How can I log all my requests and responses (headers and body) in Django using a middleware? I'm using Django 2.2 with Django rest framework, so sometimes the requests and responses are original Django type, sometimes of drf. The app is served behind gunicorn. I've developed middleware but the main problem is I can't read request's body twice as it gives me error.
-
Does this answer your question? [Exception: You cannot access body after reading from request's data stream](https://stackoverflow.com/questions/19581110/exception-you-cannot-access-body-after-reading-from-requests-data-stream) – SoroushA Jul 24 '20 at 17:32
-
@sorousha - No, I've tried using the suggested approach (using `request.read()`) but not my solution. – Masked Man Jul 24 '20 at 18:21
3 Answers
Here's an example of logging requests in the database.
Note: This will hit the database once extra with each request. So it will slow down the response time.
models.py
class Request(models.Model):
endpoint = models.CharField(max_length=100, null=True) # The url the user requested
user = models.ForeignKey(User, on_delete=models.SET_NULL, null=True) # User that made request, if authenticated
response_code = models.PositiveSmallIntegerField() # Response status code
method = models.CharField(max_length=10, null=True) # Request method
remote_address = models.CharField(max_length=20, null=True) # IP address of user
exec_time = models.IntegerField(null=True) # Time taken to create the response
date = models.DateTimeField(auto_now=True) # Date and time of request
body_response = models.TextField() # Response data
body_request = models.TextField() # Request data
middleware.py
class SaveRequest:
def __init__(self, get_response):
self.get_response = get_response
# Filter to log all request to url's that start with any of the strings below.
# With example below:
# /example/test/ will be logged.
# /other/ will not be logged.
self.prefixs = [
'/example'
]
def __call__(self, request):
_t = time.time() # Calculated execution time.
response = self.get_response(request) # Get response from view function.
_t = int((time.time() - _t)*1000)
# If the url does not start with on of the prefixes above, then return response and dont save log.
# (Remove these two lines below to log everything)
if not list(filter(request.get_full_path().startswith, self.prefixs)):
return response
# Create instance of our model and assign values
request_log = Request(
endpoint=request.get_full_path(),
response_code=response.status_code,
method=request.method,
remote_address=self.get_client_ip(request),
exec_time=_t,
body_response=str(response.content),
body_request=str(request.body)
)
# Assign user to log if it's not an anonymous user
if not request.user.is_anonymous:
request_log.user = request.user
# Save log in db
request_log.save()
return response
# get clients ip address
def get_client_ip(self, request):
x_forwarded_for = request.META.get('HTTP_X_FORWARDED_FOR')
if x_forwarded_for:
_ip = x_forwarded_for.split(',')[0]
else:
_ip = request.META.get('REMOTE_ADDR')
return _ip
settings.py
# Activate the middleware in settings.py like this.
MIDDLEWARE = [
... # Django default middleware
'<your_appname>.middleware.SaveRequest'
]

- 3,253
- 2
- 10
- 27
-
3
-
Why? You ask how you can log requests and responsens and this does exactly that... – Felix Eklöf Jul 31 '20 at 07:50
-
Also, if you write that you've done something an "got an error". No one can help you if you don't provide the error you got. – Felix Eklöf Jul 31 '20 at 07:52
-
By this approach I can't log request body, it can only be accessed once (by logger or by view) – Masked Man Jul 31 '20 at 10:22
-
Do str(response.content) instead. I tried it with my code above and it worked with both Django response and DRF response. – Felix Eklöf Jul 31 '20 at 12:05
-
-
the response has no limitation, the main PITA is request object which it's body can only be accessed once – Masked Man Jul 31 '20 at 18:27
-
Doing str(request.body) works fine for me. Tested it now. However I have a newer version of Django. So you might have to test on your version. I've updated anwser. – Felix Eklöf Aug 01 '20 at 07:53
-
Does `str(request.body)` return the actual request body content or is it empty? – Masked Man Aug 01 '20 at 17:43
-
-
1Oh I figured it out, the `body` property is actually a `bytes` object, so calling `str(request.body)` is doing the something like `copy.copy(request.body)` because strings are immutable in python. **+1** however. – Masked Man Aug 02 '20 at 13:20
Previously, I have the same problem. the request
and response
body is caputed at all. This example below is what I was configured in my previous project with a mongodb (using mongolog
module).
You can also remove the
mongolog
(mongodb) configuration to use default file logger. I choose the mongoDB to easily track therequest
,response
body orurl
also.
