Some of those 404's in our log mean that we forgot to package something; for example:
Removing the 404's would prevent us from catching important errors if ever they occur. I suggest we fix the 404's; some of them have a cascading effect, like the font-awesome missing "webfonts" folder I just fixed that leads to a lot of unnecessary 404s.
For error handling, I think that will evolve with the code-base with time as it gets leaner. One way the aforementioned flask page recommends handling errors is by creating custom errors for your particular subject-domain needs and probably send a 5xx message that's more helpful. At the moment we are using one error-handler for everything(notice the generic "Exception")! See:
@app.errorhandler(Exception)
def handle_bad_request(e):
err_msg = str(e)
logger.error(err_msg)
logger.error(request.url)
# get the stack trace and send it to the logger
exc_type, exc_value, exc_traceback = sys.exc_info()
logger.error(traceback.format_exc())
now = datetime.datetime.utcnow()
time_str = now.strftime('%l:%M%p UTC %b %d, %Y')
formatted_lines = [request.url
+ " (" + time_str + ")"] + traceback.format_exc().splitlines()
# Handle random animations
# Use a cookie to have one animation on refresh
animation = request.cookies.get(err_msg[:32])
if not animation:
list = [fn for fn in os.listdir(
"./wqflask/static/gif/error") if fn.endswith(".gif")]
animation = random.choice(list)
resp = make_response(render_template("error.html", message=err_msg,
stack=formatted_lines, error_image=animation, version=GN_VERSION))
# logger.error("Set cookie %s with %s" % (err_msg, animation))
resp.set_cookie(err_msg[:32], animation)
return resp
I'll look a way for removing this:
exc_type, exc_value, exc_traceback = sys.exc_info()
logger.error(traceback.format_exc())
with a more sane, but probably still useful, log. That's where I problem IIUC lies.
The problem of improving better logging has been documented in:
Most recently, all noisy logging was removed in:
We are aware that at the moment we are using one error-handler for everything, and this, together with any other logging issues should be tackled in another separate issue.