def collect_flow_execs(self, flow_exec_file, job_exec_file, look_back_period):
self.logger.info( "collect flow&job executions")
flow_exec_writer = FileWriter(flow_exec_file)
job_exec_writer = FileWriter(job_exec_file)
cmd = """select * from execution_flows where end_time > UNIX_TIMESTAMP(now() - INTERVAL %d MINUTE) * 1000 """ % (int(look_back_period))
self.az_cursor.execute(cmd)
rows = DbUtil.dict_cursor(self.az_cursor)
row_count = 0
for row in rows:
json_column = 'flow_data'
unzipped_content = gzip.GzipFile(mode='r', fileobj=StringIO.StringIO(row[json_column].tostring())).read()
try:
row[json_column] = json.loads(unzipped_content)
except Exception as e:
self.logger.error(e)
pass
flow_data = row[json_column]
flow_path = flow_data['projectName'] + ":" + flow_data['flowId']
flow_exec_record = AzkabanFlowExecRecord(self.app_id,
flow_data['flowId'],
flow_path,
row['version'],
row['exec_id'],
flow_data['status'],
flow_data['attempt'],
row['submit_user'],
long(row['start_time']) / 1000,
long(row['end_time']) / 1000,
self.wh_exec_id)
flow_exec_writer.append(flow_exec_record)
nodes = flow_data['nodes']
job_exec_records = []
for node in nodes:
job_exec_record = AzkabanJobExecRecord(self.app_id,
flow_path,
row['version'],
row['exec_id'],
node['id'],
flow_path + "/" + node['id'],
None,
node['status'],
node['attempt'],
long(node['startTime']) / 1000,
long(node['endTime']) / 1000,
self.wh_exec_id)
job_exec_records.append(job_exec_record)
AzkabanJobExecUtil.sortAndSet(job_exec_records)
for r in job_exec_records:
job_exec_writer.append(r)
row_count += 1
if row_count % 10000 == 0:
flow_exec_writer.flush()
job_exec_writer.flush()
flow_exec_writer.close()
job_exec_writer.close()
def collect_flow_jobs(self, flow_file, job_file, dag_file):
self.logger.info("collect flow&jobs")
flow_writer = FileWriter(flow_file)
job_writer = FileWriter(job_file)
dag_writer = FileWriter(dag_file)
query = """
SELECT a.*, b.created_time FROM
(SELECT w.app_name, w.app_path, max(w.id) as source_version, max(unix_timestamp(w.last_modified_time)) as last_modified_time
from WF_JOBS w LEFT JOIN WF_JOBS s
ON w.app_path = s.app_path AND w.created_time < s.created_time
WHERE s.created_time IS NULL GROUP BY w.app_name, w.app_path) a
JOIN
(SELECT app_path, min(unix_timestamp(created_time)) as created_time FROM WF_JOBS GROUP BY app_path) b
ON a.app_path = b.app_path
"""
self.oz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.oz_cursor)
for row in rows:
flow_record = OozieFlowRecord(self.app_id, row['app_name'],
row['app_path'], 0,
row['source_version'],
row['created_time'],
row['last_modified_time'],
self.wh_exec_id)
flow_writer.append(flow_record)
query = """
select name, type, transition from WF_ACTIONS
where wf_id = '{source_version}'
""".format(source_version=row['source_version'])
new_oz_cursor = self.oz_con.cursor()
new_oz_cursor.execute(query)
nodes = DbUtil.dict_cursor(new_oz_cursor)
for node in nodes:
job_record = OozieJobRecord(
self.app_id, row['app_path'], row['source_version'],
node['name'], row['app_path'] + "/" + node['name'],
node['type'], self.wh_exec_id)
job_writer.append(job_record)
if node['transition'] != "*" and node['transition'] is not None:
dag_edge = OozieFlowDagRecord(
self.app_id, row['app_path'], row['source_version'],
row['app_path'] + "/" + node['name'],
row['app_path'] + "/" + node['transition'],
self.wh_exec_id)
dag_writer.append(dag_edge)
new_oz_cursor.close()
dag_writer.close()
job_writer.close()
flow_writer.close()
def __init__(self):
self.logger = LoggerFactory.getLogger('jython script : ' + self.__class__.__name__)
requests.packages.urllib3.disable_warnings()
self.app_id = int(args[Constant.APP_ID_KEY])
self.wh_exec_id = long(args[Constant.WH_EXEC_ID_KEY])
self.project_writer = FileWriter(args[Constant.GIT_PROJECT_OUTPUT_KEY])
self.repo_writer = FileWriter(args[Constant.PRODUCT_REPO_OUTPUT_KEY])
self.repo_owner_writer = FileWriter(args[Constant.PRODUCT_REPO_OWNER_OUTPUT_KEY])
self.multiproduct = {}
self.git_repo = {}
self.product_repo = []
def transform(this, raw_metadata, metadata_output, field_metadata_output):
input_json_file = open(raw_metadata, 'r')
schema_file_writer = FileWriter(metadata_output)
field_file_writer = FileWriter(field_metadata_output)
i = 0
this.sort_id = 0
o_urn = ''
for line in input_json_file:
try:
j = json.loads(line)
except:
this.logger.error(" Invalid JSON:\n%s" % line)
continue
i += 1
o_field_list_ = []
this.sort_id = 0
if not j.has_key('attributes'):
o_properties = {"doc": null}
else:
o_properties = dict(j['attributes'].items())
del j['attributes']
if j.has_key('uri'):
o_urn = j['uri']
o_name = o_urn[o_urn.rfind('/') + 1:]
o_source = 'Hbase'
else:
this.logger.info('*** Warning: "uri" is not found in %s' %
j['name'])
o_urn = ''
o_name = ''
if not j.has_key('fields'):
