def get_unfinished_jobs(cls):
"""
Queries SimpleDB for a list of pending jobs that have not yet been
finished.
:rtype: list
:returns: A list of unfinished :py:class:`EncodingJob` objects.
"""
query_str = (
"SELECT * FROM %s WHERE job_state != '%s' "
"and job_state != '%s' "
"and job_state != '%s'" % (settings.SIMPLEDB_JOB_STATE_DOMAIN, "FINISHED", "ERROR", "ABANDONED")
)
results = cls._get_sdb_job_state_domain().select(query_str)
jobs = []
for item in results:
try:
job = cls._get_job_object_from_item(item)
except TypeError:
message = "JobStateBackend.get_unfinished_jobs(): " "Unable to instantiate job: %s" % item
logger.error(message_or_obj=message)
logger.error()
continue
except ImportError:
message = "JobStateBackend.get_unfinished_jobs(): " "Invalid nommer specified for job: %s" % item
logger.error(message_or_obj=message)
logger.error()
continue
jobs.append(job)
return jobs
def spawn_instances(cls, num_instances):
"""
Spawns the number of instances specified.
:param int num_instances: The number of instances to spawn.
:rtype: :py:class:`boto.ec2.instance.Reservation`
:returns: A boto Reservation for the started instance(s).
"""
logger.info("EC2InstanceManager.spawn_instances(): " \
"Spawning %d new instances" % num_instances)
conn = cls._aws_ec2_connection()
try:
image = conn.get_all_images(image_ids=settings.EC2_AMI_ID)
image = image[0]
except EC2ResponseError:
logger.error("EC2InstanceManager.spawn_instances(): " \
"No AMI with ID %s could be found." % settings.EC2_AMI_ID)
logger.error()
return
except IndexError:
logger.error("EC2InstanceManager.spawn_instances(): " \
"No AMI with ID %s could be found." % settings.EC2_AMI_ID)
logger.error()
return
# The boto Reservation object. Its 'instances' attribute is the
# important bit.
return image.run(min_count=num_instances, max_count=num_instances,
instance_type=settings.EC2_INSTANCE_TYPE,
security_groups=settings.EC2_SECURITY_GROUPS,
key_name=settings.EC2_KEY_NAME,
user_data=cls._gen_ec2_user_data())
def spawn_instances(cls, num_instances):
"""
Spawns the number of instances specified.
:param int num_instances: The number of instances to spawn.
:rtype: :py:class:`boto.ec2.instance.Reservation`
:returns: A boto Reservation for the started instance(s).
"""
logger.info("EC2InstanceManager.spawn_instances(): " \
"Spawning %d new instances" % num_instances)
conn = cls._aws_ec2_connection()
try:
image = conn.get_all_images(image_ids=settings.EC2_AMI_ID)
image = image[0]
except EC2ResponseError:
logger.error("EC2InstanceManager.spawn_instances(): " \
"No AMI with ID %s could be found." % settings.EC2_AMI_ID)
logger.error()
return
except IndexError:
logger.error("EC2InstanceManager.spawn_instances(): " \
"No AMI with ID %s could be found." % settings.EC2_AMI_ID)
logger.error()
return
# The boto Reservation object. Its 'instances' attribute is the
# important bit.
return image.run(min_count=num_instances,
max_count=num_instances,
instance_type=settings.EC2_INSTANCE_TYPE,
security_groups=settings.EC2_SECURITY_GROUPS,
key_name=settings.EC2_KEY_NAME,
user_data=cls._gen_ec2_user_data())
def refresh_jobs_with_state_changes(cls):
"""
Looks at the state SQS queue specified by the
:py:data:`SQS_JOB_STATE_CHANGE_QUEUE_NAME <media_nommer.conf.settings.SQS_JOB_STATE_CHANGE_QUEUE_NAME>`
setting and refreshes any jobs that have changed. This simply reloads
the job's details from SimpleDB_.
:rtype: ``list`` of :py:class:`EncodingJob <media_nommer.core.job_state_backend.EncodingJob>`
:returns: A list of changed :py:class:`EncodingJob` objects.
