def my_update_subtask_status(entry_id, current_task_id, new_subtask_status): """ Check whether a subtask has been updated before really updating. Check whether a subtask which has been retried has had the retry already write its results here before the code that was invoking the retry had a chance to update this status. This is the norm in "eager" mode (used by tests) where the retry is called and run to completion before control is returned to the code that invoked the retry. If the retries eventually end in failure (e.g. due to a maximum number of retries being attempted), the "eager" code will return the error for each retry as it is popped off the stack. We want to just ignore the later updates that are called as the result of the earlier retries. This should not be an issue in production, where status is updated before a task is retried, and is then updated afterwards if the retry fails. """ entry = InstructorTask.objects.get(pk=entry_id) subtask_dict = json.loads(entry.subtasks) subtask_status_info = subtask_dict['status'] current_subtask_status = SubtaskStatus.from_dict( subtask_status_info[current_task_id]) current_retry_count = current_subtask_status.get_retry_count() new_retry_count = new_subtask_status.get_retry_count() if current_retry_count <= new_retry_count: update_subtask_status(entry_id, current_task_id, new_subtask_status)
def my_update_subtask_status(entry_id, current_task_id, new_subtask_status): """ Check whether a subtask has been updated before really updating. Check whether a subtask which has been retried has had the retry already write its results here before the code that was invoking the retry had a chance to update this status. This is the norm in "eager" mode (used by tests) where the retry is called and run to completion before control is returned to the code that invoked the retry. If the retries eventually end in failure (e.g. due to a maximum number of retries being attempted), the "eager" code will return the error for each retry as it is popped off the stack. We want to just ignore the later updates that are called as the result of the earlier retries. This should not be an issue in production, where status is updated before a task is retried, and is then updated afterwards if the retry fails. """ entry = InstructorTask.objects.get(pk=entry_id) subtask_dict = json.loads(entry.subtasks) subtask_status_info = subtask_dict['status'] current_subtask_status = SubtaskStatus.from_dict(subtask_status_info[current_task_id]) current_retry_count = current_subtask_status.get_retry_count() new_retry_count = new_subtask_status.get_retry_count() if current_retry_count <= new_retry_count: update_subtask_status(entry_id, current_task_id, new_subtask_status)
def send_course_email(entry_id, email_id, to_list, global_email_context, subtask_status_dict): """ Sends an email to a list of recipients. Inputs are: * `entry_id`: id of the InstructorTask object to which progress should be recorded. * `email_id`: id of the CourseEmail model that is to be emailed. * `to_list`: list of recipients. Each is represented as a dict with the following keys: - 'profile__name': full name of User. - 'email': email address of User. - 'pk': primary key of User model. * `global_email_context`: dict containing values that are unique for this email but the same for all recipients of this email. This dict is to be used to fill in slots in email template. It does not include 'name' and 'email', which will be provided by the to_list. * `subtask_status_dict` : dict containing values representing current status. Keys are: 'task_id' : id of subtask. This is used to pass task information across retries. 'attempted' : number of attempts -- should equal succeeded plus failed 'succeeded' : number that succeeded in processing 'skipped' : number that were not processed. 'failed' : number that failed during processing 'retried_nomax' : number of times the subtask has been retried for conditions that should not have a maximum count applied 'retried_withmax' : number of times the subtask has been retried for conditions that should have a maximum count applied 'state' : celery state of the subtask (e.g. QUEUING, PROGRESS, RETRY, FAILURE, SUCCESS) Most values will be zero on initial call, but may be different when the task is invoked as part of a retry. Sends to all addresses contained in to_list that are not also in the Optout table. Emails are sent multi-part, in both plain text and html. Updates InstructorTask object with status information (sends, failures, skips) and updates number of subtasks completed. """ subtask_status = SubtaskStatus.from_dict(subtask_status_dict) current_task_id = subtask_status.task_id num_to_send = len(to_list) log.info(("Preparing to send email %s to %d recipients as subtask %s " "for instructor task %d: context = %s, status=%s, time=%s"), email_id, num_to_send, current_task_id, entry_id, global_email_context, subtask_status, datetime.now()) # Check that the requested subtask is actually known to the current InstructorTask entry. # If this fails, it throws an exception, which should fail this subtask immediately. # This can happen when the parent task has been run twice, and results in duplicate # subtasks being created for the same InstructorTask entry. This can happen when Celery # loses its connection to its broker, and any current tasks get requeued. # We hope to catch this condition in perform_delegate_email_batches() when it's the parent # task that is resubmitted, but just in case we fail to do so there, we check here as well. # There is also a possibility that this task will be run twice by Celery, for the same reason. # To deal with that, we need to confirm that the task has not already been completed. check_subtask_is_valid(entry_id, current_task_id, subtask_status) send_exception = None new_subtask_status = None try: start_time = time.time() new_subtask_status, send_exception = _send_course_email( entry_id, email_id, to_list, global_email_context, subtask_status, ) log.info( "BulkEmail ==> _send_course_email completed in : %s for task : %s with recipient count: %s", time.time() - start_time, subtask_status.task_id, len(to_list) ) except Exception: # Unexpected exception. Try to write out the failure to the entry before failing. log.exception("Send-email task %s for email %s: failed unexpectedly!", current_task_id, email_id) # We got here for really unexpected