Oozie

A small PHP client for Apache Oozie over its v2 REST API.

Quick start

use Erikwang2013\BigData\BigData;

$oozie = BigData::driver('oozie', [
    'baseUrl' => 'http://127.0.0.1:11000/oozie',
    'username' => 'oozie',
    'password' => 'secret',
]);

$jobId = $oozie->submitWorkflow(<<<XML
<configuration>
  <property>
    <name>oozie.wf.application.path</name>
    <value>hdfs:///apps/demo/workflow.xml</value>
  </property>
</configuration>
XML);

$oozie->start($jobId);
echo $oozie->status($jobId), PHP_EOL;

baseUrl is the Oozie context root, including the /oozie context path.

Direct instantiation:

use Erikwang2013\BigData\Driver\Oozie\OozieClient;
use Erikwang2013\BigData\Driver\Oozie\OozieConfig;

$oozie = new OozieClient(OozieConfig::fromArray([
    'baseUrl' => 'http://127.0.0.1:11000/oozie',
]));
// Optional HTTP client and PSR-3 logger: new OozieClient($config, $http, $logger)

Configuration

OozieConfig::fromArray() accepts:

Submitting jobs

The body is the Oozie job configuration document (<configuration> with <property> entries), not the workflow/coordinator/bundle application definition. The application definition is referenced from the configuration; for example a workflow job sets oozie.wf.application.path to the HDFS location of the workflow.xml:

$oozie->submitWorkflow($configurationXml);     // oozie.wf.application.path
$oozie->submitCoordinator($configurationXml);  // oozie.coord.application.path
$oozie->submitBundle($configurationXml);       // oozie.bundle.application.path

All three POST to /v2/jobs with Content-Type: application/xml and return the job id from the JSON response ({"id": "..."}). If a proxy or older Oozie version answers with the bare job id as plain text, the trimmed body is returned instead.

Inspecting jobs

$oozie->job($jobId);    // GET /v2/job/{id}?show=info    (decoded JSON array)
$oozie->status($jobId); // GET /v2/job/{id}?show=status  (e.g. "RUNNING")
$oozie->log($jobId);    // GET /v2/job/{id}?show=log     (plain text)

status() parses {"status": "..."}; if the response is not JSON it falls back to the trimmed plain-text body.

Job actions

$oozie->start($jobId);
$oozie->suspend($jobId);
$oozie->resume($jobId);
$oozie->kill($jobId);

Each sends PUT /v2/job/{id}?action=start|suspend|resume|kill.

rerun() sends a <configuration> XML body built from the properties array:

$oozie->rerun($jobId, [
    'oozie.wf.rerun.failnodes' => 'true',
]);

Keys and values are XML-escaped.

Logging

OozieClient accepts an optional PSR-3 logger as its third constructor argument and defaults to Psr\Log\NullLogger. Debug records contain the HTTP method and URI only — credentials and job XML are never logged.

Errors

Transport failures throw Erikwang2013\BigData\Exception\ConnectionException. Any non-2xx response (including 3xx such as 304), an Oozie error body, or invalid JSON raises Erikwang2013\BigData\Exception\RequestException, which exposes public $statusCode and $body. Both exceptions extend Erikwang2013\BigData\Exception\BigDataException.

Notes

• Uses the Oozie v2 REST API only.
• No automatic retries are performed.