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
self.cluster.fs.mkdir("/user/test/jt-test_failed_jobs")
response = self.client.post("/jobsub/new/streaming", dict(
hadoop_properties='',
inputformat_class_0='org.apache.hadoop.mapred.TextInputFormat',
input='/user/test/jt-test_failed_jobs',
mapper_cmd="python hogalabogus.py",
name="bogus stream",
num_reduce_tasks="1",
outputformat_class_0='org.apache.hadoop.mapred.TextOutputFormat',
output='/user/test/out_stream.tmp',
reducer_cmd='python hogalabogus.py',
save_submit='off',
submit='Save',
))
# Submit the job
design_id = response.context["saved"]
response = self.client.post("/jobsub/submit/%d" % design_id)
watch_id = parse_out_id(response)
response = watch_till_complete(self.client, watch_id, timeout_sec=120)
job_id = Submission.objects.get(id=watch_id).submission_handle.id
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
# 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)
# Select only failed jobs (should be present)
response = self.client.get('/jobbrowser/jobs/?state=failed')
assert_true(hadoop_job_id in response.content)
# Select only killed jobs (should be absent)
response = self.client.get('/jobbrowser/jobs/?state=killed')
assert_true(hadoop_job_id not in response.content)
# 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)
def test_new_jobs(self):
"""
Submit jobs. Let them succeed or fail and view them.
"""
# Install examples
import jobsub.management.commands.jobsub_setup as jobsub_setup
if not jobsub_setup.Command().has_been_setup():
jobsub_setup.Command().handle()
# Run the sleep example, since it doesn't require user home directory
design_id = JobDesign.objects.get(name__contains="Example: Sleep").id
response = self.client.post("/jobsub/submit/%d" % (design_id,), dict(
map_sleep_time_millis=1,
num_mappers=1,
num_reducers=1,
reduce_sleep_time_millis=1))
job_id = parse_out_id(response)
response = watch_till_complete(self.client, job_id)
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
# All jobs page
response = self.client.get('/jobbrowser/jobs/')
assert_true(hadoop_job_id.lstrip('job_') in response.content)
# Single job page
response = self.client.get('/jobbrowser/jobs/%s' % hadoop_job_id)
# Check some counters for single job.
counters = response.context['job'].counters
# These counters may be a bit dependent on current sleep task impl, but should be stable
# unless we change SleepJob
assert_equal(
response.context['job'].counters['FileSystemCounters']['counters']['FILE_BYTES_WRITTEN'],
{'map': 44L, 'reduce': 12L, 'displayName': 'FILE_BYTES_WRITTEN', 'name': 'FILE_BYTES_WRITTEN'}
)
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id,))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Select by taskstate
response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
assert_true('Showing 1 to 4 of 4 tasks' in response.content)
# Select by text
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktext=clean' % (hadoop_job_id,))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Run another sleep job but kill it
response = self.client.post("/jobsub/submit/%d" % (design_id,), dict(
map_sleep_time_millis=1,
num_mappers=2000,
num_reducers=2000,
reduce_sleep_time_millis=1))
job_id = parse_out_id(response)
time.sleep(15) # 15 seconds should be enough to start the job
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
self.client.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id,))
# It should say killed
response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id,))
html = response.content.lower()
assert_true(hadoop_job_id in html)
assert_true('killed' in html)
# Exercise select by taskstate
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=pending' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=running' % (hadoop_job_id,))
def test_new_jobs(self):
"""
Submit jobs. Let them succeed or fail and view them.
"""
# Install examples
import jobsub.management.commands.jobsub_setup as jobsub_setup
if not jobsub_setup.Command().has_been_setup():
jobsub_setup.Command().handle()
# Run the sleep example, since it doesn't require user home directory
design_id = JobDesign.objects.get(name__contains="Example: Sleep").id
response = self.client.post(
"/jobsub/submit/%d" % (design_id, ),
dict(map_sleep_time_millis=1,
num_mappers=1,
num_reducers=1,
reduce_sleep_time_millis=1))
watch_id = parse_out_id(response)
response = watch_till_complete(self.client, watch_id)
job_id = Submission.objects.get(id=watch_id).submission_handle.id
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
# All jobs page
response = self.client.get('/jobbrowser/jobs/')
assert_true(hadoop_job_id.lstrip('job_') in response.content)
# 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['FileSystemCounters'][
'counters']['FILE_BYTES_WRITTEN']
assert_true(counters_file_bytes_written['map'] > 0)
assert_true(counters_file_bytes_written['reduce'] > 0)
assert_equal(counters_file_bytes_written['displayName'],
'FILE_BYTES_WRITTEN')
assert_equal(counters_file_bytes_written['displayName'],
'FILE_BYTES_WRITTEN')
# 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.mapred.JobInProgress$Counter']
['counters']['TOTAL_LAUNCHED_MAPS']['total'], 1)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapred.JobInProgress$Counter']
['counters']['TOTAL_LAUNCHED_REDUCES']['total'], 1)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapred.JobInProgress$Counter']
['counters']['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapred.JobInProgress$Counter']
['counters']['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0)
assert_true(response.context['job'].