"""
logger.debug("JobCache.refresh_jobs_with_state_changes(): " \
"Checking state change queue.")
# Pops up to 10 changed jobs that we think may have changed. There are
# some false alarms in here, whch brings us to...
popped_changed_jobs = JobStateBackend.pop_state_changes_from_queue(10)
# A temporary list that stores the jobs that actually changed. This
# will be returned at the completion of this method's path.
changed_jobs = []
if popped_changed_jobs:
logger.debug("Potential job state changes found: %s" %
popped_changed_jobs)
for job in popped_changed_jobs:
if cls.is_job_cached(job):
current_state = cls.get_job(job).job_state
new_state = job.job_state
if current_state != new_state:
logger.info("* Job state changed %s: %s -> %s" % (
job.unique_id,
# Current job state in cache
current_state,
# New incoming job state
new_state,
))
cls.update_job(job)
# This one actually changed, append this for returning.
changed_jobs.append(job)
if new_state == 'ERROR':
logger.error('Error trace from ec2nommerd:')
logger.error(job.job_state_details)
return changed_jobs
def refresh_jobs_with_state_changes(cls):
"""
Looks at the state SQS queue specified by the
:py:data:`SQS_JOB_STATE_CHANGE_QUEUE_NAME <media_nommer.conf.settings.SQS_JOB_STATE_CHANGE_QUEUE_NAME>`
setting and refreshes any jobs that have changed. This simply reloads
the job's details from SimpleDB_.
:rtype: ``list`` of :py:class:`EncodingJob <media_nommer.core.job_state_backend.EncodingJob>`
:returns: A list of changed :py:class:`EncodingJob` objects.
"""
logger.debug("JobCache.refresh_jobs_with_state_changes(): " \
"Checking state change queue.")
# Pops up to 10 changed jobs that we think may have changed. There are
# some false alarms in here, whch brings us to...
popped_changed_jobs = JobStateBackend.pop_state_changes_from_queue(10)
# A temporary list that stores the jobs that actually changed. This
# will be returned at the completion of this method's path.
changed_jobs = []
if popped_changed_jobs:
logger.debug("Potential job state changes found: %s" % popped_changed_jobs)
for job in popped_changed_jobs:
if cls.is_job_cached(job):
current_state = cls.get_job(job).job_state
new_state = job.job_state
if current_state != new_state:
logger.info("* Job state changed %s: %s -> %s" % (
job.unique_id,
# Current job state in cache
current_state,
# New incoming job state
new_state,
))
cls.update_job(job)
# This one actually changed, append this for returning.
changed_jobs.append(job)
if new_state == 'ERROR':
logger.error('Error trace from ec2nommerd:')
logger.error(job.job_state_details)
return changed_jobs
def __run_ffmpeg(self, fobj):
"""
Fire up ffmpeg and toss the results into a temporary file.
:rtype: file-like object or ``None``
:returns: If the encoding succeeds, a file-like object is returned.
If an error happens, ``None`` is returned, and the ERROR job
state is set.
"""
is_two_pass = len(self.job.job_options) > 1
out_fobj = tempfile.NamedTemporaryFile(mode='w+b', delete=True)
pass_counter = 1
for encoding_pass_options in self.job.job_options:
is_second_pass = pass_counter == 2
# Based on the given options, assemble the command list to
# pass on to Popen.
ffmpeg_cmd = self.__assemble_ffmpeg_cmd_list(
encoding_pass_options,
fobj,
out_fobj,
is_two_pass=is_two_pass,
is_second_pass=is_second_pass)
# Do this for ffmpeg's sake. Allows more than one concurrent
# encoding job per EC2 instance.
os.chdir(self.temp_cwd)
# Fire up ffmpeg.
process = subprocess.Popen(
ffmpeg_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
# Get back to home dir. Not sure this is necessary, but meh.
os.chdir(os.path.expanduser('~'))
# Block here while waiting for output
cmd_output = process.communicate()
# 0 is success, so anything but that is bad.
error_happened = process.returncode != 0
if not error_happened and (is_second_pass or not is_two_pass):
# No errors, return the file object for uploading.
out_fobj.seek(0)
return out_fobj
elif error_happened:
# Error found, return nothing so the nommer can die.
logger.error(
message_or_obj="Error encountered while running ffmpeg.")
