Collate output in Python logging MemoryHandler wit

Question

I have the logging module MemoryHandler set up to queue debug and error messages for the SMTPHandler target. What I want is for an email to be sent when the process errors that contains all debug statements up to that point (one per line). What I get instead is a separate email for every debug message.

This seems like it should be trivial, and part of the logging package, but I can't find anything about it, no examples, nothing on Google.

log = logging.getLogger()
log.setLevel(logging.DEBUG)
debug_format = logging.Formatter("%(levelname)s at %(asctime)s in %(filename)s (line %(lineno)d):: %(message)s")

# write errors to email
error_mail_subject = "ERROR: Script error in %s on %s" % (sys.argv[0], os.uname()[1])
error_mail_handler = logging.handlers.SMTPHandler(SMTP_HOST, 'errors@'+os.uname()[1], [LOG_EMAIL], error_mail_subject)
error_mail_handler.setLevel(logging.ERROR)
#error_mail_handler.setLevel(logging.DEBUG)
error_mail_handler.setFormatter(debug_format)

# buffer debug messages so they can be sent with error emails
memory_handler = logging.handlers.MemoryHandler(1024*10, logging.ERROR, error_mail_handler)
memory_handler.setLevel(logging.DEBUG)

# attach handlers
log.addHandler(memory_handler)
log.addHandler(error_mail_handler)

Related to this:

Do I need to add the error_mail_handler to the logger explicitly if it is a target of memory_handler anyway? Should error_mail_handler be set to DEBUG or ERROR target? Does it even need a target when it is being fed from memory_handler?

Would love to see some working code from anyone who has solved this problem.

Answer 1

You might want to use or adapt the BufferingSMTPHandler which is in this test script.

In general, you don't need to add a handler to a logger if it's the target of a MemoryHandler handler which has been added to a logger. If you set the level of a handler, that will affect what the handler actually processes - it won't process anything which is less severe than its level setting.

Answer 2

If you are using django - here is simple buffering handler, which will use standard django email methods:

import logging

from django.conf import settings
from django.core.mail import EmailMessage


class DjangoBufferingSMTPHandler(logging.handlers.BufferingHandler):
    def __init__(self, capacity, toaddrs=None, subject=None):
        logging.handlers.BufferingHandler.__init__(self, capacity)

        if toaddrs:
            self.toaddrs = toaddrs
        else:
            # Send messages to site administrators by default
            self.toaddrs = zip(*settings.ADMINS)[-1]

        if subject:
            self.subject = subject
        else:
            self.subject = 'logging'

    def flush(self):
        if len(self.buffer) == 0:
            return

        try:
            msg = "\r\n".join(map(self.format, self.buffer))
            emsg = EmailMessage(self.subject, msg, to=self.toaddrs)
            emsg.send()
        except Exception:
            # handleError() will print exception info to stderr if logging.raiseExceptions is True
            self.handleError(record=None)
        self.buffer = []

In django settings.py you will need to configure email and logging like this:

EMAIL_USE_TLS = True
EMAIL_PORT = 25  
EMAIL_HOST = ''  # example: 'smtp.yandex.ru'
EMAIL_HOST_USER = ''  # example: '[e-mail]'
EMAIL_HOST_PASSWORD = ''
DEFAULT_FROM_EMAIL = EMAIL_HOST_USER
SERVER_EMAIL = EMAIL_HOST_USER

LOGGING = {
    'handlers': {
        ...
        'mail_buffer': {
            'level': 'WARN',
            'capacity': 9999,
            'class': 'utils.logging.DjangoBufferingSMTPHandler',
            # optional: 
            # 'toaddrs': '[e-mail]'
            # 'subject': 'log messages'
        }
    },
    ...
}

Answer 3

I think the point about the SMTP logger is that it is meant to send out a significant log message functioning as some kind of alert if sent to a human recipient or else to be further processed by an automated recipient.

If a collection of log messages is to be sent by email then that constitutes a report being sent at the end of execution of a task and writing that log to a file and then emailing the file would seem to be a reasonable solution.

I took a look at the basic FileHandler log handler and how to build a mechanism to write to a temp file then attach that temp file when the script exits.

