) def get_metric_name_type(self, req): if req.path.startswith('/v1/'): try: stat_type = [None, 'account', 'container', 'object'][req.path.strip('/').count('/')] except IndexError: stat_type = 'object' else: stat_type = req.environ.get('swift.source') return stat_type def statsd_metric_name(self, req, status_int, method): stat_type = self.get_metric_name_type(req) if stat_type is None: return None stat_method = method if method in self.valid_methods \ else 'BAD_METHOD' return '.'.join((stat_type, stat_method, str(status_int))) def statsd_metric_name_policy(self, req, status_int, method, policy_index): if policy_index is None: return None stat_type = self.get_metric_name_type(req) if stat_type == 'object': stat_method = method if method in self.valid_methods \ else 'BAD_METHOD' # The policy may not exist policy = POLICIES.get_by_index(policy_index) if policy: return '.'.join((stat_type, 'policy', str(policy_index), stat_method, str(status_int))) else: return None else: return None def __call__(self, env, start_response): if self.req_already_logged(env): return self.app(env, start_response) self.mark_req_logged(env) start_response_args = [None] input_proxy = InputProxy(env['wsgi.input']) env['wsgi.input'] = input_proxy start_time = time.time() def my_start_response(status, headers, exc_info=None): start_response_args[0] = (status, list(headers), exc_info) def status_int_for_logging(client_disconnect=False, start_status=None): # log disconnected clients as '499' status code if client_disconnect or input_proxy.client_disconnect: ret_status_int = 499 elif start_status is None: ret_status_int = int( start_response_args[0][0].split(' ', 1)[0]) else: ret_status_int = start_status return ret_status_int def iter_response(iterable): iterator = iter(iterable) try: chunk = next(iterator) while not chunk: chunk = next(iterator) except StopIteration: chunk = '' for h, v in start_response_args[0][1]: if h.lower() in ('content-length', 'transfer-encoding'): break else: if not chunk: start_response_args[0][1].append(('Content-Length', '0')) elif isinstance(iterable, list): start_response_args[0][1].append( ('Content-Length', str(sum(len(i) for i in iterable)))) resp_headers = dict(start_response_args[0][1]) start_response(*start_response_args[0]) req = Request(env) # Log timing information for time-to-first-byte (GET requests only) method = self.method_from_req(req) if method == 'GET': status_int = status_int_for_logging() policy_index = get_policy_index(req.headers, resp_headers) metric_name = self.statsd_metric_name(req, status_int, method) metric_name_policy = self.statsd_metric_name_policy( req, status_int, method, policy_index) if metric_name: self.access_logger.timing_since( metric_name + '.first-byte.timing', start_time) if metric_name_policy: self.access_logger.timing_since( metric_name_policy + '.first-byte.timing', start_time) bytes_sent = 0 client_disconnect = False try: while chunk: bytes_sent += len(chunk) yield chunk chunk = next(iterator) except GeneratorExit: # generator was closed before we finished client_disconnect = True raise finally: status_int = status_int_for_logging(client_disconnect) self.log_request( req, status_int, input_proxy.bytes_received, bytes_sent, start_time, time.time(), resp_headers=resp_headers) close_method = getattr(iterable, 'close', None) if callable(close_method): close_method() try: iterable = self.app(env, my_start_response) except Exception: exc_type, exc_value, exc_traceback = sys.exc_info() req = Request(env) status_int = status_int_for_logging(start_status=500) self.log_request( req, status_int, input_proxy.bytes_received, 0, start_time, time.time()) six.reraise(exc_type, exc_value, exc_traceback) else: return iter_response(iterable) def filter_factory(global_conf, **local_conf): conf = global_conf.copy() conf.update(local_conf) def proxy_logger(app): return ProxyLoggingMiddleware(app, conf) return proxy_logger