o_fields = {"doc": None}
else:
o_fields = {}
for f in j['fields']:
o_field_name = f['name']
o_fields[o_field_name] = dict(f)
acp = AvroColumnParser(j, o_urn)
o_field_list_ += acp.get_column_list_result()
dataset_schema_record = DatasetSchemaRecord(
o_name, json.dumps(j, sort_keys=True),
json.dumps(o_properties, sort_keys=True), json.dumps(o_fields),
o_urn, o_source, 'HBase', 'Table', None, None, None)
schema_file_writer.append(dataset_schema_record)
for fields in o_field_list_:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
field_file_writer.close()
schema_file_writer.close()
input_json_file.close()
def collect_flow_execs(self, flow_exec_file, job_exec_file,
look_back_period):
self.logger.info("collect flow&job executions")
flow_exec_writer = FileWriter(flow_exec_file)
job_exec_writer = FileWriter(job_exec_file)
cmd = "SELECT * FROM workflow_info WHERE status is NULL"
self.lz_cursor.execute(cmd)
# rows = DbUtil.dict_cursor(self.lz_cursor)
rows = DbUtil.copy_dict_cursor(self.lz_cursor)
row_count = 0
for row in rows:
flow_path = row['project_name'] + ":" + row['workflow_name']
flow_exec_record = LhotseFlowExecRecord(
self.app_id, row["workflow_name"], flow_path, 0, 1,
"SUCCEEDED", 1, row['owner'],
long(time.mktime(row['create_time'].timetuple())),
long(time.mktime(row['modify_time'].timetuple())),
self.wh_exec_id)
flow_exec_writer.append(flow_exec_record)
job_exec_records = []
task_query = "SELECT * FROM task_info WHERE workflow_id = \"{0}\"".format(
row['workflow_id'])
new_lz_cursor = self.lz_cursor
new_lz_cursor.execute(task_query)
task_rows = DbUtil.dict_cursor(new_lz_cursor)
for task in task_rows:
if task['real_task_id'] is None:
continue
job_exec_record = LhotseJobExecRecord(
self.app_id, flow_path, 0, 1, task['task_name'],
flow_path + "/" + task['task_name'],
long(task['real_task_id']), 'SUCCEEDED', 1,
int(time.mktime(task['create_time'].timetuple())),
int(time.mktime(task['modify_time'].timetuple())),
self.wh_exec_id)
job_exec_records.append(job_exec_record)
## LhotseJobExecRecord.sortAndSet(job_exec_records)
for r in job_exec_records:
job_exec_writer.append(r)
row_count += 1
if row_count % 10000 == 0:
flow_exec_writer.flush()
job_exec_writer.flush()
flow_exec_writer.close()
job_exec_writer.close()
def __init__(self, args):
self.logger = LoggerFactory.getLogger('jython script : ' + self.__class__.__name__)
self.base_url = args[Constant.BASE_URL_KEY]
temp_dir = FileUtil.etl_temp_dir(args, "CODESEARCH")
self.code_search_committer_writer = FileWriter(
os.path.join(temp_dir, args[Constant.DATABASE_SCM_REPO_OUTPUT_KEY]))
def collect_flow_schedules(self, schedule_file):
# load flow scheduling info from table triggers
self.logger.info("collect flow schedule")
timezone = "ALTER SESSION SET TIME_ZONE = 'US/Pacific'"
self.aw_cursor.execute(timezone)
schema = "ALTER SESSION SET CURRENT_SCHEMA=APPWORX"
self.aw_cursor.execute(schema)
schedule_writer = FileWriter(schedule_file)
query = \
"""SELECT J.SO_APPLICATION, J.SO_MODULE, S.AW_SCH_NAME, S.AW_SCH_INTERVAL, S.AW_ACTIVE,
ROUND((cast((FROM_TZ(CAST(S.AW_SCH_START as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as EFFECT_STARTED,
ROUND((cast((FROM_TZ(CAST(S.AW_SCH_END as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as EFFECT_END
FROM SO_JOB_TABLE J
JOIN AW_MODULE_SCHED S ON J.SO_JOB_SEQ = S.AW_JOB_SEQ
WHERE J.SO_COMMAND_TYPE = 'CHAIN' AND S.AW_ACTIVE = 'Y' """
self.aw_cursor.execute(query)
rows = DbUtil.dict_cursor(self.aw_cursor)
for row in rows:
schedule_record = AppworxFlowScheduleRecord(
self.app_id, row['SO_APPLICATION'] + ":" + row['SO_MODULE'],
row['AW_SCH_NAME'], int(row['AW_SCH_INTERVAL']),
long(row['EFFECT_STARTED']), long(row['EFFECT_END']), '0',
self.wh_exec_id)
schedule_writer.append(schedule_record)
schedule_writer.close()
def __init__(self):
self.logger = LoggerFactory.getLogger('jython script : ' + self.__class__.__name__)
username = args[Constant.HIVE_METASTORE_USERNAME]
password = args[Constant.HIVE_METASTORE_PASSWORD]
jdbc_driver = args[Constant.HIVE_METASTORE_JDBC_DRIVER]
jdbc_url = args[Constant.HIVE_METASTORE_JDBC_URL]
self.conn_hms = zxJDBC.connect(jdbc_url, username, password, jdbc_driver)
self.curs = self.conn_hms.cursor()
dependency_instance_file = args[Constant.HIVE_DEPENDENCY_CSV_FILE_KEY]
self.instance_writer = FileWriter(dependency_instance_file)
def collect_flow_schedules(self, schedule_file):
# load flow scheduling info from table triggers
self.logger.info("collect flow schedule")
schedule_writer = FileWriter(schedule_file)
query = "select * from triggers"
self.az_cursor.execute(query)
rows = DbUtil.dict_cursor(self.az_cursor)
for row in rows:
json_column = 'data'
if row[json_column] != None:
unzipped_content = gzip.GzipFile(
mode='r',
fileobj=StringIO.StringIO(
