AWS Data Pipeline configures and manages a data-driven workflow called a
pipeline.
AWS Data Pipeline
handles the details of scheduling and ensuring that data dependencies are met so
that your application
can focus on processing the data.
AWS Data Pipeline provides a JAR implementation of a task runner called AWS Data
Pipeline Task Runner.
AWS Data Pipeline Task Runner provides logic for common data management
scenarios, such as performing
database queries and running data analysis using Amazon Elastic MapReduce
(Amazon EMR). You can use
AWS Data Pipeline Task Runner as your task runner, or you can write your own
task runner to provide
custom data management.
AWS Data Pipeline implements two main sets of functionality. Use the first set
to create a pipeline
and define data sources, schedules, dependencies, and the transforms to be
performed on the data.
Use the second set in your task runner application to receive the next task
ready for processing.
The logic for performing the task, such as querying the data, running data
analysis, or converting
the data from one format to another, is contained within the task runner. The
task runner performs
the task assigned to it by the web service, reporting progress to the web
service as it does so.
When the task is done, the task runner reports the final success or failure of
the task to the web service.
The pipeline is set to the DEACTIVATING
state until the deactivation process completes.
To resume a deactivated pipeline, use ActivatePipeline. By default, the
pipeline resumes from the last completed execution.
Optionally, you can specify the date and time to resume the pipeline.
Deletes a pipeline, its pipeline definition, and its run history.
AWS Data Pipeline attempts to cancel instances associated with the pipeline that
are currently being processed by task runners.
Deleting a pipeline cannot be undone. You cannot query or restore a deleted
pipeline.
To temporarily pause a pipeline instead of deleting it, call SetStatus with
the status set to PAUSE on individual components.
Components that are paused by SetStatus can be resumed.
The information retrieved includes the name of the pipeline, the pipeline
identifier,
its current state, and the user account that owns the pipeline. Using account
credentials, you can retrieve metadata about pipelines
that you or your IAM users have created. If you are using an IAM user account,
you can retrieve metadata about only those pipelines
for which you have read permissions.
To retrieve the full pipeline definition instead of metadata about the pipeline,
call GetPipelineDefinition.
Task runners call EvaluateExpression to evaluate a string in the context of
the specified object.
For example, a task runner can evaluate SQL queries stored in Amazon S3.
POST / HTTP/1.1
Content-Type: application/x-amz-json-1.1
X-Amz-Target: DataPipeline.DescribePipelines
Content-Length: 164
Host: datapipeline.us-east-1.amazonaws.com
X-Amz-Date: Mon, 12 Nov 2012 17:49:52 GMT
Authorization: AuthParams
{"pipelineId": "df-08785951KAKJEXAMPLE",
"objectId": "Schedule",
"expression": "Transform started at startDateTime and finished at
endDateTime"}
x-amzn-RequestId: 02870eb7-0736-11e2-af6f-6bc7a6be60d9
Content-Type: application/x-amz-json-1.1
Content-Length: 103
Date: Mon, 12 Nov 2012 17:50:53 GMT
{"evaluatedExpression": "Transform started at 2012-12-12T00:00:00 and finished
at 2012-12-21T18:00:00"}
Task runners call PollForTask to receive a task to perform from AWS Data
Pipeline.
The task runner specifies which tasks it can perform
by setting a value for the workerGroup parameter. The task returned can come
from any of the pipelines that
match the workerGroup value passed in by the task runner and that was launched
using the IAM user credentials
specified by the task runner.
If tasks are ready in the work queue, PollForTask returns a response
immediately. If no tasks are available in the queue,
PollForTask uses long-polling and holds on to a poll connection for up to a 90
seconds, during which time the first newly
scheduled task is handed to the task runner. To accomodate this, set the socket
timeout in your task runner to 90 seconds. The task
runner should not call PollForTask again on the same workerGroup until it
receives a response, and this can take up to 90 seconds.
Adds tasks, schedules, and preconditions to the specified pipeline.
You can use PutPipelineDefinition to populate a new pipeline.
PutPipelineDefinition also validates the configuration as it adds it to the
pipeline. Changes to the pipeline are saved unless one
of the following three validation errors exists in the pipeline.
An object is missing a name or identifier field.
A string or reference field is empty.
The number of objects in the pipeline exceeds the maximum allowed
objects.
The pipeline is in a FINISHED state.
Pipeline object definitions are passed to the PutPipelineDefinition action and
returned by the GetPipelineDefinition action.
Example 1This example sets an valid pipeline configuration and returns success.
