airflow-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (AIRFLOW-100) Add flexibility to ExternalTaskSensor
Date Thu, 30 Jun 2016 23:43:10 GMT

    [ https://issues.apache.org/jira/browse/AIRFLOW-100?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15358054#comment-15358054
] 

ASF subversion and git services commented on AIRFLOW-100:
---------------------------------------------------------

Commit efdbbb5d3beba49f9b633f0a25ce768f896c0a6a in incubator-airflow's branch refs/heads/master
from jlowin
[ https://git-wip-us.apache.org/repos/asf?p=incubator-airflow.git;h=efdbbb5 ]

[AIRFLOW-100] Add execution_date_fn to ExternalTaskSensor

Currently, ExternalTaskSensor only supports querying execution_dates
that are either the same as the ExternalTaskSensor's execution_date
or a fixed interval from that date (using `execution_delta`). This
adds the ability to provide a fn (`execution_date_fn`) that accepts
the current execution_date and can return any desired date for
querying. This is much more flexible. For example, it could
supply the last date of the previous month.


> Add flexibility to ExternalTaskSensor
> -------------------------------------
>
>                 Key: AIRFLOW-100
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-100
>             Project: Apache Airflow
>          Issue Type: Improvement
>          Components: operators
>            Reporter: Jeremiah Lowin
>            Assignee: Jeremiah Lowin
>            Priority: Minor
>              Labels: operator
>
> The ExternalTaskSensor defaults to sensing tasks with the same {{execution_date}} as
it does, and has an {{execution_delta}} parameter for looking back farther in time. However,
this doesn't support the case where the sensing task has a smaller schedule_interval than
the target task.
> For example, if the ETS were run every hour, one couldn't use a fixed {{execution_delta}}
to sense a task that only ran daily (since each instance of the ETS would need a different
execution_delta). 
> However, a Daily task can wait for multiple hourly tasks, because it knows in advance
that it needs 24 ETS's with deltas == range(24).
> Concrete suggestion:
> - add a param ({{execution_delta_fn}}?) that takes in the current execution_date and
is expected to return the desired sense date (for example, it could always return midnight
of the previous day, no matter what the ETS was executed).
> cc [~criccomini]



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message