row[json_column].tostring())).read()
try:
row[json_column] = json.loads(unzipped_content)
except Exception as e:
self.logger.error(e)
pass
if not "projectId" in row[json_column]["actions"][0][
"actionJson"]:
continue
# print json.dumps(row[json_column], indent=4)
if row[json_column]["triggerCondition"]["checkers"][0][
"checkerJson"]["isRecurring"] == 'true':
unit, frequency, cron_expr = None, None, None
period = row[json_column]["triggerCondition"]["checkers"][
0]["checkerJson"]["period"]
if period is not None and period != "null" and period[
-1:] in self._period_unit_table:
unit = self._period_unit_table[period[-1:]]
frequency = int(
row[json_column]["triggerCondition"]["checkers"][0]
["checkerJson"]["period"][:-1])
if "cronExpression" in row[json_column][
"triggerCondition"]["checkers"][0]["checkerJson"]:
cron_expr = row[json_column]["triggerCondition"][
"checkers"][0]["checkerJson"]["cronExpression"]
schedule_record = AzkabanFlowScheduleRecord(
self.app_id, row[json_column]["actions"][0]
["actionJson"]["projectName"] + ':' + row[json_column]
["actions"][0]["actionJson"]["flowName"], unit,
frequency, cron_expr,
long(row[json_column]["triggerCondition"]["checkers"]
[0]["checkerJson"]["firstCheckTime"]) / 1000,
int(
time.mktime(
datetime.date(2099, 12, 31).timetuple())), '0',
self.wh_exec_id)
schedule_writer.append(schedule_record)
schedule_writer.close()
def collect_flow_owners(self, owner_file):
self.logger.info("collect owners")
owner_writer = FileWriter(owner_file)
query = "SELECT DISTINCT app_name, app_path, user_name from WF_JOBS"
self.oz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.oz_cursor)
for row in rows:
owner_record = OozieFlowOwnerRecord(self.app_id, row['app_path'],
row['user_name'],
self.wh_exec_id)
owner_writer.append(owner_record)
owner_writer.close()
def collect_flow_owners(self, owner_file):
# load user info from table project_permissions
self.logger.info("collect owner&permissions")
user_writer = FileWriter(owner_file)
query = "SELECT project_name, workflow_name, owner FROM workflow_info WHERE status is NULL"
self.lz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.lz_cursor)
for row in rows:
record = LhotseFlowOwnerRecord(
self.app_id, row['project_name'] + ':' + row["workflow_name"],
row["owner"], 'ADMIN', 'LDAP', self.wh_exec_id)
user_writer.append(record)
user_writer.close()
def collect_flow_execs(self, flow_exec_file, lookback_period):
self.logger.info("collect flow execs")
flow_exec_writer = FileWriter(flow_exec_file)
query = "select id, app_name, app_path, unix_timestamp(start_time) as start_time, unix_timestamp(end_time) as end_time, run, status, user_name from WF_JOBS where end_time > now() - INTERVAL %d MINUTE" % (
int(lookback_period))
self.oz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.oz_cursor)
for row in rows:
flow_exec_record = OozieFlowExecRecord(
self.app_id, row['app_name'], row['app_path'], row['id'],
row['id'], row['status'], row['run'], row['user_name'],
row['start_time'], row['end_time'], self.wh_exec_id)
flow_exec_writer.append(flow_exec_record)
flow_exec_writer.close()
def collect_flow_owners(self, owner_file):
# load user info from table project_permissions
self.logger.info("collect owner&permissions")
user_writer = FileWriter(owner_file)
query = "select f.flow_id, p.name as project_name, p.version as project_verison, pp.name as owner, pp.permissions, pp.isGroup " \
"from project_flows f join project_permissions pp on f.project_id = pp.project_id join projects p on f.project_id = p.id where p.active = 1"
self.az_cursor.execute(query)
rows = DbUtil.dict_cursor(self.az_cursor)
for row in rows:
record = AzkabanFlowOwnerRecord(
self.app_id, row['project_name'] + ':' + row["flow_id"],
row["owner"],
AzkabanPermission(row["permissions"]).toFlatString(),
'GROUP' if row['isGroup'] == 1 else 'LDAP', self.wh_exec_id)
user_writer.append(record)
user_writer.close()
def collect_job_execs(self, job_exec_file, lookback_period):
self.logger.info("collect job execs")
job_exec_writer = FileWriter(job_exec_file)
query = """
select a.id as job_exec_id, a.name as job_name, j.id as flow_exec_id, a.status, a.user_retry_count,
unix_timestamp(a.start_time) start_time, unix_timestamp(a.end_time) end_time,
j.app_name as jname, j.app_path, transition from WF_ACTIONS a JOIN WF_JOBS j on a.wf_id = j.id where j.end_time > now() - INTERVAL %d MINUTE
""" % (int(lookback_period))
self.oz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.oz_cursor)
for row in rows:
job_exec_record = OozieJobExecRecord(
self.app_id, row['app_path'], row['flow_exec_id'],
row['flow_exec_id'], row['job_name'],
row['app_path'] + "/" + row['job_name'], row['job_exec_id'],
row['status'], row['user_retry_count'], row['start_time'],
row['end_time'], self.wh_exec_id)