1. middleware.py
import json
import logging
import traceback
from django.http import Http404
from django.http.request import RawPostDataException
from django.utils.deprecation import MiddlewareMixin
# if you want to track it
# from sentry_sdk import capture_exception
def get_client_ip(request):
"""
get client ip address
used in:
- apps/models_helper/visitor.py
"""
ip_address = request.META.get('HTTP_X_FORWARDED_FOR', None)
if ip_address:
ip_address = ip_address.split(', ')[0]
else:
ip_address = request.META.get('REMOTE_ADDR', '')
return ip_address
class SessionLogMiddleware(MiddlewareMixin):
"""
this middleware to create a log
by current user
"""
def save_mongodb_logging(self, request, response=None, exception=None, status_code=None):
"""
{
"_id" : ObjectId("5d2d7200b2b76e29fc94080f"),
"module" : "middleware",
"line" : 94,
"thread" : NumberLong(140471566464768),
"created" : ISODate("2019-07-16T13:43:12.820Z"),
"level" : "INFO",
"path" : "/home/projectname/envs/env-projectname/projectname-django/apps/utils/middleware.py",
"msg" : {
"REQUEST_DATA" : {
"start_time" : "2019-07-04 16:38:01",
"finish_time" : "2019-07-04 16:39:22",
"quiz_choices_id" : [5, 30, 1, 3, 9]
},
"METHOD" : "POST",
"URL" : "http://localhost:8000/api/quiz",
"RESPONSE_DATA" : {
"message" : "Answer successfully created!",
"customer_quiz" : {
"customer_quiz_id" : 34,
"created_at" : "2019-07-16T13:43:12.538089",
"total_correct_answers" : 3,
"customer" : 1,
"start_time" : "2019-07-16T13:43:12.538062",
"total_incorrect_answers" : 2,
"deleted_at" : null,
"finish_time" : "2019-07-04T16:39:22",
"updated_at" : "2019-07-16T13:43:12.538098"
},
"action" : true
},
"HEADERS" : {
"Cache-Control" : "no-cache",
"Accept-Encoding" : "gzip, deflate",
"User-Agent" : "PostmanRuntime/7.6.1",
"Connection" : "keep-alive",
"Content-Type" : "application/json",
"Cookie" : "sessionid=ugs3iaw9ltqtdrh65rhfbpo2ct6uq83o",
"Accept" : "*/*",
"Content-Length" : "116",
"Host" : "localhost:8000",
"Authorization" : "Token 5dace55366d8f36102c04c7b9fbdbdd7352f2ffc",
"Postman-Token" : "61873e1e-0c1b-43f0-b048-a84bbd87ca4f"
},
"USER" : {
"ip_address": "186.291.22.5",
"email" : "user@gmail.com",
"phone_number" : "0821xxxxx",
"username" : "si-fulan",
"model": "Customer",
"id" : 1
},
"ERROR_MESSAGE": null,
"STATUS_CODE": 200
},
"func" : "save_mongodb_logging",
"dates" : [
ISODate("2019-07-16T13:43:12.820Z")
],
"name" : "logger:mongolog",
"counter" : 1,
"uuid" : "ebe40b9a7bfb598ab3ec1c473fb93e31",
"process" : 26388,
"filename" : "middleware.py"
}
"""
# this `user` usage for models.User or models.Customer
user = request._cached_user if hasattr(request, '_cached_user') else request.user
if hasattr(request, 'auth') and request.auth:
if hasattr(request.auth, 'customer'):
user = request.auth.customer or user
headers = eval(str(request.headers))
response_data = None
request_data = None
def clean_text(text):
if isinstance(text, bytes):
try:
return text.decode('utf-8')\
.replace('\\n', '')\
.replace('\\t', '')\
.replace('\\r', '')
except Exception:
pass
return str(text)
try:
request_data = json.loads(clean_text(request.body))
except RawPostDataException:
response_data = "RawPostDataException: You cannot access body after reading from request's data stream"
except Exception:
try:
request_data = clean_text(request.body)
except Exception:
pass
if not headers.get('Content-Type') == 'application/x-www-form-urlencoded':
try:
response_data = json.loads(clean_text(response.content))
except RawPostDataException:
response_data = "RawPostDataException: You cannot access body after reading from request's data stream"
except Exception:
try:
response_data = clean_text(response.content)
except Exception:
pass
log_data = {
'HEADERS': headers,
'METHOD': request.method,
'USER': {
'id': user.pk if hasattr(user, 'pk') else None,
'username': user.username if hasattr(user, 'username') else None,
'phone_number': user.phone_number if hasattr(user, 'phone_number') else None,
'email': user.email if hasattr(user, 'email') else None,
'model': 'User' if hasattr(user, 'is_superuser') else 'Customer',
'ip_address': get_client_ip(request)
},
'URL': request.build_absolute_uri(),
'REQUEST_DATA': request_data,
'RESPONSE_DATA': response_data,
'ERROR_MESSAGE': exception,
'STATUS_CODE': status_code
}
log = logging.getLogger('logger:mongolog')
log.error(log_data) if exception else log.debug(log_data)
def process_exception(self, request, exception):
# print(type(exception), exception) # just for debug
status_code = 404 if isinstance(exception, Http404) else 500
try:
self.save_mongodb_logging(request,
exception=exception,
status_code=status_code)
except Exception:
# error = traceback.format_exc()
# capture_exception(error)
print(error)
return # no return anything
def process_response(self, request, response):
if not response.status_code >= 400:
try:
response_data = {'request': request,
'response': response,
'status_code': response.status_code}
self.save_mongodb_logging(**response_data)
except Exception:
pass
return response
2. logger.py
import datetime
from django.conf import settings
MONGODB_CONNECTION_URL = getattr(settings, 'MONGODB_CONNECTION_URL', None)