POST / HTTP/1.1
Content-Type: application/x-amz-json-1.1
X-Amz-Target: DataPipeline.PutPipelineDefinition
Content-Length: 914
Host: datapipeline.us-east-1.amazonaws.com
X-Amz-Date: Mon, 12 Nov 2012 17:49:52 GMT
Authorization: AuthParams
{"pipelineId": "df-0937003356ZJEXAMPLE",
"pipelineObjects":
[
{"id": "Default",
"name": "Default",
"fields":
[
{"key": "workerGroup",
"stringValue": "workerGroup"}
]
},
{"id": "Schedule",
"name": "Schedule",
"fields":
[
{"key": "startDateTime",
"stringValue": "2012-12-12T00:00:00"},
{"key": "type",
"stringValue": "Schedule"},
{"key": "period",
"stringValue": "1 hour"},
{"key": "endDateTime",
"stringValue": "2012-12-21T18:00:00"}
]
},
{"id": "SayHello",
"name": "SayHello",
"fields":
[
{"key": "type",
"stringValue": "ShellCommandActivity"},
{"key": "command",
"stringValue": "echo hello"},
{"key": "parent",
"refValue": "Default"},
{"key": "schedule",
"refValue": "Schedule"}
]
}
]
}
HTTP/1.1 200
x-amzn-RequestId: f74afc14-0754-11e2-af6f-6bc7a6be60d9
Content-Type: application/x-amz-json-1.1
Content-Length: 18
Date: Mon, 12 Nov 2012 17:50:53 GMT
{"errored": false}
Example 2This example sets an invalid pipeline configuration (the value for `workerGroup`
is an empty string) and returns an error message.
POST / HTTP/1.1
Content-Type: application/x-amz-json-1.1
X-Amz-Target: DataPipeline.PutPipelineDefinition
Content-Length: 903
Host: datapipeline.us-east-1.amazonaws.com
X-Amz-Date: Mon, 12 Nov 2012 17:49:52 GMT
Authorization: AuthParams
{"pipelineId": "df-06372391ZG65EXAMPLE",
"pipelineObjects":
[
{"id": "Default",
"name": "Default",
"fields":
[
{"key": "workerGroup",
"stringValue": ""}
]
},
{"id": "Schedule",
"name": "Schedule",
"fields":
[
{"key": "startDateTime",
"stringValue": "2012-09-25T17:00:00"},
{"key": "type",
"stringValue": "Schedule"},
{"key": "period",
"stringValue": "1 hour"},
{"key": "endDateTime",
"stringValue": "2012-09-25T18:00:00"}
]
},
{"id": "SayHello",
"name": "SayHello",
"fields":
[
{"key": "type",
"stringValue": "ShellCommandActivity"},
{"key": "command",
"stringValue": "echo hello"},
{"key": "parent",
"refValue": "Default"},
{"key": "schedule",
"refValue": "Schedule"}
]
}
]
}
HTTP/1.1 200
x-amzn-RequestId: f74afc14-0754-11e2-af6f-6bc7a6be60d9
Content-Type: application/x-amz-json-1.1
Content-Length: 18
Date: Mon, 12 Nov 2012 17:50:53 GMT
{"__type": "com.amazon.setl.webservice#InvalidRequestException",
"message": "Pipeline definition has errors: Could not save the pipeline
definition due to FATAL errors:
[com.amazon.setl.webservice.ValidationError@108d7ea9] Please call Validate to
validate your pipeline"}
Task runners call ReportTaskProgress when assigned a task to acknowledge that
it has the task.
If the web service does not
receive this acknowledgement within 2 minutes, it assigns the task in a
subsequent PollForTask call. After this initial acknowledgement,
the task runner only needs to report progress every 15 minutes to maintain its
ownership of the task. You can change this reporting time
from 15 minutes by specifying a reportProgressTimeout field in your pipeline.
If a task runner does not report its status after 5 minutes, AWS Data Pipeline
assumes that the task runner is unable to process the task
and reassigns the task in a subsequent response to PollForTask. Task runners
should call ReportTaskProgress every 60 seconds.
Task runners call ReportTaskRunnerHeartbeat every 15 minutes to indicate that
they are operational.
If the AWS Data Pipeline Task Runner is launched on a resource managed by AWS
Data Pipeline, the web service can use
this call to detect when the task runner application has failed and restart a
new instance.
Requests that the status of the specified physical or logical pipeline objects
be updated in the specified pipeline.
This update might not occur immediately, but is eventually consistent. The
status that can be set depends on the type of object (for example, DataNode or
Activity).
You cannot perform this operation on FINISHED pipelines and attempting to do
so returns InvalidRequestException.
Task runners call SetTaskStatus to notify AWS Data Pipeline that a task is
completed and provide information about the final status.
A task runner makes this call regardless of whether the task was sucessful. A
task runner does not need to call SetTaskStatus for
tasks that are canceled by the web service during a call to
ReportTaskProgress.