Python OozieServerProvider.wait_until_completion方法代码示例

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def setUp(self):
    """
    To clean: creating test1, test2, test3...users
    """
    TestJobBrowserWithHadoop.user_count += 1
    self.username = 'test' + str(TestJobBrowserWithHadoop.user_count)
    self.home_dir = '/user/%s' % self.username
    self.cluster.fs.do_as_user(self.username, self.cluster.fs.create_home_dir, self.home_dir)

    self.client = make_logged_in_client(username=self.username, is_superuser=False, groupname='test')
    self.user = User.objects.get(username=self.username)
    grant_access(self.username, 'test', 'jobsub')
    grant_access(self.username, 'test', 'jobbrowser')
    grant_access(self.username, 'test', 'oozie')
    add_to_group(self.username)

    self.prev_user = self.cluster.fs.user
    self.cluster.fs.setuser(self.username)

    self.install_examples()
    self.design = self.create_design()

    # Run the sleep example, since it doesn't require user home directory
    design_id = self.design.id
    response = self.client.post(reverse('oozie:submit_workflow',
                                args=[design_id]),
                                data={u'form-MAX_NUM_FORMS': [u''],
                                      u'form-INITIAL_FORMS': [u'1'],
                                      u'form-0-name': [u'REDUCER_SLEEP_TIME'],
                                      u'form-0-value': [u'1'],
                                      u'form-TOTAL_FORMS': [u'1']},
                                follow=True)
    oozie_jobid = response.context['oozie_workflow'].id
    OozieServerProvider.wait_until_completion(oozie_jobid, timeout=120, step=1)

    self.hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
    self.hadoop_job_id_short = views.get_shorter_id(self.hadoop_job_id)

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def test_job(self):
    """
    Test new job views.

    The status of the jobs should be the same as the status reported back by oozie.
    In this case, all jobs should succeed.
    """
    # Run the sleep example, since it doesn't require user home directory
    design_id = self.design.id
    response = self.client.post(reverse('oozie:submit_workflow',
                                args=[design_id]),
                                data={u'form-MAX_NUM_FORMS': [u''],
                                      u'form-INITIAL_FORMS': [u'1'],
                                      u'form-0-name': [u'REDUCER_SLEEP_TIME'],
                                      u'form-0-value': [u'1'],
                                      u'form-TOTAL_FORMS': [u'1']},
                                follow=True)
    oozie_jobid = response.context['oozie_workflow'].id
    OozieServerProvider.wait_until_completion(oozie_jobid, timeout=120, step=1)
    hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
    hadoop_job_id_short = views.get_shorter_id(hadoop_job_id)

    # All jobs page and fetch job ID
    # Taking advantage of the fact new jobs are at the top of the list!
    response = self.client.get('/jobbrowser/jobs/')
    assert_true(hadoop_job_id_short in response.content, response.content)

    # Make sure job succeeded
    response = self.client.get('/jobbrowser/jobs/?state=completed')
    assert_true(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=failed')
    assert_false(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=running')
    assert_false(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=killed')
    assert_false(hadoop_job_id_short in response.content)

    # Check sharing permissions
    # Login as ourself
    finish = SHARE_JOBS.set_for_testing(True)
    try:
      response = self.client.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    finish = SHARE_JOBS.set_for_testing(False)
    try:
      response = self.client.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    # Login as someone else
    client_not_me = make_logged_in_client(username='not_me', is_superuser=False, groupname='test')
    grant_access("not_me", "test", "jobbrowser")

    finish = SHARE_JOBS.set_for_testing(True)
    try:
      response = client_not_me.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    finish = SHARE_JOBS.set_for_testing(False)
    try:
      response = client_not_me.get('/jobbrowser/jobs/?user=')
      assert_false(hadoop_job_id_short in response.content)
    finally:
      finish()

    # Single job page
    response = self.client.get('/jobbrowser/jobs/%s' % hadoop_job_id)

    # Check some counters for single job.
    counters = response.context['job'].counters
    counters_file_bytes_written = counters['org.apache.hadoop.mapreduce.FileSystemCounter']['counters']['FILE_BYTES_WRITTEN']
    assert_true(counters_file_bytes_written['map'] > 0)
    assert_true(counters_file_bytes_written['reduce'] > 0)

    # We can't just check the complete contents of the python map because the
    # SLOTS_MILLIS_* entries have a variable number of milliseconds from
    # run-to-run.
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_MAPS']['total'], 2L)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_REDUCES']['total'], 1L)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0L)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0L)
    assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
    assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)

    # There should be 4 tasks for this job: cleanup, setup, map, reduce
    response = self.client.get('/jobbrowser/jobs/%s/tasks' % (hadoop_job_id,))
    assert_true(len(response.context['page'].object_list), 4)
    # Select by tasktype
    response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id,))
    assert_true(len(response.context['page'].object_list), 1)
    # Select by taskstate
    response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
    assert_true(len(response.context['page'].object_list), 4)
    # Select by text
#.........这里部分代码省略.........