counters['org.apache.hadoop.mapred.JobInProgress$Counter']
['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
assert_true(response.context['job'].
counters['org.apache.hadoop.mapred.JobInProgress$Counter']
['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)
# Check conf keys made it
assert_equal(response.context['job'].conf_keys['mapredReducerClass'],
'org.apache.hadoop.examples.SleepJob')
# 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('Showing 1 to 4 of 4 tasks' in response.content)
# Select by tasktype
response = self.client.get(
'/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id, ))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Select by taskstate
response = self.client.get(
'/jobbrowser/jobs/%s/tasks?taskstate=succeeded' %
(hadoop_job_id, ))
assert_true('Showing 1 to 4 of 4 tasks' in response.content)
# Select by text
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktext=clean' %
(hadoop_job_id, ))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Run another sleep job but kill it
response = self.client.post(
"/jobsub/submit/%d" % (design_id, ),
dict(map_sleep_time_millis=1,
num_mappers=2000,
num_reducers=2000,
reduce_sleep_time_millis=1))
job_id = parse_out_id(response)
time.sleep(15) # 15 seconds should be enough to start the job
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
client2 = make_logged_in_client('test_non_superuser',
is_superuser=False)
response = client2.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id, ))
assert_equal(
"Permission denied. User test_non_superuser cannot delete user test's job.",
response.context["error"])
self.client.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id, ))
# It should say killed
response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id, ))
html = response.content.lower()
assert_true(hadoop_job_id in html)
assert_true('killed' in html)
# Exercise select by taskstate
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=pending' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=running' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=killed' %
(hadoop_job_id, ))
# Test single task page
late_task_id = hadoop_job_id.replace('job', 'task') + '_r_001999'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' %
(hadoop_job_id, late_task_id))
assert_false('succeed' in response.content)
assert_true('killed' in response.content)
# The first task should've succeeded
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' %
(hadoop_job_id, early_task_id))
assert_true('succeed' in response.content)
assert_false('failed' in response.content)
# Test single attempt page
attempt_id = early_task_id.replace('task', 'attempt') + '_0'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s/attempts/%s' %
(hadoop_job_id, early_task_id, attempt_id))
assert_true('syslog' in response.content)
# Test dock jobs
response = self.client.get('/jobbrowser/dock_jobs/')
assert_true('completed' in response.content)
# TODO(atm): I'm pretty sure the following test only passes because of
# failed jobs which are run in test_failed_jobs
assert_true('failed' in response.content)
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
self.cluster.fs.mkdir("/user/test/jt-test_failed_jobs")
response = self.client.post(
"/jobsub/new/streaming",
dict(
hadoop_properties='',
inputformat_class_0='org.apache.hadoop.mapred.TextInputFormat',
input='/user/test/jt-test_failed_jobs',
mapper_cmd="python hogalabogus.py",
name="bogus stream",
num_reduce_tasks="1",
outputformat_class_0=
'org.apache.hadoop.mapred.TextOutputFormat',
output='/user/test/out_stream.tmp',
reducer_cmd='python hogalabogus.py',
save_submit='off',
submit='Save',
))
# Submit the job
design_id = response.context["saved"]
response = self.client.post("/jobsub/submit/%d" % design_id)
watch_id = parse_out_id(response)
response = watch_till_complete(self.client, watch_id, timeout_sec=120)
job_id = Submission.objects.get(id=watch_id).submission_handle.id
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
# 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)
# Select only failed jobs (should be present)
response = self.client.get('/jobbrowser/jobs/?state=failed')
assert_true(hadoop_job_id in response.content)
# Select only killed jobs (should be absent)
response = self.client.get('/jobbrowser/jobs/?state=killed')
assert_true(hadoop_job_id not in response.content)
# 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)
def test_new_jobs(self):
"""
Submit jobs. Let them succeed or fail and view them.