job_exec_writer.append(job_exec_record)
job_exec_writer.close()
def collect_flow_owners(self, owner_file):
self.logger.info("collect owner&permissions")
timezone = "ALTER SESSION SET TIME_ZONE = 'US/Pacific'"
self.aw_cursor.execute(timezone)
schema = "ALTER SESSION SET CURRENT_SCHEMA=APPWORX"
self.aw_cursor.execute(schema)
user_writer = FileWriter(owner_file)
query = \
"""SELECT DISTINCT J.SO_JOB_SEQ, J.SO_MODULE, J.SO_APPLICATION, U.SO_USER_NAME FROM SO_JOB_TABLE J
JOIN SO_JOB_HISTORY H ON J.SO_JOB_SEQ = H.SO_JOB_SEQ
JOIN SO_USER_TABLE U ON H.SO_USER_SEQ = U.SO_USER_SEQ
WHERE J.SO_COMMAND_TYPE = 'CHAIN' """
self.aw_cursor.execute(query)
rows = DbUtil.dict_cursor(self.aw_cursor)
for row in rows:
record = AppworxFlowOwnerRecord(
self.app_id, row['SO_APPLICATION'] + ':' + row["SO_MODULE"],
row["SO_USER_NAME"], 'EXECUTE', 'GROUP', self.wh_exec_id)
user_writer.append(record)
user_writer.close()
def collect_flow_schedules(self, schedule_file):
self.logger.info("collect flow schedule")
schedule_writer = FileWriter(schedule_file)
query = """
SELECT DISTINCT cj.id as ref_id, cj.frequency, cj.time_unit,
unix_timestamp(cj.start_time) as start_time, unix_timestamp(cj.end_time) as end_time,
wj.app_path
FROM COORD_JOBS cj JOIN COORD_ACTIONS ca ON ca.job_id = cj.id JOIN WF_JOBS wj ON ca.external_id = wj.id
WHERE cj.status = 'RUNNING'
"""
self.oz_cursor.execute(query)
rows = DbUtil.dict_cursor(self.oz_cursor)
for row in rows:
schedule_record = OozieFlowScheduleRecord(
self.app_id, row['app_path'], row['time_unit'],
int(row['frequency']), row['start_time'], row['end_time'],
row['ref_id'], self.wh_exec_id)
schedule_writer.append(schedule_record)
schedule_writer.close()
def collect_dali_view_owner(self, file):
# dataset_urn, owner_id, sort_id, namespace, db_name, source_time
dali_prefix = "hive:///prod_tracking_views/"
namespace = "urn:li:corpuser"
db_name = "hive-nertz"
file_writer = FileWriter(file)
cmd = """
select distinct file_name, email, owner_id, last_commit_time from (
select distinct file_name, committer_email as email, trim(substring_index(committer_email, '@', 1)) owner_id, max(commit_time) last_commit_time from source_code_commit_info
where file_name like "%.hive" and repository_urn = '{git_urn}'
group by file_name, committer_email
union
select distinct file_name, author_email as email, trim(substring_index(author_email, '@', 1)) owner_id, max(commit_time) last_commit_time from source_code_commit_info
where file_name like "%.hive" and repository_urn = '{git_urn}'
group by file_name, author_email
) a where owner_id not in ({blacklist}) order by file_name, last_commit_time desc;
""".format(git_urn=self.git_urn,
blacklist=','.join('?' * len(self.owner_blacklist)))
print cmd
self.wh_cursor.execute(cmd, self.owner_blacklist)
rows = DbUtil.dict_cursor(self.wh_cursor)
prev_dataset = ""
sort_id = 0
for row in rows:
dataset_urn = dali_prefix + re.split("\.", row['file_name'])[0]
owner_id = row['owner_id']
if dataset_urn == prev_dataset:
sort_id += 1
else:
sort_id = 0
prev_dataset = dataset_urn
source_time = row['last_commit_time']
dataset_owner_record = DatasetOwnerRecord(dataset_urn, owner_id,
sort_id, namespace,
db_name, source_time)
file_writer.append(dataset_owner_record)
file_writer.close()
self.format_table_metadata(rows,schema)
end = datetime.datetime.now().strftime("%H:%M:%S")
self.logger.info("Collecting table info [%s -> %s]" % (str(begin), str(end)))
csv_columns = ['name', 'columns', 'schema_type', 'properties', 'urn', 'source', 'location_prefix',
'parent_name', 'storage_type', 'dataset_type', 'is_partitioned']
self.write_csv(table_output_file, csv_columns, self.table_output_list)
csv_columns = ['dataset_urn', 'sort_id', 'name', 'data_type', 'nullable',
'size', 'precision', 'scale', 'default_value', 'doc']
self.write_csv(field_output_file, csv_columns, self.field_output_list)
scaned_dict = {}
if sample:
open(sample_output_file, 'wb')
os.chmod(sample_output_file, 0666)
sample_file_writer = FileWriter(sample_output_file)
##writer = csv.DictWriter(csvfile, fieldnames=csv_columns, delimiter='\x1A', lineterminator='\n',
## quoting=csv.QUOTE_NONE, quotechar='\1', escapechar='\0')
self.logger.info("Writing to CSV file {}".format(sample_output_file))
# collect sample data
for onedatabase in schema:
database_name = onedatabase['database']
if 'tables' in onedatabase:
alltables = onedatabase['tables']
else:
alltables = onedatabase['views']
for onetable in alltables:
table_name = onetable['original_name'].split('.')[1]
if table_name in scaned_dict:
def run(self, database_name, table_name, schema_output_file,
sample_output_file):
"""
The entrance of the class, extract schema and sample data
Notice the database need to have a order that the databases have more info (DWH_STG) should be scaned first.
