|
Profiling
How to use the integrated profiler.
Enabling the profilerIn your settings.py just add the following: ENABLE_PROFILER = True Tuning the resultsThe profiler is configured via settings.py. You can restrict the profile to certain functions by specifying a regular expression: PROFILE_PATTERN = '.*token.*' # matches all function names containing 'token' The maximum number of functions can be set with: MAX_PROFILE_RESULTS = 160 # default is 80 You can change the sorting criteria with: SORT_PROFILE_RESULTS_BY = 'cumulative' # default is 'time' It's possible to additionally get the callers and callees profile: EXTRA_PROFILE_OUTPUT = 'callers' # or 'callees' (you can also provide a tuple with multiple values) Profiling under loadIf you profile a website under load your logs will be flooded. You can limit the profiler to just a certain percentage of all requests: PROFILE_PERCENTAGE = 2.5 # Profile only 2.5% of all requests Alternatively, you can limit the profiler to requests that contain "profile=forced" in their query string: ONLY_FORCED_PROFILE = True How to profile datastore callsPut this in your settings.py to profile all important datastore calls. ENABLE_PROFILER = True SORT_PROFILE_RESULTS_BY = 'cumulative' # default is 'time' PROFILE_PATTERN = 'ext.db..+\((?:get|get_by_key_name|get_by_id|fetch|count|put)\)' Note that the db.Model.get(), db.Model.get_by_key_name(), db.Model.get_by_id(), and db.Model.put() calls translate into db.get() and db.put(), so for every call of these model functions you'll also get a call of the global db functions. Please take a look at the ragendja.dbutils.prefetch_references() function (see the documentation). It can help optimize your code. |
