Xiaodong DENG created AIRFLOW-2814:
--------------------------------------

             Summary: Default Arg "file_process_interval" for class 
SchedulerJob is inconsistent with doc
                 Key: AIRFLOW-2814
                 URL: https://issues.apache.org/jira/browse/AIRFLOW-2814
             Project: Apache Airflow
          Issue Type: Bug
          Components: scheduler
            Reporter: Xiaodong DENG
            Assignee: Xiaodong DENG


h2. Backgrond

In 
[https://github.com/XD-DENG/incubator-airflow/blob/master/airflow/jobs.py#L592] 
, it was mentioned the default value of argument *file_process_interval* should 
be 3 minutes (*file_process_interval:* Parse and schedule each file no faster 
than this interval).

The value is normally parsed from the default configuration. However, in the 
default config_template, its value is 0 rather than 180 seconds 
([https://github.com/XD-DENG/incubator-airflow/blob/master/airflow/config_templates/default_airflow.cfg#L432]
 ). 
h2. Issue

This means that actually that each file is parsed and scheduled without letting 
Airflow "rest". This conflicts with the design purpose (by default let it be 
180 seconds) and may affect performance significantly.
h2. My Proposal

Change the value in the config template from 0 to 180.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to