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def test_failed_jobs(self):
    """
    Test jobs with genuine failure, not just killed
    """
    # Create design that will fail because the script file isn't there
    INPUT_DIR = self.home_dir + '/input'
    OUTPUT_DIR = self.home_dir + '/output'
    try:
        self.cluster.fs.mkdir(self.home_dir + "/jt-test_failed_jobs")
        self.cluster.fs.mkdir(INPUT_DIR)
        self.cluster.fs.rmtree(OUTPUT_DIR)
    except:
        # rmtree probably failed here.
        pass
    response = self.client.post(reverse('jobsub.views.new_design', kwargs={'node_type': 'mapreduce'}), {
        'name': ['test_failed_jobs-1'],
        'description': ['description test_failed_jobs-1'],
        'args': '',
        'jar_path': '/user/hue/oozie/workspaces/lib/hadoop-examples.jar',
        'prepares': '[]',
        'archives': '[]',
        'files': '[]',
        'job_properties': ['[{"name":"mapred.input.dir","value":"%s"},\
            {"name":"mapred.output.dir","value":"%s"},\
            {"name":"mapred.mapper.class","value":"org.apache.hadoop.mapred.lib.dne"},\
            {"name":"mapred.combiner.class","value":"org.apache.hadoop.mapred.lib.dne"},\
            {"name":"mapred.reducer.class","value":"org.apache.hadoop.mapred.lib.dne"}]' % (INPUT_DIR, OUTPUT_DIR)]
        }, HTTP_X_REQUESTED_WITH='XMLHttpRequest', follow=True)

    # Submit the job
    design_dict = json.loads(response.content)
    design_id = int(design_dict['id'])
    response = self.client.post(reverse('oozie:submit_workflow',
                                args=[design_id]),
                                data={u'form-MAX_NUM_FORMS': [u''],
                                      u'form-INITIAL_FORMS': [u'1'],
                                      u'form-0-name': [u'REDUCER_SLEEP_TIME'],
                                      u'form-0-value': [u'1'],
                                      u'form-TOTAL_FORMS': [u'1']},
                                follow=True)
    oozie_jobid = response.context['oozie_workflow'].id
    job = OozieServerProvider.wait_until_completion(oozie_jobid, timeout=120, step=1)
    hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
    hadoop_job_id_short = views.get_shorter_id(hadoop_job_id)

    # Select only killed jobs (should be absent)
    # Taking advantage of the fact new jobs are at the top of the list!
    response = self.client.get('/jobbrowser/jobs/?state=killed')
    assert_false(hadoop_job_id_short in response.content)

    # Select only failed jobs (should be present)
    # Map job should succeed. Reduce job should fail.
    response = self.client.get('/jobbrowser/jobs/?state=failed')
    assert_true(hadoop_job_id_short in response.content)

    # The single job view should have the failed task table
    response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id,))
    html = response.content.lower()
    assert_true('failed task' in html)

    # The map task should say success (empty input)
    map_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
    response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, map_task_id))
    assert_true('succeed' in response.content)
    assert_true('failed' not in response.content)

    # The reduce task should say failed
    reduce_task_id = hadoop_job_id.replace('job', 'task') + '_r_000000'
    response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, reduce_task_id))
    assert_true('succeed' not in response.content)
    assert_true('failed' in response.content)

    # Selecting by failed state should include the failed map
    response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' % (hadoop_job_id,))
    assert_true('r_000000' in response.content)
    assert_true('m_000000' not in response.content)

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def test_job(self):
    """
    Test new job views.

