This plugins runs Jobs and Transformations through Carte servers.
It allows to orchestrate a massive number of trans/jobs taking care
of the dependencies between them, even between different instances.
This is done by using CarteJobOperator
and CarteTransOperator
It also runs Pan (transformations) and Kitchen (Jobs) in local mode,
both from repository and local XML files. For this approach, use
KitchenOperator
and PanOperator
The same setup process must be performed on webserver, scheduler and workers (that runs this tasks) to get it working. If you want to deploy specific workers to run this kind of tasks, see Queues, in Airflow Concepts section.
First of all, the package should be installed via pip install
command.
pip install airflow-pentaho-plugin
Then, a new connection needs to be added to Airflow Connections, to do this,
go to Airflow web UI, and click on Admin -> Connections
on the top menu.
Now, click on Create
tab.
Use HTTP connection type. Enter the Conn Id, this plugin uses pdi_default
by default, the username and the password for your Pentaho Repository.
At the bottom of the form, fill the Extra field with pentaho_home
, the
path where your pdi-ce is placed, and rep
, the repository name for this
connection, using a json formatted string like it follows.
{
"pentaho_home": "/opt/pentaho",
"rep": "Default"
}
In order to use CarteJobOperator
, the connection should be set different. Fill
host
(including http://
or https://
) and port
for Carte hostname and port,
username
and password
for PDI repository, and extra
as it follows.
{
"rep": "Default",
"carte_username": "cluster",
"carte_password": "cluster"
}
CarteJobOperator is responsible for running jobs in remote slave servers. Here
it is an example of CarteJobOperator
usage.
# For versions before 2.0
# from airflow.operators.airflow_pentaho import CarteJobOperator
from airflow_pentaho.operators.carte import CarteJobOperator
# ... #
# Define the task using the CarteJobOperator
avg_spent = CarteJobOperator(
conn_id='pdi_default',
task_id="average_spent",
job="/home/bi/average_spent",
params={"date": "{{ ds }}"}, # Date in yyyy-mm-dd format
dag=dag)
# ... #
some_task >> avg_spent >> another_task
Kitchen operator is responsible for running Jobs. Lets suppose that we have
a defined Job saved on /home/bi/average_spent
in our repository with
the argument date
as input parameter. Lets define the task using the
KitchenOperator
.
# For versions before 2.0
# from airflow.operators.airflow_pentaho import KitchenOperator
from airflow_pentaho.operators.kettle import KitchenOperator
# ... #
# Define the task using the KitchenOperator
avg_spent = KitchenOperator(
conn_id='pdi_default',
queue="pdi",
task_id="average_spent",
directory="/home/bi",
job="average_spent",
params={"date": "{{ ds }}"}, # Date in yyyy-mm-dd format
dag=dag)
# ... #
some_task >> avg_spent >> another_task
CarteTransOperator is responsible for running transformations in remote slave
servers. Here it is an example of CarteTransOperator
usage.
# For versions before 2.0
# from airflow.operators.airflow_pentaho import CarteTransOperator
from airflow_pentaho.operators.carte import CarteTransOperator
# ... #
# Define the task using the CarteJobOperator
enriche_customers = CarteTransOperator(
conn_id='pdi_default',
task_id="enrich_customer_data",
job="/home/bi/enrich_customer_data",
params={"date": "{{ ds }}"}, # Date in yyyy-mm-dd format
dag=dag)
# ... #
some_task >> enrich_customers >> another_task
Pan operator is responsible for running transformations. Lets suppose that
we have one saved on /home/bi/clean_somedata
. Lets define the task using the
PanOperator
. In this case, the transformation receives a parameter that
determines the file to be cleaned.
# For versions before 2.0
# from airflow.operators.airflow_pentaho import PanOperator
from airflow_pentaho.operators.kettle import PanOperator
# ... #
# Define the task using the PanOperator
clean_input = PanOperator(
conn_id='pdi_default',
queue="pdi",
task_id="cleanup",
directory="/home/bi",
trans="clean_somedata",
params={"file": "/tmp/input_data/{{ ds }}/sells.csv"},
dag=dag)
# ... #
some_task >> clean_input >> another_task
For more information, please see sample_dags/pdi_flow.py