Log exception with traceback in Python

Heres a simple example taken from the python 2.6 documentation:

import logging
LOG_FILENAME = '/tmp/logging_example.out'
logging.basicConfig(filename=LOG_FILENAME,level=logging.DEBUG,)

logging.debug('This message should go to the log file')

My job recently tasked me with logging all the tracebacks/exceptions from our application. I tried numerous techniques that others had posted online such as the one above but settled on a different approach. Overriding traceback.print_exception.

I have a write up at http://www.bbarrows.com/ That would be much easier to read but Ill paste it in here as well.

When tasked with logging all the exceptions that our software might encounter in the wild I tried a number of different techniques to log our python exception tracebacks. At first I thought that the python system exception hook, sys.excepthook would be the perfect place to insert the logging code. I was trying something similar to:

import traceback
import StringIO
import logging
import os, sys

def my_excepthook(excType, excValue, traceback, logger=logger):
    logger.error("Logging an uncaught exception",
                 exc_info=(excType, excValue, traceback))

sys.excepthook = my_excepthook  

This worked for the main thread but I soon found that the my sys.excepthook would not exist across any new threads my process started. This is a huge issue because most everything happens in threads in this project.

After googling and reading plenty of documentation the most helpful information I found was from the Python Issue tracker.

The first post on the thread shows a working example of the sys.excepthook NOT persisting across threads (as shown below). Apparently this is expected behavior.

import sys, threading

def log_exception(*args):
    print 'got exception %s' % (args,)
sys.excepthook = log_exception

def foo():
    a = 1 / 0

threading.Thread(target=foo).start()

The messages on this Python Issue thread really result in 2 suggested hacks. Either subclass Thread and wrap the run method in our own try except block in order to catch and log exceptions or monkey patch threading.Thread.run to run in your own try except block and log the exceptions.

The first method of subclassing Thread seems to me to be less elegant in your code as you would have to import and use your custom Thread class EVERYWHERE you wanted to have a logging thread. This ended up being a hassle because I had to search our entire code base and replace all normal Threads with this custom Thread. However, it was clear as to what this Thread was doing and would be easier for someone to diagnose and debug if something went wrong with the custom logging code. A custome logging thread might look like this:

class TracebackLoggingThread(threading.Thread):
    def run(self):
        try:
            super(TracebackLoggingThread, self).run()
        except (KeyboardInterrupt, SystemExit):
            raise
        except Exception, e:
            logger = logging.getLogger('')
            logger.exception("Logging an uncaught exception")

The second method of monkey patching threading.Thread.run is nice because I could just run it once right after __main__ and instrument my logging code in all exceptions. Monkey patching can be annoying to debug though as it changes the expected functionality of something. The suggested patch from the Python Issue tracker was:

def installThreadExcepthook():
    """
    Workaround for sys.excepthook thread bug
    From
http://spyced.blogspot.com/2007/06/workaround-for-sysexcepthook-bug.html

(https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1230540&group_id=5470).
    Call once from __main__ before creating any threads.
    If using psyco, call psyco.cannotcompile(threading.Thread.run)
    since this replaces a new-style class method.
    """
    init_old = threading.Thread.__init__
    def init(self, *args, **kwargs):
        init_old(self, *args, **kwargs)
        run_old = self.run
        def run_with_except_hook(*args, **kw):
            try:
                run_old(*args, **kw)
            except (KeyboardInterrupt, SystemExit):
                raise
            except:
                sys.excepthook(*sys.exc_info())
        self.run = run_with_except_hook
    threading.Thread.__init__ = init

It was not until I started testing my exception logging I realized that I was going about it all wrong.

