How Jobs are handled via Talend Administration Center
Talend Jobs are scheduled from Talend Administration Center, which provides a browser management interface, and are parameterized with Context variables. But often you may prefer to have programmatic control of Jobs via an API. This post covers how to expose TalendJobs via the Talend Administration Center API. It provides sample Jobs, some useful browser utilities, and an example of wrapping the API in a RESTful service layer using Data Services.
The Job Conductor page of Talend Administration Center is easy to use and powerful: its allows you to schedule Jobs with simple, more complex Cron triggers, or file triggers. You also have the possibility to launch Jobs manually from the browser. With exception of the file trigger, the Job is still running at a pre-determined time or with explicit human intervention. Using file triggers as the mechanism for inter-process communication may require access privileges which are not allowed in a secure environment, which is why it may be preferable to invoke a Talend Job via a real API.
Jobs can also be parametrized with Context variables. Context variables can be overridden by system administrators on the Job Conductor page to provide additional flexibility. But when the Job is running, it always runs with the same set of pre-configured Context variables, whether they are the default values or the overridden values, they cannot be changed without human intervention. It is preferable to be able to pass parameters via an API.
One option is to build Jobs as self-contained .zip files. The generated archives will include launching scripts and all necessary .jar files. However, the resulting Jobs are running in isolation and lack the monitoring, management, and control parts provided by Talend Administration Center. No centralized logging is provided, and there is no concept of Job Servers or a Job grid. Instead, these responsibilities fall on the developer. As individual solutions proliferate, the management of the broader system becomes more difficult and the maintenance tail becomes more unwieldy.
So while exported Jobs provide flexibility, they sacrifice manageability. The Talend Administration Center API provides a very simple and powerful alternative.
Talend Administration Center MetaServlet API
The Talend Administration Center MetaServlet API is an RPC style HTTP API, (not restful) that it is very easy to use and can be easily wrapped with a RESTful interface if desired.
All MetaServlet operations are invoked via an HTTP Get request. All parameters to the operation are encoded as a single, unnamed base-64 encoded parameter to the get request.
The Talend Administration Center MetaServlet command-line tool is available in the following folder of the Talend Administration Center installation directory:
<tomcat_path>/webapps/org.talend.administrator/WEB-INF/classes/MetaServletCaller.bat on Windows
<tomcat_path>/webapps/org.talend.administrator/WEB-INF/classes/MetaServletCaller.sh on Linux
Running the MetaServletCaller with no arguments shows the top level help message:
usage: Missing required option: url
-f,--format-output format Json output
-h,--help print this help message
-json,--json-params <arg> Required params in a Json object
-url,--tac-url <arg> TAC's http url
-v,--verbose display more informations
In order to get the full detailed help message, the Talend Administration Center service must be up and running and you must pass the --tac-urlparameter.
--help all, -h all
Use --help all to get the full help, and the -h all for an abbreviated version. The examples below capture the output to a text file for subsequent reference.
--tac-url=http://localhost:8080/org.talend.administrator/ -help runTask
Description : Allows to run a task defined in Job conductor by its id.
Mode can be 'asynchronous' or 'synchronous'
Requires authentication : true
Since : 4.2
In order to run a task, you need to know its system generated taskId. This information can be retrieved by executing the getTaskIdByName command.
Gets the corresponding ID of the task by looking for its taskName.
-url=http://localhost:8080/org.talend.administrator/ -help getTaskIdByName
Description : Get task id by given taskName
Requires authentication : true
Since : 5.1
Invoking the Talend Administration Center API interactively
When developers work with the MetaServlet API, it can be useful for them to interactively invoke the Talend Administration Center API.
Go to the Job Conductor page of Talend Administration Center, select the arrow icon next to any of the columns and make sure the Id column is checked so that it will be displayed.
Once the HTTP Get message is sent, monitor the progress of the execution through the Execution History page of Talend Administration Center.
Invoking the Talend Administration Center API programmatically
Once the JSON arguments are base-64 encoded, they can be passed as the sole parameter to the HTTP Get request. If you are integrating with Talend, your application might be written in regular Java, or possibly some other language, which should not cause any issues since the HTTP Get and base-64 are interoperable standards.
To invoke the Talend Administration Center API, the JSON objects must be base-64 encoded.
If you happen to be integrating with a Java application, use the following Apache Commons Base64 class method:org.apache.commons.codec.binary.Base64.encodeBase64()
If you are using Talend, use the tLibraryLoad component to add the Apache Commons library.
You can retrieve the MetaServlet Job archive attached to the Downloads tab on the left panel of this page to invoke the Talend Administration Center API from the Talend Job.
It uses the encodeBase64() method within a tMap prior to the tRESTclient invocation of the Talend Administration Center API operations. Three operations are invoked, and each operation is invoked within its own SubJob. Each SubJob starts by initializing the request from the Context Parameters:
The first invocation looks up the taskId based on the human readable Job name.
The second invocation uses the taskId returned from the first invocation to trigger the Job.
The third invocation uses the returned execRequestId handle as the argument to the getTaskExecutionStatus operation to monitor the Job status.