reasons. Since we don't know how far # the task got in emailing, we count all recipients as having failed. # It at least keeps the counts consistent. subtask_status.increment(failed=num_to_send, state=FAILURE) update_subtask_status(entry_id, current_task_id, subtask_status) raise if send_exception is None: # Update the InstructorTask object that is storing its progress. log.info("Send-email task %s for email %s: succeeded", current_task_id, email_id) update_subtask_status(entry_id, current_task_id, new_subtask_status) elif isinstance(send_exception, RetryTaskError): # If retrying, a RetryTaskError needs to be returned to Celery. # We assume that the the progress made before the retry condition # was encountered has already been updated before the retry call was made, # so we only log here. log.warning("Send-email task %s for email %s: being retried", current_task_id, email_id) raise send_exception # pylint: disable=raising-bad-type else: log.error("Send-email task %s for email %s: failed: %s", current_task_id, email_id, send_exception) update_subtask_status(entry_id, current_task_id, new_subtask_status) raise send_exception # pylint: disable=raising-bad-type # return status in a form that can be serialized by Celery into JSON: log.info("Send-email task %s for email %s: returning status %s", current_task_id, email_id, new_subtask_status) return new_subtask_status.to_dict()
def send_course_email(entry_id, email_id, to_list, global_email_context, subtask_status_dict): """ Sends an email to a list of recipients. Inputs are: * `entry_id`: id of the InstructorTask object to which progress should be recorded. * `email_id`: id of the CourseEmail model that is to be emailed. * `to_list`: list of recipients. Each is represented as a dict with the following keys: - 'profile__name': full name of User. - 'email': email address of User. - 'pk': primary key of User model. * `global_email_context`: dict containing values that are unique for this email but the same for all recipients of this email. This dict is to be used to fill in slots in email template. It does not include 'name' and 'email', which will be provided by the to_list. * `subtask_status_dict` : dict containing values representing current status. Keys are: 'task_id' : id of subtask. This is used to pass task information across retries. 'attempted' : number of attempts -- should equal succeeded plus failed 'succeeded' : number that succeeded in processing 'skipped' : number that were not processed. 'failed' : number that failed during processing 'retried_nomax' : number of times the subtask has been retried for conditions that should not have a maximum count applied 'retried_withmax' : number of times the subtask has been retried for conditions that should have a maximum count applied 'state' : celery state of the subtask (e.g. QUEUING, PROGRESS, RETRY, FAILURE, SUCCESS) Most values will be zero on initial call, but may be different when the task is invoked as part of a retry. Sends to all addresses contained in to_list that are not also in the Optout table. Emails are sent multi-part, in both plain text and html. Updates InstructorTask object with status information (sends, failures, skips) and updates number of subtasks completed. """ subtask_status = SubtaskStatus.from_dict(subtask_status_dict) current_task_id = subtask_status.task_id num_to_send = len(to_list) log.info((u"Preparing to send email %s to %d recipients as subtask %s " u"for instructor task %d: context = %s, status=%s"), email_id, num_to_send, current_task_id, entry_id, global_email_context, subtask_status) # Check that the requested subtask is actually known to the current InstructorTask entry. # If this fails, it throws an exception, which should fail this subtask immediately. # This can happen when the parent task has been run twice, and results in duplicate # subtasks being created for the same InstructorTask entry. This can happen when Celery # loses its connection to its broker, and any current tasks get requeued. # We hope to catch this condition in perform_delegate_email_batches() when it's the parent # task that is resubmitted, but just in case we fail to do so there, we check here as well. # There is also a possibility that this task will be run twice by Celery, for the same reason. # To deal with that, we need to confirm that the task has not already been completed. check_subtask_is_valid(entry_id, current_task_id, subtask_status) send_exception = None new_subtask_status = None try: course_title = global_email_context['course_title'] with dog_stats_api.timer('course_email.single_task.time.overall', tags=[_statsd_tag(course_title)]): new_subtask_status, send_exception = _send_course_email( entry_id, email_id, to_list, global_email_context, subtask_status, ) except Exception: # Unexpected exception. Try to write out the failure to the entry before failing. log.exception("Send-email task %s for email %s: failed unexpectedly!", current_task_id, email_id) # We got here for really unexpected reasons. Since we don't know how far # the task got in emailing, we count all recipients as having failed. # It at least keeps the counts consistent. subtask_status.increment(failed=num_to_send, state=FAILURE) update_subtask_status(entry_id, current_task_id, subtask_status) raise if send_exception is None: # Update the InstructorTask object that is storing its progress. log.info("Send-email task %s for email %s: succeeded", current_task_id, email_id) update_subtask_status(entry_id, current_task_id, new_subtask_status) elif isinstance(send_exception, RetryTaskError): # If retrying, a RetryTaskError needs to be returned to Celery. # We assume that the the progress made before the retry condition # was encountered has already been updated before the retry call was made, # so we only log here. log.warning("Send-email task %s for email %s: being retried", current_task_id, email_id) raise send_exception # pylint: disable=raising-bad-type else: log.error("Send-email task %s for email %s: failed: %s", current_task_id, email_id, send_exception) update_subtask_status(entry_id, current_task_id, new_subtask_status) raise send_exception # pylint: disable=raising-bad-type # return status in a form that can be serialized by Celery into JSON: log.info("Send-email task %s for email %s: returning status %s", current_task_id, email_id, new_subtask_status) return new_subtask_status.to_dict()