To test I had placed a

raise Exception("Test")

somewhere in my code. However, wrapping a a method that called this method was a try except block that printed out the traceback and swallowed the exception. This was very frustrating because I saw the traceback bring printed to STDOUT but not being logged. It was I then decided that a much easier method of logging the tracebacks was just to monkey patch the method that all python code uses to print the tracebacks themselves, traceback.print_exception.
I ended up with something similar to the following:

def add_custom_print_exception():
    old_print_exception = traceback.print_exception
    def custom_print_exception(etype, value, tb, limit=None, file=None):
        tb_output = StringIO.StringIO()
        traceback.print_tb(tb, limit, tb_output)
        logger = logging.getLogger('customLogger')
        logger.error(tb_output.getvalue())
        tb_output.close()
        old_print_exception(etype, value, tb, limit=None, file=None)
    traceback.print_exception = custom_print_exception

This code writes the traceback to a String Buffer and logs it to logging ERROR. I have a custom logging handler set up the ‘customLogger’ logger which takes the ERROR level logs and send them home for analysis.

maybe not as stylish, but easier:

#!/bin/bash
log="/var/log/yourlog"
/path/to/your/script.py 2>&1 | (while read; do echo "$REPLY" >> $log; done)

Use exc_info options may be better, remains warning or error title:

try:
    # coode in here
except Exception as e:
    logging.error(e, exc_info=True)

Uncaught exception messages go to STDERR, so instead of implementing your logging in Python itself you could send STDERR to a file using whatever shell you’re using to run your Python script. In a Bash script, you can do this with output redirection, as described in the BASH guide.

Examples

Append errors to file, other output to the terminal:

./test.py 2>> mylog.log

Overwrite file with interleaved STDOUT and STDERR output:

./test.py &> mylog.log

You can log all uncaught exceptions on the main thread by assigning a handler to sys.excepthook, perhaps using the exc_info parameter of Python’s logging functions:

import sys
import logging

logging.basicConfig(filename='/tmp/foobar.log')

def exception_hook(exc_type, exc_value, exc_traceback):
    logging.error(
        "Uncaught exception",
        exc_info=(exc_type, exc_value, exc_traceback)
    )

sys.excepthook = exception_hook

raise Exception('Boom')

If your program uses threads, however, then note that threads created using threading.Thread will not trigger sys.excepthook when an uncaught exception occurs inside them, as noted in Issue 1230540 on Python’s issue tracker. Some hacks have been suggested there to work around this limitation, like monkey-patching Thread.__init__ to overwrite self.run with an alternative run method that wraps the original in a try block and calls sys.excepthook from inside the except block. Alternatively, you could just manually wrap the entry point for each of your threads in try/except yourself.

What I was looking for:

import sys
import traceback

exc_type, exc_value, exc_traceback = sys.exc_info()
traceback_in_var = traceback.format_tb(exc_traceback)

See:

• https://docs.python.org/3/library/traceback.html

You can get the traceback using a logger, at any level (DEBUG, INFO, …). Note that using logging.exception, the level is ERROR.

# test_app.py
import sys
import logging

logging.basicConfig(level="DEBUG")

def do_something():
    raise ValueError(":(")

try:
    do_something()
except Exception:
    logging.debug("Something went wrong", exc_info=sys.exc_info())

DEBUG:root:Something went wrong
Traceback (most recent call last):
  File "test_app.py", line 10, in <module>
    do_something()
  File "test_app.py", line 7, in do_something
    raise ValueError(":(")
ValueError: :(

EDIT:

This works too (using python 3.6)

logging.debug("Something went wrong", exc_info=True)

Here is a version that uses sys.excepthook

import traceback
import sys

logger = logging.getLogger()

def handle_excepthook(type, message, stack):
     logger.error(f'An unhandled exception occured: {message}. Traceback: {traceback.format_tb(stack)}')

sys.excepthook = handle_excepthook

This is how I do it.

try:
    do_something()
except:
    # How can I log my exception here, complete with its traceback?
    import traceback
    traceback.format_exc() # this will print a complete trace to stout.

To key off of others that may be getting lost in here, the way that works best with capturing it in logs is to use the traceback.format_exc() call and then split this string for each line in order to capture in the generated log file:

import logging
import sys
import traceback

try:
  ...
except Exception as ex:
  # could be done differently, just showing you can split it apart to capture everything individually
  ex_t = type(ex).__name__
  err = str(ex)
  err_msg = f'[{ex_t}] - {err}'
  logging.error(err_msg)

  # go through the trackback lines and individually add those to the log as an error
  for l in traceback.format_exc().splitlines():
    logging.error(l)

For some reason simply using log.exception(exc) didn’t work for me and I had to manually build the exc_info param as in this answer.

log.exception(exc, exc_info=(type(exc), exc, exc.__traceback__))