    The status of the jobs should be the same as the status reported back by oozie.
    In this case, all jobs should succeed.
    """
    # Clone design
    assert_equal(0, OozieDesign.objects.filter(owner__username=self.username).count())
    self.client.post('/jobsub/clone_design/%d' % self.sleep_design_id)
    assert_equal(1, OozieDesign.objects.filter(owner__username=self.username).count())

    # Run the sleep example, since it doesn't require user home directory
    design_id = OozieDesign.objects.get(owner__username=self.username).id
    response = self.client.post("/jobsub/submit_design/%d" % (design_id,),
      dict(map_sleep_time=1,
           num_maps=1,
           num_reduces=1,
           reduce_sleep_time=1),
      follow=True)
    oozie_jobid = response.context['jobid']
    job = OozieServerProvider.wait_until_completion(oozie_jobid, timeout=120, step=1)
    hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
    hadoop_job_id_short = views.get_shorter_id(hadoop_job_id)

    # All jobs page and fetch job ID
    # Taking advantage of the fact new jobs are at the top of the list!
    response = self.client.get('/jobbrowser/jobs/')
    assert_true(hadoop_job_id_short in response.content)

    # Make sure job succeeded
    response = self.client.get('/jobbrowser/jobs/?state=completed')
    assert_true(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=failed')
    assert_false(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=running')
    assert_false(hadoop_job_id_short in response.content)
    response = self.client.get('/jobbrowser/jobs/?state=killed')
    assert_false(hadoop_job_id_short in response.content)

    # Check sharing permissions
    # Login as ourself
    finish = SHARE_JOBS.set_for_testing(True)
    try:
      response = self.client.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    finish = SHARE_JOBS.set_for_testing(False)
    try:
      response = self.client.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    # Login as someone else
    client_not_me = make_logged_in_client(username='not_me', is_superuser=False, groupname='test')
    grant_access("not_me", "test", "jobbrowser")

    finish = SHARE_JOBS.set_for_testing(True)
    try:
      response = client_not_me.get('/jobbrowser/jobs/?user=')
      assert_true(hadoop_job_id_short in response.content)
    finally:
      finish()

    finish = SHARE_JOBS.set_for_testing(False)
    try:
      response = client_not_me.get('/jobbrowser/jobs/?user=')
      assert_false(hadoop_job_id_short in response.content)
    finally:
      finish()

    # Single job page
    response = self.client.get('/jobbrowser/jobs/%s' % hadoop_job_id)

    # Check some counters for single job.
    counters = response.context['job'].counters
    counters_file_bytes_written = counters['org.apache.hadoop.mapreduce.FileSystemCounter']['counters']['FILE_BYTES_WRITTEN']
    assert_true(counters_file_bytes_written['map'] > 0)
    assert_true(counters_file_bytes_written['reduce'] > 0)

    # We can't just check the complete contents of the python map because the
    # SLOTS_MILLIS_* entries have a variable number of milliseconds from
    # run-to-run.
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_MAPS']['total'], 1)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_REDUCES']['total'], 1)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0)
    assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0)
    assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
    assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)

    # There should be 4 tasks for this job: cleanup, setup, map, reduce
    response = self.client.get('/jobbrowser/jobs/%s/tasks' % (hadoop_job_id,))
    assert_true(len(response.context['page'].object_list), 4)
    # Select by tasktype
    response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id,))
    assert_true(len(response.context['page'].object_list), 1)
    # Select by taskstate
#.........这里部分代码省略.........

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def test_failed_jobs(self):
    """
    Test jobs with genuine failure, not just killed
    """
    # Create design that will fail because the script file isn't there
    INPUT_DIR = self.home_dir + '/input'
    OUTPUT_DIR = self.home_dir + '/output'
    try:
        self.cluster.fs.mkdir(self.home_dir + "/jt-test_failed_jobs")
        self.cluster.fs.mkdir(INPUT_DIR)
        self.cluster.fs.rmtree(OUTPUT_DIR)
    except:
        # rmtree probably failed here.
        pass
    response = self.client.post('/jobsub/new_design/mapreduce', {
        'wf-name': ['test_failed_jobs-1'],
        'wf-description': ['description test_failed_jobs-1'],
        'action-args': [''],
        'action-jar_path': ['/user/hue/jobsub/examples/hadoop-examples.jar'],
        'action-archives': ['[]'],
        'action-job_properties': ['[{"name":"mapred.input.dir","value":"%s"},\
            {"name":"mapred.output.dir","value":"%s"},\
            {"name":"mapred.mapper.class","value":"org.apache.hadoop.mapred.lib.dne"},\
            {"name":"mapred.combiner.class","value":"org.apache.hadoop.mapred.lib.dne"},\
            {"name":"mapred.reducer.class","value":"org.apache.hadoop.mapred.lib.dne"}]' % (INPUT_DIR, OUTPUT_DIR)],
        'action-files': ['[]']}, follow=True)
    designs = json.loads(response.context['designs'])

