def run_hive_jobs(start_dt, end_dt):
jobname = "Daily Exercise Stats (%s to %s)" % (start_dt, end_dt)
jobflow = emr.create_hive_cluster(jobname, {})
# TODO(jace): make sure the required data (ProblemLogs, etc)
# is available before running these downstream summaries
day = datetime.datetime.strptime(end_dt, DATE_FORMAT)
start_day = datetime.datetime.strptime(start_dt, DATE_FORMAT)
day -= datetime.timedelta(days=1) # end date is exclusive
while day >= start_day:
emr.add_hive_step(jobflow, {},
's3://ka-mapreduce/code/hive/daily_ex_stats.q',
script_args={
"dt": day.strftime(DATE_FORMAT),
"branch": "/"
})
day -= datetime.timedelta(days=1)
return jobflow, emr.wait_for_completion(jobflow)
def run_hive_jobs(start_dt, end_dt):
jobname = "Daily Exercise Stats (%s to %s)" % (start_dt, end_dt)
jobflow = emr.create_hive_cluster(jobname, {})
# TODO(jace): make sure the required data (ProblemLogs, etc)
# is available before running these downstream summaries
day = datetime.datetime.strptime(end_dt, DATE_FORMAT)
start_day = datetime.datetime.strptime(start_dt, DATE_FORMAT)
day -= datetime.timedelta(days=1) # end date is exclusive
while day >= start_day:
emr.add_hive_step(
jobflow,
{},
"s3://ka-mapreduce/code/hive/daily_ex_stats.q",
script_args={"dt": day.strftime(DATE_FORMAT), "branch": "/"},
)
day -= datetime.timedelta(days=1)
return jobflow, emr.wait_for_completion(jobflow)
def run_hive_jobs(jobname, steps, num_instances):
"""Run hive steps.
Arguments:
jobname: Name for the Amazon EMR job.
steps: A sequence of dictionaries describing the job steps to add.
Each step may specify the keys "hive_script" and "hive_args". If
"hive_script" is missing, no job step will be added. These steps
usually come directly from a configuration file.
num_instances: The number of instances to run this job on. Equivalent
to the EMR CLI option --num-instances.
Calls sys.exit() when a job does not complete successfully.
"""
jobflow = emr.create_hive_cluster(
jobname, {"num_instances": num_instances})
for step in steps:
# It's possible to leave out hive_script and hive_args, for
# when the step just wants to move data from hive into mongo,
# and not run any hive script.
if 'hive_script' not in step:
continue
emr.add_hive_step(jobflow, {},
hive_script=step["hive_script"],
script_args=step["hive_args"])
status = emr.wait_for_completion(jobflow, logger=g_logger)
listing = emr.list_steps(jobflow)
failures = ["FAILED", "CANCELLED", "TERMINATED"]
if any(s in listing for s in failures):
subject = "Reporting jobflow FAILED: %s" % jobname
notify.send_email(subject, listing)
notify.send_hipchat(subject)
else:
subject = "Reporting jobflow SUCCEEDED: %s" % jobname
notify.send_email(subject, listing)
if status != "COMPLETED":
g_logger.fatal("Hive jobs failed")
g_logger.fatal(emr.list_steps(jobflow))
sys.exit(1)
def run_hive_jobs(jobname, steps, num_instances):
"""Run hive steps.
Arguments:
jobname: Name for the Amazon EMR job.
steps: A sequence of dictionaries describing the job steps to add.
Each step may specify the keys "hive_script" and "hive_args". If
"hive_script" is missing, no job step will be added. These steps
usually come directly from a configuration file.
num_instances: The number of instances to run this job on. Equivalent
to the EMR CLI option --num-instances.
Calls sys.exit() when a job does not complete successfully.
"""
jobflow = emr.create_hive_cluster(
jobname, {"num_instances": num_instances})
for step in steps:
# It's possible to leave out hive_script and hive_args, for
# when the step just wants to move data from hive into mongo,
# and not run any hive script.
if 'hive_script' not in step:
continue
emr.add_hive_step(jobflow, {},
hive_script=step["hive_script"],
script_args=step.get("hive_args", {}))
status = emr.wait_for_completion(jobflow, logger=g_logger)
listing = emr.list_steps(jobflow)
failures = ["FAILED", "CANCELLED", "TERMINATED"]
if any(s in listing for s in failures):
subject = "Reporting jobflow FAILED: %s" % jobname
notify.send_email(subject, listing)
notify.send_hipchat(subject)
else:
subject = "Reporting jobflow SUCCEEDED: %s" % jobname
notify.send_email(subject, listing)
if status != "COMPLETED":
g_logger.fatal("Hive jobs failed")
g_logger.fatal(emr.list_steps(jobflow))
sys.exit(1)
def run_hive_jobs(start_dt, end_dt, earliest_dt):
jobname = "Growth Reporting (%s to %s)" % (start_dt, end_dt)
jobflow = emr.create_hive_cluster(jobname, {})
# TODO(jace): make sure the required data (ProblemLogs, etc)
# is available before running these downstream summaries
emr.add_hive_step(jobflow, {},
hive_script='s3://ka-mapreduce/code/hive/user_daily_activity.q',
script_args={"start_dt": start_dt, "end_dt": end_dt})
emr.add_hive_step(jobflow, {},
hive_script='s3://ka-mapreduce/code/hive/user_growth.q',
script_args={"start_dt": earliest_dt, "end_dt": end_dt})
emr.add_hive_step(jobflow, {},
hive_script='s3://ka-mapreduce/code/hive/company_metrics.q',
script_args={"start_dt": earliest_dt, "end_dt": end_dt})
return jobflow, emr.wait_for_completion(jobflow)