:param database_name:
:param table_name:
:param schema_output_file:
:return:
"""
cur = self.conn_td.cursor()
schema = []
f_log = open(self.log_file, "a")
schema_json = open(schema_output_file, 'wb')
os.chmod(schema_output_file, 0666)
open(sample_output_file, 'wb')
os.chmod(sample_output_file, 0666)
sample_file_writer = FileWriter(sample_output_file)
if database_name is None and table_name is None: # default route: process everything
for database_name in self.databases:
self.logger.info("Collecting tables in database : " +
database_name)
# table info
rows = []
begin = datetime.datetime.now().strftime("%H:%M:%S")
rows.extend(self.get_table_info(database_name, table_name))
if len(rows) > 0:
self.format_table_metadata(rows, schema)
end = datetime.datetime.now().strftime("%H:%M:%S")
f_log.write("Get table info %12s [%s -> %s]\n" %
(database_name, str(begin), str(end)))
# view info
rows = []
begin = datetime.datetime.now().strftime("%H:%M:%S")
rows.extend(self.get_view_info(database_name, table_name))
if len(rows) > 0:
self.format_view_metadata(rows, schema)
end = datetime.datetime.now().strftime("%H:%M:%S")
f_log.write("Get view info %12s [%s -> %s]\n" %
(database_name, str(begin), str(end)))
scaned_dict = {
} # a cache of {name : {urn : _, data : _}} to avoid repeat computing
# collect sample data
for onedatabase in schema:
database_name = onedatabase['database']
if 'tables' in onedatabase:
alltables = onedatabase['tables']
else:
alltables = onedatabase['views']
for onetable in alltables:
table_name = onetable['original_name'].split('.')[1]
if table_name in scaned_dict:
sample_record = SampleDataRecord(
'teradata', '/' + database_name + '/' + table_name,
scaned_dict[table_name]['ref_urn'],
scaned_dict[table_name]['data'])
else:
(ref_urn, sample_data) = self.get_sample_data(
database_name, table_name)
sample_record = SampleDataRecord(
'teradata', '/' + database_name + '/' + table_name,
'', sample_data)
scaned_dict[table_name] = {
'ref_urn': ref_urn,
'data': sample_data
}
sample_file_writer.append(sample_record)
sample_file_writer.close()
# print 'byte size of schema : ' + str(sys.getsizeof(schema))
schema_json.write(json.dumps(schema, indent=None) + '\n')
cur.close()
schema_json.close()
f_log.close()
return rows
if __name__ == "__main__":
args = sys.argv[1]
# connection
username = args[Constant.HIVE_METASTORE_USERNAME]
password = args[Constant.HIVE_METASTORE_PASSWORD]
jdbc_driver = args[Constant.HIVE_METASTORE_JDBC_DRIVER]
jdbc_url = args[Constant.HIVE_METASTORE_JDBC_URL]
if Constant.HIVE_DATABASE_WHITELIST_KEY in args:
database_white_list = args[Constant.HIVE_DATABASE_WHITELIST_KEY]
else:
database_white_list = ''
e = HiveExtract()
e.dataset_instance_file = args[Constant.HIVE_INSTANCE_CSV_FILE_KEY]
e.instance_writer = FileWriter(e.dataset_instance_file)
e.wh_exec_id = long(args[Constant.WH_EXEC_ID_KEY])
e.db_id = args[Constant.DB_ID_KEY]
e.conn_hms = zxJDBC.connect(jdbc_url, username, password, jdbc_driver)
try:
e.databases = e.get_all_databases(database_white_list)
e.run(args[Constant.HIVE_SCHEMA_JSON_FILE_KEY], None)
finally:
e.conn_hms.close()
e.instance_writer.close()
def collect_flow_execs(self, flow_exec_file, job_exec_file,
look_back_period):
self.logger.info(
"collect flow&job executions [last_execution_unix_time=%s lookback_period=%s]"
% (self.last_execution_unix_time, self.lookback_period))
flow_exec_writer = FileWriter(flow_exec_file)
job_exec_writer = FileWriter(job_exec_file)
timezone = "ALTER SESSION SET TIME_ZONE = 'US/Pacific'"
self.aw_cursor.execute(timezone)
schema = "ALTER SESSION SET CURRENT_SCHEMA=APPWORX"
self.aw_cursor.execute(schema)
flow_id_list = []
if self.last_execution_unix_time:
flow_cmd = \
"""SELECT J.SO_JOB_SEQ, J.SO_MODULE, J.SO_APPLICATION, H.SO_STATUS_NAME, H.SO_JOBID, H.SO_CHAIN_ID,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_STARTED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_STARTED,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_FINISHED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_FINISHED,
U.SO_USER_NAME FROM SO_JOB_TABLE J
JOIN (
SELECT * FROM SO_JOB_HISTORY WHERE SO_JOB_FINISHED >= DATE '1970-01-01' - interval '8' hour + (%d - 3600) / 86400
AND SO_CHILD_COUNT > 0
UNION ALL
SELECT * FROM SO_JOB_QUEUE WHERE SO_STATUS_NAME IN ('INITIATED', 'RUNNING', 'FINISHED')
AND SO_CHILD_COUNT > 0
) H ON J.SO_JOB_SEQ = H.SO_JOB_SEQ
LEFT JOIN SO_USER_TABLE U ON H.SO_USER_SEQ = U.SO_USER_SEQ
WHERE J.SO_COMMAND_TYPE = 'CHAIN' """ % long(self.last_execution_unix_time)
else:
flow_cmd = \
"""SELECT J.SO_JOB_SEQ, J.SO_MODULE, J.SO_APPLICATION, H.SO_STATUS_NAME, H.SO_JOBID, H.SO_CHAIN_ID,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_STARTED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_STARTED,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_FINISHED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_FINISHED,
U.SO_USER_NAME FROM SO_JOB_TABLE J
JOIN (
SELECT * FROM SO_JOB_HISTORY WHERE SO_JOB_FINISHED >= SYSDATE - %d
AND SO_CHILD_COUNT > 0
UNION ALL
SELECT * FROM SO_JOB_QUEUE WHERE SO_STATUS_NAME IN ('INITIATED', 'RUNNING', 'FINISHED')
AND SO_CHILD_COUNT > 0
) H ON J.SO_JOB_SEQ = H.SO_JOB_SEQ
LEFT JOIN SO_USER_TABLE U ON H.SO_USER_SEQ = U.SO_USER_SEQ
WHERE J.SO_COMMAND_TYPE = 'CHAIN' """ % int(self.lookback_period)
''' SO_CHAIN_ID = :flow_exec_id will find all job executions under the top level flow
