You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
152 lines
6.0 KiB
152 lines
6.0 KiB
3 years ago
|
.. Licensed to the Apache Software Foundation (ASF) under one
|
||
|
or more contributor license agreements. See the NOTICE file
|
||
|
distributed with this work for additional information
|
||
|
regarding copyright ownership. The ASF licenses this file
|
||
|
to you under the Apache License, Version 2.0 (the
|
||
|
"License"); you may not use this file except in compliance
|
||
|
with the License. You may obtain a copy of the License at
|
||
|
|
||
|
.. http://www.apache.org/licenses/LICENSE-2.0
|
||
|
|
||
|
.. Unless required by applicable law or agreed to in writing,
|
||
|
software distributed under the License is distributed on an
|
||
|
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
|
||
|
KIND, either express or implied. See the License for the
|
||
|
specific language governing permissions and limitations
|
||
|
under the License.
|
||
|
|
||
|
Concepts
|
||
|
========
|
||
|
|
||
|
In this section, you would know the core concepts of *PyDolphinScheduler*.
|
||
|
|
||
|
Process Definition
|
||
|
------------------
|
||
|
|
||
|
Process definition describe the whole things except `tasks`_ and `tasks dependence`_, which including
|
||
|
name, schedule interval, schedule start time and end time. You would know scheduler
|
||
|
|
||
|
Process definition could be initialized in normal assign statement or in context manger.
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
# Initialization with assign statement
|
||
|
pd = ProcessDefinition(name="my first process definition")
|
||
|
|
||
|
# Or context manger
|
||
|
with ProcessDefinition(name="my first process definition") as pd:
|
||
|
pd.submit()
|
||
|
|
||
|
Process definition is the main object communicate between *PyDolphinScheduler* and DolphinScheduler daemon.
|
||
|
After process definition and task is be declared, you could use `submit` and `run` notify server your definition.
|
||
|
|
||
|
If you just want to submit your definition and create workflow, without run it, you should use attribute `submit`.
|
||
|
But if you want to run the workflow after you submit it, you could use attribute `run`.
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
# Just submit definition, without run it
|
||
|
pd.submit()
|
||
|
|
||
|
# Both submit and run definition
|
||
|
pd.run()
|
||
|
|
||
|
Schedule
|
||
|
~~~~~~~~
|
||
|
|
||
|
We use parameter `schedule` determine the schedule interval of workflow, *PyDolphinScheduler* support seven
|
||
|
asterisks expression, and each of the meaning of position as below
|
||
|
|
||
|
.. code-block:: text
|
||
|
|
||
|
* * * * * * *
|
||
|
┬ ┬ ┬ ┬ ┬ ┬ ┬
|
||
|
│ │ │ │ │ │ │
|
||
|
│ │ │ │ │ │ └─── year
|
||
|
│ │ │ │ │ └───── day of week (0 - 7) (0 to 6 are Sunday to Saturday, or use names; 7 is Sunday, the same as 0)
|
||
|
│ │ │ │ └─────── month (1 - 12)
|
||
|
│ │ │ └───────── day of month (1 - 31)
|
||
|
│ │ └─────────── hour (0 - 23)
|
||
|
│ └───────────── min (0 - 59)
|
||
|
└─────────────── second (0 - 59)
|
||
|
|
||
|
Here we add some example crontab:
|
||
|
|
||
|
- `0 0 0 * * ? *`: Workflow execute every day at 00:00:00.
|
||
|
- `10 2 * * * ? *`: Workflow execute hourly day at ten pass two.
|
||
|
- `10,11 20 0 1,2 * ? *`: Workflow execute first and second day of month at 00:20:10 and 00:20:11.
|
||
|
|
||
|
Tenant
|
||
|
~~~~~~
|
||
|
|
||
|
Tenant is the user who run task command in machine or in virtual machine. it could be assign by simple string.
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
#
|
||
|
pd = ProcessDefinition(name="process definition tenant", tenant="tenant_exists")
|
||
|
|
||
|
.. note::
|
||
|
|
||
|
Make should tenant exists in target machine, otherwise it will raise an error when you try to run command
|
||
|
|
||
|
Tasks
|
||
|
-----
|
||
|
|
||
|
Task is the minimum unit running actual job, and it is nodes of DAG, aka directed acyclic graph. You could define
|
||
|
what you want to in the task. It have some required parameter to make uniqueness and definition.
|
||
|
|
||
|
Here we use :py:meth:`pydolphinscheduler.tasks.Shell` as example, parameter `name` and `command` is required and must be provider. Parameter
|
||
|
`name` set name to the task, and parameter `command` declare the command you wish to run in this task.
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
# We named this task as "shell", and just run command `echo shell task`
|
||
|
shell_task = Shell(name="shell", command="echo shell task")
|
||
|
|
||
|
If you want to see all type of tasks, you could see :doc:`tasks/index`.
|
||
|
|
||
|
Tasks Dependence
|
||
|
~~~~~~~~~~~~~~~~
|
||
|
|
||
|
You could define many tasks in on single `Process Definition`_. If all those task is in parallel processing,
|
||
|
then you could leave them alone without adding any additional information. But if there have some tasks should
|
||
|
not be run unless pre task in workflow have be done, we should set task dependence to them. Set tasks dependence
|
||
|
have two mainly way and both of them is easy. You could use bitwise operator `>>` and `<<`, or task attribute
|
||
|
`set_downstream` and `set_upstream` to do it.
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
# Set task1 as task2 upstream
|
||
|
task1 >> task2
|
||
|
# You could use attribute `set_downstream` too, is same as `task1 >> task2`
|
||
|
task1.set_downstream(task2)
|
||
|
|
||
|
# Set task1 as task2 downstream
|
||
|
task1 << task2
|
||
|
# It is same as attribute `set_upstream`
|
||
|
task1.set_upstream(task2)
|
||
|
|
||
|
# Beside, we could set dependence between task and sequence of tasks,
|
||
|
# we set `task1` is upstream to both `task2` and `task3`. It is useful
|
||
|
# for some tasks have same dependence.
|
||
|
task1 >> [task2, task3]
|
||
|
|
||
|
Task With Process Definition
|
||
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||
|
|
||
|
In most of data orchestration cases, you should assigned attribute `process_definition` to task instance to
|
||
|
decide workflow of task. You could set `process_definition` in both normal assign or in context manger mode
|
||
|
|
||
|
.. code-block:: python
|
||
|
|
||
|
# Normal assign, have to explicit declaration and pass `ProcessDefinition` instance to task
|
||
|
pd = ProcessDefinition(name="my first process definition")
|
||
|
shell_task = Shell(name="shell", command="echo shell task", process_definition=pd)
|
||
|
|
||
|
# Context manger, `ProcessDefinition` instance pd would implicit declaration to task
|
||
|
with ProcessDefinition(name="my first process definition") as pd:
|
||
|
shell_task = Shell(name="shell", command="echo shell task",
|
||
|
|
||
|
With both `Process Definition`_, `Tasks`_ and `Tasks Dependence`_, we could build a workflow with multiple tasks.
|