You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by "MeiK (JIRA)" <ji...@apache.org> on 2019/08/02 07:23:00 UTC
[jira] [Created] (AIRFLOW-5096) reduce the number of times the
pickle is inserted into the database by modifying the hash field of Dag
MeiK created AIRFLOW-5096:
-----------------------------
Summary: reduce the number of times the pickle is inserted into the database by modifying the hash field of Dag
Key: AIRFLOW-5096
URL: https://issues.apache.org/jira/browse/AIRFLOW-5096
Project: Apache Airflow
Issue Type: Improvement
Components: DAG
Affects Versions: 1.10.3
Reporter: MeiK
Assignee: MeiK
After the scheduler has the --do_pickle option turned on, the scheduler will insert all the file pickles into the database each time it scans the file, which will cause the database to swell rapidly.
In my opinion, the main reason is because the hash function that determines whether the dag is the same as the pickle version uses the last_loaded field, which changes every time it is read instead of modified. Therefore, airflow inserts a large amount of unchanging data into the database.
I created a commit in which the last modified time of the file was used instead of last_loaded as the hash field, which works fine on my computer. Please let me know if you have a better way.
English is not my native language; please excuse typing errors.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)