# All logging handlers configurations.
# 'propagate': False = mean is the error logs ins't duplicates to another file logs.
NOW = datetime.datetime.now()
DAY_NAME = NOW.strftime('%A').lower()
MAXIMUM_FILE_LOGS = 1024 * 1024 * 10 # 10 MB
BACKUP_COUNT = 5
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'formatters': {
'standard': {
'format': '[%(levelname)s] %(asctime)s %(name)s: %(message)s'
},
},
'handlers': {
'default': {
'level': 'DEBUG',
'class': 'logging.handlers.RotatingFileHandler',
'filename': 'logs/default.log',
'maxBytes': MAXIMUM_FILE_LOGS,
'backupCount': BACKUP_COUNT,
'formatter': 'standard',
},
'mongolog': {
'level': 'DEBUG',
'class': 'mongolog.SimpleMongoLogHandler',
# Set the connection string to the mongo instance.
# mongodb://[username]:[password]@[host]:[port]/[database]
'connection': MONGODB_CONNECTION_URL,
# define mongo collection the log handler should use. Default is mongolog
# This is useful if you want different handlers to use different collections
'collection': 'mongolog',
'formatter': 'standard',
},
'request_debug_handler': {
'level': 'DEBUG',
'class': 'logging.handlers.RotatingFileHandler',
'filename': 'logs/request_debug.log',
'maxBytes': MAXIMUM_FILE_LOGS,
'backupCount': BACKUP_COUNT,
'formatter': 'standard',
},
'request_error_handler': {
'level': 'ERROR',
'class': 'logging.handlers.RotatingFileHandler',
'filename': 'logs/request_error.log',
'maxBytes': MAXIMUM_FILE_LOGS,
'backupCount': BACKUP_COUNT,
'formatter': 'standard',
},
'mail_admins_handler': {
'level': 'ERROR',
'class': 'django.utils.log.AdminEmailHandler',
'email_backend': 'django.core.mail.backends.smtp.EmailBackend'
},
},
'root': {
'handlers': ['default', 'mongolog'],
'level': 'DEBUG'
},
'loggers': {
'django.request': {
'handlers': [
'mongolog',
'request_debug_handler',
'request_error_handler',
'mail_admins_handler'
],
'level': 'DEBUG',
'propagate': False
},
}
}
3. settings.py
LOGGING = LOGGING # from the `logger.py` file.
MONGODB_CONNECTION_URL = 'mongodb://db_user:db_password@127.0.0.1:27017/mongolog'
MIDDLEWARE = [
....
'path.to.middleware.SessionLogMiddleware',
]

- 3,994
- 3
- 17
- 54
-
I don't understand where you are getting actual response processed by view. – Masked Man Aug 01 '20 at 17:46
I originally tried to do something like request = copy.copy(request)
but clearly it's a mistake because shallow copying does not copy nested objects. So the correct approach is (__call__
is middleware's class instance method):
def __call__(self, request):
request_body = copy.copy(request.body)
# Here goes more code for further processing
# and now I can safely use request_body before or after
# the view code runs
Update
As Felix Eklöf suggested, you can also use str(request.body)
and python will handle copying the body contents, cause strings are immutable in python. (I guess it has better readability too).

- 2,176
- 2
- 22
- 41