select SO_EXECUTE_ORDER, SO_JOBID, SO_PARENTS_JOBID, SO_DIRECT_PARENT_JOBID, SO_CHAIN_ID
from so_job_history where SO_JOBID = SO_CHAIN_ID or SO_PARENTS_JOBID <> SO_CHAIN_ID
'''
if self.last_execution_unix_time:
job_cmd = \
"""SELECT D.SO_TASK_NAME, U.SO_USER_NAME, H.SO_STATUS_NAME, H.SO_JOBID, D.SO_DET_SEQ as JOB_ID,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_STARTED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_STARTED,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_FINISHED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_FINISHED
FROM SO_JOB_HISTORY H
JOIN SO_CHAIN_DETAIL D ON D.SO_CHAIN_SEQ = H.SO_CHAIN_SEQ AND D.SO_DET_SEQ = H.SO_DET_SEQ
LEFT JOIN SO_USER_TABLE U ON H.SO_USER_SEQ = U.SO_USER_SEQ
WHERE --H.SO_JOB_FINISHED >= DATE '1970-01-01' - interval '8' hour + (%d - 3600) / 86400) and
H.SO_CHAIN_ID = %d"""
else:
job_cmd = \
"""SELECT D.SO_TASK_NAME, U.SO_USER_NAME, H.SO_STATUS_NAME, H.SO_JOBID, D.SO_DET_SEQ as JOB_ID,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_STARTED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_STARTED,
ROUND((cast((FROM_TZ(CAST(H.SO_JOB_FINISHED as timestamp), 'US/Pacific') at time zone 'GMT') as date) -
to_date('01-JAN-1970','DD-MON-YYYY'))* (86400)) as JOB_FINISHED
FROM SO_JOB_HISTORY H
JOIN SO_CHAIN_DETAIL D ON D.SO_CHAIN_SEQ = H.SO_CHAIN_SEQ AND D.SO_DET_SEQ = H.SO_DET_SEQ
LEFT JOIN SO_USER_TABLE U ON H.SO_USER_SEQ = U.SO_USER_SEQ
WHERE H.SO_JOB_FINISHED >= SYSDATE - %d and
H.SO_CHAIN_ID = %d"""
try:
self.aw_cursor.execute(flow_cmd)
except Exception as e:
self.logger.error(e + "\n" + flow_cmd)
rows = DbUtil.dict_cursor(self.aw_cursor)
row_count = 0
for row in rows:
flow_path = row['SO_APPLICATION'] + ":" + row['SO_MODULE']
so_flow_id = row['SO_JOBID']
flow_attempt = 0
flow_exec_id = 0
try:
flow_attempt = int(
float(str(so_flow_id - int(so_flow_id))[1:]) * 100)
flow_exec_id = int(so_flow_id)
except Exception as e:
self.logger.error(e)
self.logger.debug("processing flow_exec_id: %8d" % flow_exec_id)
flow_exec_record = AppworxFlowExecRecord(
self.app_id, long(row['SO_JOB_SEQ']), row['SO_MODULE'],
flow_path, 0, flow_exec_id, row['SO_STATUS_NAME'],
flow_attempt,
row['SO_USER_NAME'] if row['SO_USER_NAME'] else '',
long(row['JOB_STARTED']),
long(row['JOB_FINISHED'] if row['JOB_FINISHED'] else 0),
self.wh_exec_id)
flow_exec_writer.append(flow_exec_record)
new_appworx_cursor = self.aw_con.cursor()
if self.last_execution_unix_time:
new_appworx_cursor.execute(
job_cmd %
(long(self.last_execution_unix_time), flow_exec_id))
else:
new_appworx_cursor.execute(
job_cmd % (int(self.lookback_period), flow_exec_id))
job_rows = DbUtil.dict_cursor(new_appworx_cursor)
for job in job_rows:
so_job_id = job['SO_JOBID']
job_attempt = 0
job_exec_id = 0
try:
job_attempt = int(
float(str(so_job_id - int(so_job_id))[1:]) * 100)
job_exec_id = int(so_job_id)
except Exception as e:
self.logger.error(e)
job_exec_record = AppworxJobExecRecord(
self.app_id, long(row['SO_JOB_SEQ']), flow_path, 0,
flow_exec_id, long(job['JOB_ID']), job['SO_TASK_NAME'],
flow_path + "/" + job['SO_TASK_NAME'],
job_exec_id, job['SO_STATUS_NAME'], job_attempt,
long(job['JOB_STARTED']), long(job['JOB_FINISHED']),
self.wh_exec_id)
job_exec_writer.append(job_exec_record)
row_count += 1
if row_count % 10000 == 0:
flow_exec_writer.flush()
job_exec_writer.flush()
flow_exec_writer.close()
job_exec_writer.close()
def transform(self, input, td_metadata, td_field_metadata):
'''
convert from json to csv
:param input: input json file
:param td_metadata: output data file for teradata metadata
:param td_field_metadata: output data file for teradata field metadata
:return:
'''
f_json = open(input)
data = json.load(f_json)
f_json.close()
schema_file_writer = FileWriter(td_metadata)
field_file_writer = FileWriter(td_field_metadata)
for d in data:
i = 0
for k in d.keys():
if k not in ['tables', 'views']:
continue
self.logger.info("%s %4d %s" %
(datetime.datetime.now().strftime("%H:%M:%S"),
len(d[k]), k))
for t in d[k]:
self.logger.info("%4d %s" % (i, t['name']))
if t['name'] == 'HDFStoTD_2464_ERR_1':
continue
i += 1
output = {}
prop_json = {}
output['name'] = t['name']
output['original_name'] = t['original_name']
prop_json["createTime"] = t["createTime"] if t.has_key(
"createTime") else None
prop_json[
"lastAlterTime"] = t["lastAlterTime"] if t.has_key(
"lastAlterTime") else None
prop_json[
"lastAccessTime"] = t["lastAccessTime"] if t.has_key(
"lastAccessTime") else None
prop_json["accessCount"] = t["accessCount"] if t.has_key(
"accessCount") else None
prop_json["sizeInMbytes"] = t["sizeInMbytes"] if t.has_key(
"sizeInMbytes") else None
if "type" in t:
prop_json["storage_type"] = t["type"]
if "partition" in t:
prop_json["partition"] = t["partition"]
if "partitions" in t:
prop_json["partitions"] = t["partitions"]
if "hashKey" in t:
prop_json["hashKey"] = t["hashKey"]
if "indices" in t:
prop_json["indices"] = t["indices"]
if "referenceTables" in t:
prop_json["referenceTables"] = t["referenceTables"]
if "viewSqlText" in t:
prop_json["viewSqlText"] = t["viewSqlText"]
output['fields'] = []
flds = {}
field_detail_list = []
sort_id = 0
for c in t['columns']:
# output['fields'].append(
# { 'name' : t['name'].encode('latin-1'),
# 'type' : None if c['data_type'] is None else c['data_type'].encode('latin-1'),
# 'attributes_json' : c}
# output['fields'][c['name'].encode('latin-1')].append({ "doc" : "", "type" : [None if c['data_type'] is None else c['data_type'].encode('latin-1')]})
sort_id += 1