I found the "atexit" module that allows for a method to be registered that will be executed against an object when the script is exiting.

import logging
import smtplib
from email.mime.multipart import MIMEMultipart
from email.mime.text import MIMEText
from email.mime.base import MIMEBase
import os
from email import encoders
import uuid
# atexit allows for a method to be set to handle an object when the script             exits
import atexit
filename = uuid.uuid4().hex
class MailLogger:

def __init__(self, filePath, smtpDict):
    self.filePath = filePath
    self.smtpDict = smtpDict
    # Generate random file name
    filename = '%s.txt' % ( uuid.uuid4().hex )
    # Create full filename
    filename = '%s/%s' % (filePath,filename)
    self.filename = filename
    self.fileLogger = logging.getLogger('mailedLog')
    self.fileLogger.setLevel(logging.INFO)
    self.fileHandler = logging.FileHandler(filename)
    self.fileHandler.setLevel(logging.INFO)
    formatter = logging.Formatter('%(name)s - %(levelname)s - %(message)s')
    self.fileHandler.setFormatter(formatter)
    self.fileLogger.addHandler(self.fileHandler)
    atexit.register(self.mailOut)

def mailOut(self):
    '''
    Script is exiting so time to mail out the log file

    "emailSettings":    {
                        "smtpServer"    :     "smtp.dom.com",
                        "smtpPort"        :    25,
                        "sender"        :    "[e-mail]>",
                        "recipients"    :    [
                                                "[e-mail]"
                                            ],
                        "subject"        :    "Email Subject"
},
    '''
    # Close the file handler
    smtpDict = self.smtpDict
    self.fileHandler.close()
    msg = MIMEMultipart('alternative')
    s = smtplib.SMTP(smtpDict["smtpServer"], smtpDict["smtpPort"] )        
    msg['Subject'] = smtpDict["subject"]
    msg['From'] = smtpDict["sender"]
    msg['To'] = ','.join(smtpDict["recipients"])
    body = 'See attached report file'    
    content = MIMEText(body, 'plain')
    msg.attach(content)
    attachment = MIMEBase('application', 'octet-stream')
    attachment.set_payload(open(self.filename, 'rb').read())
    encoders.encode_base64(attachment)
    attachment.add_header('Content-Disposition', 'attachment; filename="%s"' % os.path.basename(self.filename))
    msg.attach(attachment)
    s.send_message(msg)
    s.quit()

My basic test script is:

from EmailLogRpt import MailLogger
import time
smtpDict = {
                        "smtpServer"    :     "smtp.dom.com",
                        "smtpPort"        :    25,
                        "sender"        :    "[e-mail]",
                        "recipients"    :    [
                                                "[e-mail]>"
                                            ],
                        "subject"        :    "Email Subject"
}
myMailLogger = MailLogger("/home/ed/tmp",smtpDict).fileLogger
myMailLogger.info("test msg 1")
time.sleep(5)
myMailLogger.info("test msg 2")

Hope this helps somebody.

Answer 4

For this purpose I use the BufferingSMTPHandler suggested by Vinay Sajip with one minor tweak: I set the buffer length to something really big (say 5000 log records) and manualy call the flush method of the handler every some seconds and after checking for internet conectivity.

# init
log_handler1 = BufferingSMTPHandler(
    'smtp.host.lala', "[e-mail]", ['[e-mail]'], 'Log event(s)',5000)
...
logger.addHandler(log_handler1)
...

# main code
...
if internet_connection_ok and seconds_since_last_flush>60:
    log_handler1.flush() # send buffered log records (if any)

Answer 5

I wrote my own beastly thread-safe implementation of BufferingSMTPHandler which sends emails from a separate thread. The primary goal is to not block the main thread.

As written, it uses two queues - this seemed necessary in order to implement some useful class-level parameters that are defined in the "Configurable parameters" section of the code. Although you can use the code as-is, it's probably better if you study and use it to write your own class.

Issues:

• Some class-level parameters can perhaps be instance-level instead.
• Either threading.Timer or the signal module could perhaps be used to avoid loops that run forever.