logger.error(message_or_obj=cmd_output[0])
logger.error(message_or_obj=cmd_output[1])
self.wrapped_set_job_state('ERROR', details=cmd_output[1])
return None
pass_counter += 1
def get_unfinished_jobs(cls):
"""
Queries SimpleDB for a list of pending jobs that have not yet been
finished.
:rtype: list
:returns: A list of unfinished :py:class:`EncodingJob` objects.
"""
query_str = "SELECT * FROM %s WHERE job_state != '%s' " \
"and job_state != '%s' " \
"and job_state != '%s'" % (
settings.SIMPLEDB_JOB_STATE_DOMAIN,
'FINISHED',
'ERROR',
'ABANDONED',
)
results = cls._get_sdb_job_state_domain().select(query_str)
jobs = []
for item in results:
try:
job = cls._get_job_object_from_item(item)
except TypeError:
message = "JobStateBackend.get_unfinished_jobs(): " \
"Unable to instantiate job: %s" % item
logger.error(message_or_obj=message)
logger.error()
continue
except ImportError:
message = "JobStateBackend.get_unfinished_jobs(): " \
"Invalid nommer specified for job: %s" % item
logger.error(message_or_obj=message)
logger.error()
continue
jobs.append(job)
return jobs
def __run_ffmpeg(self, fobj):
"""
Fire up ffmpeg and toss the results into a temporary file.
:rtype: file-like object or ``None``
:returns: If the encoding succeeds, a file-like object is returned.
If an error happens, ``None`` is returned, and the ERROR job
state is set.
"""
is_two_pass = len(self.job.job_options) > 1
out_fobj = tempfile.NamedTemporaryFile(mode='w+b', delete=True)
pass_counter = 1
for encoding_pass_options in self.job.job_options:
is_second_pass = pass_counter == 2
# Based on the given options, assemble the command list to
# pass on to Popen.
ffmpeg_cmd = self.__assemble_ffmpeg_cmd_list(
encoding_pass_options,
fobj, out_fobj,
is_two_pass=is_two_pass,
is_second_pass=is_second_pass)
# Do this for ffmpeg's sake. Allows more than one concurrent
# encoding job per EC2 instance.
os.chdir(self.temp_cwd)
# Fire up ffmpeg.
process = subprocess.Popen(ffmpeg_cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
# Get back to home dir. Not sure this is necessary, but meh.
os.chdir(os.path.expanduser('~'))
# Block here while waiting for output
cmd_output = process.communicate()
# 0 is success, so anything but that is bad.
error_happened = process.returncode != 0
if error_happened:
# Error found, return nothing so the nommer can die.
logger.error(
message_or_obj="Error encountered while running ffmpeg.")
logger.error(message_or_obj=cmd_output[0])
logger.error(message_or_obj=cmd_output[1])
self.wrapped_set_job_state('ERROR', details=cmd_output[1])
return None
move_atom = encoding_pass_options.get('move_atom_to_front', False)
if move_atom:
qtfaststart_cmd = self.__assemble_qtfaststart_cmd_list(out_fobj)
process = subprocess.Popen(qtfaststart_cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
cmd_output = process.communicate()
error_happened = process.returncode != 0
if error_happened:
# Error found, return nothing so the nommer can die.
logger.error(
message_or_obj="Error encountered while running qtfaststart.")
logger.error(message_or_obj=cmd_output[0])
logger.error(message_or_obj=cmd_output[1])
self.wrapped_set_job_state('ERROR', details=cmd_output[1])
return None
if is_second_pass or not is_two_pass:
# No errors, return the file object for uploading.
out_fobj.seek(0)
return out_fobj
pass_counter += 1