output['fields'].append({
"name":
c['name'],
"doc":
'',
"type":
c['dataType'] if c['dataType'] else None,
"nullable":
c['nullable'],
"maxByteLength":
c['maxByteLength'],
"format":
c['columnFormat']
if c.has_key('columnFormat') else None,
"accessCount":
c['accessCount']
if c.has_key('accessCount') else None,
"lastAccessTime":
c['lastAccessTime']
if c.has_key("lastAccessTime") else None
})
flds[c['name']] = {
'type': c['dataType'],
"maxByteLength": c['maxByteLength']
}
field_detail_list.append([
"teradata:///%s/%s" %
(d['database'], output['name']),
str(sort_id), '0', '', c['name'], '',
c['dataType'] if 'dataType' in c
and c['dataType'] is not None else '',
str(c['maxByteLength'])
if 'maxByteLength' in c else '0',
str(c['precision']) if 'precision' in c
and c['precision'] is not None else '',
str(c['scale'])
if 'scale' in c and c['scale'] is not None else '',
c['nullable'] if 'nullable' in c
and c['nullable'] is not None else 'Y', '', '', '',
'', '', '', ''
])
dataset_scehma_record = DatasetSchemaRecord(
output['name'], json.dumps(output),
json.dumps(prop_json), json.dumps(flds),
"teradata:///%s/%s" % (d['database'], output['name']),
'Teradata', output['original_name'],
(self.convert_timestamp(t["createTime"])
if t.has_key("createTime") else None),
(self.convert_timestamp(t["lastAlterTime"])
if t.has_key("lastAlterTime") else None))
schema_file_writer.append(dataset_scehma_record)
for fields in field_detail_list:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
schema_file_writer.flush()
field_file_writer.flush()
self.logger.info("%20s contains %6d %s" %
(d['database'], i, k))
schema_file_writer.close()
field_file_writer.close()
def transform(self, input, hive_metadata, hive_field_metadata):
"""
convert from json to csv
:param input: input json file
:param hive_metadata: output data file for hive table metadata
:param hive_field_metadata: output data file for hive field metadata
:return:
"""
f_json = open(input)
all_data = json.load(f_json)
f_json.close()
schema_file_writer = FileWriter(hive_metadata)
field_file_writer = FileWriter(hive_field_metadata)
lineageInfo = LineageInfo()
# one db info : 'type', 'database', 'tables'
# one table info : required : 'name' , 'type', 'serializationFormat' ,'createTime', 'DB_ID', 'TBL_ID', 'SD_ID'
# optional : 'schemaLiteral', 'schemaUrl', 'fieldDelimiter', 'fieldList'
for one_db_info in all_data:
i = 0
for table in one_db_info['tables']:
i += 1
schema_json = {}
prop_json = {} # set the prop json
for prop_name in TableInfo.optional_prop:
if prop_name in table and table[prop_name] is not None:
prop_json[prop_name] = table[prop_name]
if TableInfo.view_expended_text in prop_json:
text = prop_json[TableInfo.view_expended_text].replace(
'`', '')
array = HiveViewDependency.getViewDependency(text)
l = []
for a in array:
l.append(a)
prop_json['view_depends_on'] = l
# process either schema
flds = {}
field_detail_list = []
if TableInfo.schema_literal in table and table[
TableInfo.schema_literal] is not None:
sort_id = 0
try:
schema_data = json.loads(
table[TableInfo.schema_literal])
except ValueError:
self.logger.error("Schema json error for table : \n" +
str(table))
schema_json = schema_data
# extract fields to field record
urn = "hive:///%s/%s" % (one_db_info['database'],
table['name'])
acp = AvroColumnParser(schema_data, urn=urn)
result = acp.get_column_list_result()
field_detail_list += result
elif TableInfo.field_list in table:
# Convert to avro
uri = "hive:///%s/%s" % (one_db_info['database'],
table['name'])
hcp = HiveColumnParser(table, urn=uri)
schema_json = {
'fields': hcp.column_type_dict['fields'],
'type': 'record',
'name': table['name'],
'uri': uri
}
field_detail_list += hcp.column_type_list
dataset_scehma_record = DatasetSchemaRecord(
table['name'], json.dumps(schema_json),
json.dumps(prop_json), json.dumps(flds),
"hive:///%s/%s" % (one_db_info['database'], table['name']),
'Hive', '', (table[TableInfo.create_time] if table.has_key(
TableInfo.create_time) else None),
(table["lastAlterTime"])
if table.has_key("lastAlterTime") else None)
schema_file_writer.append(dataset_scehma_record)
for fields in field_detail_list:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
schema_file_writer.flush()
field_file_writer.flush()
self.logger.info("%20s contains %6d tables" %
(one_db_info['database'], i))
schema_file_writer.close()
field_file_writer.close()
def collect_flow_jobs(self, flow_file, job_file, dag_file):
self.logger.info("collect flow&jobs")
query = "SELECT * FROM workflow_info WHERE status is NULL"
self.lz_cursor.execute(query)
## rows = DbUtil.dict_cursor(self.lz_cursor)
rows = DbUtil.copy_dict_cursor(self.lz_cursor)
flow_writer = FileWriter(flow_file)
job_writer = FileWriter(job_file)
dag_writer = FileWriter(dag_file)
row_count = 0
for row in rows:
self.logger.info("collect flow %d!" % row_count)
flow_path = row['project_name'] + ":" + row['workflow_name']
flow_record = LhotseFlowRecord(
self.app_id, row['workflow_name'], row['project_name'],
flow_path, 0, int(time.mktime(row['create_time'].timetuple())),
int(time.mktime(row['modify_time'].timetuple())), 0, 'Y',
self.wh_exec_id)
## for debug
## self.logger.info("the flow record is: %s" % flow_record.toCsvString())
flow_writer.append(flow_record)