"""
# Install examples
import jobsub.management.commands.jobsub_setup as jobsub_setup
if not jobsub_setup.Command().has_been_setup():
jobsub_setup.Command().handle()
# Run the sleep example, since it doesn't require user home directory
design_id = JobDesign.objects.get(name__contains="Example: Sleep").id
response = self.client.post("/jobsub/submit/%d" % (design_id,), dict(
map_sleep_time_millis=1,
num_mappers=1,
num_reducers=1,
reduce_sleep_time_millis=1))
watch_id = parse_out_id(response)
response = watch_till_complete(self.client, watch_id)
job_id = Submission.objects.get(id=watch_id).submission_handle.id
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
# All jobs page
response = self.client.get('/jobbrowser/jobs/')
assert_true(hadoop_job_id.lstrip('job_') in response.content)
# 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['FileSystemCounters']['counters']['FILE_BYTES_WRITTEN']
assert_true(counters_file_bytes_written['map'] > 0)
assert_true(counters_file_bytes_written['reduce'] > 0)
assert_equal(counters_file_bytes_written['displayName'], 'FILE_BYTES_WRITTEN')
assert_equal(counters_file_bytes_written['displayName'], 'FILE_BYTES_WRITTEN')
# 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.mapred.JobInProgress$Counter']['counters']['TOTAL_LAUNCHED_MAPS']['total'], 1)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapred.JobInProgress$Counter']['counters']['TOTAL_LAUNCHED_REDUCES']['total'], 1)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapred.JobInProgress$Counter']['counters']['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapred.JobInProgress$Counter']['counters']['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0)
assert_true(response.context['job'].counters['org.apache.hadoop.mapred.JobInProgress$Counter']['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
assert_true(response.context['job'].counters['org.apache.hadoop.mapred.JobInProgress$Counter']['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)
# Check conf keys made it
assert_equal(response.context['job'].conf_keys['mapredReducerClass'],
'org.apache.hadoop.examples.SleepJob')
# 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('Showing 1 to 4 of 4 tasks' in response.content)
# Select by tasktype
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id,))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Select by taskstate
response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
assert_true('Showing 1 to 4 of 4 tasks' in response.content)
# Select by text
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktext=clean' % (hadoop_job_id,))
assert_true('Showing 1 to 1 of 1 tasks' in response.content)
# Run another sleep job but kill it
response = self.client.post("/jobsub/submit/%d" % (design_id,), dict(
map_sleep_time_millis=1,
num_mappers=2000,
num_reducers=2000,
reduce_sleep_time_millis=1))
job_id = parse_out_id(response)
hadoop_job_id = get_hadoop_job_id(self.jobsubd, job_id)
client2 = make_logged_in_client('test_non_superuser', is_superuser=False)
response = client2.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id,))
assert_equal("Permission denied. User test_non_superuser cannot delete user test's job.",
response.context["error"])
# Make sure that the first map task succeeds before moving on
# This will keep us from hitting timing-related failures
first_mapper = hadoop_job_id.replace('job', 'task') + '_m_000000'
start = time.time()
timeout_sec = 60
while first_mapper not in \
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,)).content:
time.sleep(1)
# If this assert fails, something has probably really failed
assert_true(time.time() - start < timeout_sec,
"Timed out waiting for first mapper to complete")
self.client.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id,))
# It should say killed
response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id,))
html = response.content.lower()
assert_true(hadoop_job_id in html)
assert_true('killed' in html)
# Exercise select by taskstate
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=pending' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=running' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=killed' % (hadoop_job_id,))
# Test single task page
late_task_id = hadoop_job_id.replace('job', 'task') + '_r_001999'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, late_task_id))
assert_false('succeed' in response.content)
assert_true('killed' in response.content)
# The first task should've succeeded
# We use a different method of checking success for this one
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, early_task_id))
assert_true('succeed' in response.content)
assert_false('failed' in response.content)
# Test single attempt page
attempt_id = early_task_id.replace('task', 'attempt') + '_0'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s/attempts/%s' %
(hadoop_job_id, early_task_id, attempt_id))
assert_true('syslog' in response.content)
# Test dock jobs
response = self.client.get('/jobbrowser/dock_jobs/')
assert_true('completed' in response.content)
# TODO(atm): I'm pretty sure the following test only passes because of
# failed jobs which are run in test_failed_jobs
assert_true('failed' in response.content)
