def pre_process_data(ns, tf, log_details):
"""
For convenience, perform operations on the input stream before passing
along to other data processing code
"""
if ns.sample:
log.info('sampling a percentage of input data without replacement',
extra=dict(sampling_pct=ns.sample, **log_details))
tf = tf.sample(False, ns.sample, 0)
if ns.mapjson:
log.info('converting all elements in data stream to json')
tf = tf.map(simplejson.loads)
return tf
def apply_data_transform(ns, sc, log_details, pjob_id, module):
"""Pass control to the module.main method. If module.main specifies a
`textFile` parameter, pass the textFile instance. Otherwise, just map
module.main on the RDD
"""
func_args = inspect.getargspec(module.main).args
if 'sc' in func_args:
log.info(
'passing spark context to a module.main function',
extra=log_details)
try:
module.main(sc=sc, ns=ns, **pjob_id)
except Exception as err:
log_and_raise(
"Job failed with error: %s" % err, log_details)
else:
read_fp = format_fp(ns.read_fp, ns, pjob_id)
log_details = dict(read_fp=read_fp, **log_details)
tf = sc.textFile(read_fp, ns.minPartitions)
tf = pre_process_data(ns=ns, tf=tf, log_details=log_details)
if 'textFile' in func_args:
log.info(
'passing textFile instance to a module.main function',
extra=log_details)
try:
module.main(textFile=tf, ns=ns, **pjob_id)
except Exception as err:
log_and_raise(
"Job failed with error: %s" % err, log_details)
else:
write_fp = format_fp(ns.write_fp, ns, pjob_id)
log.info(
'mapping a module.main function to all elements in a textFile'
' and writing output',
extra=dict(write_fp=write_fp, **log_details))
try:
(
tf
.map(functools.partial(module.main, ns=ns, **pjob_id))
.saveAsTextFile(write_fp)
)
except Exception as err:
log_and_raise(err, log_details)
def main(ns):
"""
A generic plugin that schedules arbitrary bash jobs using Stolos
Assume code is written in Python. For Scala or R code, use another option.
"""
job_id = ns.job_id
ld = dict(app_name=ns.app_name, job_id=ns.job_id)
log.info('Running bash job', extra=ld)
cmd = get_bash_cmd(ns.app_name)
if ns.bash_cmd:
cmd += ' '.join(ns.bash_cmd)
log.debug("Appending user-supplied bash options to defaults",
extra=dict(app_name=ns.app_name, job_id=job_id, cmd=cmd))
ld.update(cmd=cmd)
if not cmd:
raise UserWarning(
"You need to specify bash options or configure default bash"
" options")
_cmdargs = dict(**ns.__dict__)
_cmdargs.update(api.parse_job_id(ns.app_name, job_id))
cmd = cmd.format(**_cmdargs)
if ns.redirect_to_stderr:
_std = sys.stderr
else:
_std = PIPE
log.info('running command', extra=ld)
returncode, stdout, stderr = run(cmd,
shell=True,
timeout=ns.watch,
stdout=_std,
stderr=_std)
ld = dict(bash_returncode=returncode, stdout=stdout, stderr=stderr, **ld)
if returncode == -9:
log_and_raise("Bash job timed out", ld)
elif returncode != 0:
# this raises an error and logs output:
log_and_raise("Bash job failed", ld)
else:
log.info("Bash job succeeded", extra=ld)
def main(ns):
"""
A generic plugin that schedules arbitrary bash jobs using Stolos
Assume code is written in Python. For Scala or R code, use another option.
"""
job_id = ns.job_id
ld = dict(app_name=ns.app_name, job_id=ns.job_id)
log.info('Running bash job', extra=ld)
cmd = get_bash_cmd(ns.app_name)
if ns.bash_cmd:
cmd += ' '.join(ns.bash_cmd)
log.debug(
"Appending user-supplied bash options to defaults", extra=dict(
app_name=ns.app_name, job_id=job_id, cmd=cmd))
ld.update(cmd=cmd)
if not cmd:
raise UserWarning(
"You need to specify bash options or configure default bash"
" options")
_cmdargs = dict(**ns.__dict__)
_cmdargs.update(api.parse_job_id(ns.app_name, job_id))
cmd = cmd.format(**_cmdargs)
if ns.redirect_to_stderr:
_std = sys.stderr
else:
_std = PIPE
log.info('running command', extra=ld)
returncode, stdout, stderr = run(
cmd, shell=True, timeout=ns.watch, stdout=_std, stderr=_std)
ld = dict(bash_returncode=returncode, stdout=stdout, stderr=stderr, **ld)
if returncode == -9:
log_and_raise("Bash job timed out", ld)
elif returncode != 0:
# this raises an error and logs output:
log_and_raise("Bash job failed", ld)
else:
log.info("Bash job succeeded", extra=ld)