    # Submit the job
    design_id = designs[0]['id']
    response = self.client.post("/jobsub/submit_design/%d" % design_id, follow=True)
    oozie_jobid = response.context['jobid']
    OozieServerProvider.wait_until_completion(oozie_jobid, timeout=500, step=1)
    hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
    hadoop_job_id_short = views.get_shorter_id(hadoop_job_id)

    # Select only killed jobs (should be absent)
    # Taking advantage of the fact new jobs are at the top of the list!
    response = self.client.get('/jobbrowser/jobs/?state=killed')
    assert_false(hadoop_job_id_short in response.content)

    # Select only failed jobs (should be present)
    # Map job should succeed. Reduce job should fail.
    response = self.client.get('/jobbrowser/jobs/?state=failed')
    assert_true(hadoop_job_id_short in response.content)

    # The single job view should have the failed task table
    response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id,))
    html = response.content.lower()
    assert_true('failed task' in html)

    # The map task should say success (empty input)
    map_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
    response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, map_task_id))
    assert_true('succeed' in response.content)
    assert_true('failed' not in response.content)

    # The reduce task should say failed
    reduce_task_id = hadoop_job_id.replace('job', 'task') + '_r_000000'
    response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, reduce_task_id))
    assert_true('succeed' not in response.content)
    assert_true('failed' in response.content)

    # Selecting by failed state should include the failed map
    response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' % (hadoop_job_id,))
    assert_true('r_000000' in response.content)
    assert_true('m_000000' not in response.content)

# 需要导入模块: from liboozie.oozie_api_test import OozieServerProvider [as 别名]
# 或者: from liboozie.oozie_api_test.OozieServerProvider import wait_until_completion [as 别名]
  def test_jobsub_setup_and_run_samples(self):
    """
    Merely exercises jobsub_setup, and then runs the sleep example.
    """
    if not jobsub_setup.Command().has_been_setup():
      jobsub_setup.Command().handle()
    self.cluster.fs.setuser('jobsub_test')

    assert_equal(3, OozieDesign.objects.filter(owner__username='sample').count())
    assert_equal(2, OozieMapreduceAction.objects.filter(ooziedesign__owner__username='sample').count())
    assert_equal(1, OozieStreamingAction.objects.filter(ooziedesign__owner__username='sample').count())

    # Make sure sample user got created.
    assert_equal(1, User.objects.filter(username='sample').count())

    # Clone design
    assert_equal(0, OozieDesign.objects.filter(owner__username='jobsub_test').count())
    jobid = OozieDesign.objects.get(name='sleep_job', owner__username='sample').id

    self.client.post('/jobsub/clone_design/%d' % jobid)
    assert_equal(1, OozieDesign.objects.filter(owner__username='jobsub_test').count())
    jobid = OozieDesign.objects.get(owner__username='jobsub_test').id

    # And now submit and run the sleep sample
    response = self.client.post('/jobsub/submit_design/%d' % jobid, {
        'num_reduces': 1,
        'num_maps': 1,
        'map_sleep_time': 1,
        'reduce_sleep_time': 1}, follow=True)

    assert_true(sum([status in response.content for status in ('PREP', 'OK', 'DONE')]) > 0)
    assert_true(str(jobid) in response.content)

    oozie_job_id = response.context['jobid']
    job = OozieServerProvider.wait_until_completion(oozie_job_id, timeout=120, step=1)
    logs = OozieServerProvider.oozie.get_job_log(oozie_job_id)

    assert_equal('SUCCEEDED', job.status, logs)


    # Grep
    n = OozieDesign.objects.filter(owner__username='jobsub_test').count()
    jobid = OozieDesign.objects.get(name='grep_example').id

    self.client.post('/jobsub/clone_design/%d' % jobid)
    assert_equal(n + 1, OozieDesign.objects.filter(owner__username='jobsub_test').count())
    jobid = OozieDesign.objects.get(owner__username='jobsub_test', name__contains='sleep_job').id

    # And now submit and run the sleep sample
    response = self.client.post('/jobsub/submit_design/%d' % jobid, {
        'num_reduces': 1,
        'num_maps': 1,
        'map_sleep_time': 1,
        'reduce_sleep_time': 1}, follow=True)

    assert_true(sum([status in response.content for status in ('PREP', 'OK', 'DONE')]) > 0)
    assert_true(str(jobid) in response.content)

    oozie_job_id = response.context['jobid']
    job = OozieServerProvider.wait_until_completion(oozie_job_id, timeout=60, step=1)
    logs = OozieServerProvider.oozie.get_job_log(oozie_job_id)

    assert_equal('SUCCEEDED', job.status, logs)