You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by "Sergei Iakhnin (JIRA)" <ji...@apache.org> on 2016/05/30 12:57:12 UTC

[jira] [Created] (AIRFLOW-193) Allow a series of tasks to be executed on the same worker

Sergei Iakhnin created AIRFLOW-193:
--------------------------------------

             Summary: Allow a series of tasks to be executed on the same worker
                 Key: AIRFLOW-193
                 URL: https://issues.apache.org/jira/browse/AIRFLOW-193
             Project: Apache Airflow
          Issue Type: New Feature
            Reporter: Sergei Iakhnin


Currently the only way to limit the execution of a series of tasks to a single worker is via pools, however this is not a very convenient method when managing hundreds of workers.

In the context of scientific workflows it is a common desire to be able to retrieve a (possibly large) sample from a data repository (or object store), then progressively elaborate it via a series of transformations, and finally deposit the result back. From a modelling perspective it makes sense to have the series of transformations each be encapsulated in a separate task. From practical considerations (performance, network bandwidth) it would be desirable to retrieve the sample to a single worker's local storage, where it would then be worked upon until completion. This, of course, requires the ability to have a slew of tasks to be bound to a particular worker. 



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