# get relative task of this workflow.
task_query = "SELECT * FROM task_info WHERE workflow_id = \"{0}\"".format(
row['workflow_id'])
new_lz_cursor = self.lz_cursor
new_lz_cursor.execute(task_query)
task_rows = DbUtil.dict_cursor(new_lz_cursor)
for task in task_rows:
job_record = LhotseJobRecord(
self.app_id, flow_path, 0, task['task_name'],
flow_path + '/' + task['task_name'],
task['task_type_name'], 'Y', self.wh_exec_id)
job_writer.append(job_record)
# task bridge
# bridge's status need to be considered in the next stage
task_bridge_query = "SELECT * FROM task_bridge WHERE workflow_id = \"{0}\"".format(
row['workflow_id'])
self.lz_cursor.execute(task_bridge_query)
# task_bridge_rows = DbUtil.dict_cursor(self.lz_cursor)
task_bridge_rows = DbUtil.copy_dict_cursor(self.lz_cursor)
for bridge in task_bridge_rows:
origin_task_query = "SELECT task_name FROM task_info WHERE task_id = \"{0}\"".format(
bridge['origin_id'])
self.lz_cursor.execute(origin_task_query)
origin_tasks = self.lz_cursor.fetchone()
target_task_query = "SELECT task_name FROM task_info WHERE task_id = \"{0}\"".format(
bridge['target_id'])
self.lz_cursor.execute(target_task_query)
target_tasks = self.lz_cursor.fetchone()
dag_edge = LhotseFlowDagRecord(
self.app_id, flow_path, 0,
flow_path + '/' + origin_tasks[0],
flow_path + '/' + target_tasks[0], self.wh_exec_id)
dag_writer.append(dag_edge)
row_count += 1
if row_count % 1000 == 0:
flow_writer.flush()
job_writer.flush()
dag_writer.flush()
flow_writer.close()
job_writer.close()
dag_writer.close()
def transform(self, input, hive_metadata, hive_field_metadata):
"""
convert from json to csv
:param input: input json file
:param hive_metadata: output data file for hive table metadata
:param hive_field_metadata: output data file for hive field metadata
:return:
"""
f_json = open(input)
all_data = json.load(f_json)
f_json.close()
schema_file_writer = FileWriter(hive_metadata)
field_file_writer = FileWriter(hive_field_metadata)
lineageInfo = LineageInfo()
# one db info : 'type', 'database', 'tables'
# one table info : required : 'name' , 'type', 'serializationFormat' ,'createTime', 'DB_ID', 'TBL_ID', 'SD_ID'
# optional : 'schemaLiteral', 'schemaUrl', 'fieldDelimiter', 'fieldList'
for one_db_info in all_data:
i = 0
for table in one_db_info['tables']:
i += 1
schema_json = {}
prop_json = {} # set the prop json
for prop_name in TableInfo.optional_prop:
if prop_name in table and table[prop_name] is not None:
prop_json[prop_name] = table[prop_name]
if TableInfo.view_expended_text in prop_json:
text = prop_json[TableInfo.view_expended_text].replace(
'`', '')
array = HiveViewDependency.getViewDependency(text)
l = []
for a in array:
l.append(a)
prop_json['view_depends_on'] = l
# process either schema
flds = {}
field_detail_list = []
if TableInfo.schema_literal in table and table[
TableInfo.schema_literal] is not None:
sort_id = 0
try:
schema_data = json.loads(
table[TableInfo.schema_literal])
except ValueError:
self.logger.error("Schema json error for table : \n" +
str(table))
schema_json = schema_data
# process each field
for field in schema_data['fields']:
field_name = field['name']
type = field['type'] # could be a list
default_value = field[
'default'] if 'default' in field else None
doc = field['doc'] if 'doc' in field else None
attributes_json = json.loads(
field['attributes_json']
) if 'attributes_json' in field else None
pk = delta = is_nullable = is_indexed = is_partitioned = inside_type = format = data_size = None
if attributes_json:
pk = attributes_json[
'pk'] if 'pk' in attributes_json else None
delta = attributes_json[
'delta'] if 'delta' in attributes_json else None
is_nullable = attributes_json[
'nullable'] if 'nullable' in attributes_json else None
inside_type = attributes_json[
'type'] if 'type' in attributes_json else None
format = attributes_json[
'format'] if 'format' in attributes_json else None
flds[field_name] = {'type': type}
# String urn, Integer sortId, Integer parentSortId, String parentPath, String fieldName,
#String dataType, String isNullable, String defaultValue, Integer dataSize, String namespace, String description
sort_id += 1
field_detail_list.append([
"hive:///%s/%s" %
(one_db_info['database'], table['name']),
str(sort_id), '0', None, field_name, '', type,
data_size, None, None, is_nullable, is_indexed,
is_partitioned, default_value, None,
json.dumps(attributes_json)
])
elif TableInfo.field_list in table:
schema_json = {
'type': 'record',
'name': table['name'],
'fields': table[TableInfo.field_list]
} # construct a schema for data came from COLUMN_V2
for field in table[TableInfo.field_list]:
field_name = field['ColumnName']
type = field['TypeName']
# ColumnName, IntegerIndex, TypeName, Comment
flds[field_name] = {'type': type}
pk = delta = is_nullable = is_indexed = is_partitioned = inside_type = format = data_size = default_value = None # TODO ingest
field_detail_list.append([
"hive:///%s/%s" %
(one_db_info['database'], table['name']),
field['IntegerIndex'], '0', None, field_name, '',
field['TypeName'], None, None, None, is_nullable,
is_indexed, is_partitioned, default_value, None,
None
])
dataset_scehma_record = DatasetSchemaRecord(
table['name'], json.dumps(schema_json),
json.dumps(prop_json), json.dumps(flds),
"hive:///%s/%s" % (one_db_info['database'], table['name']),
'Hive', '', (table[TableInfo.create_time] if table.has_key(
TableInfo.create_time) else None),
(table["lastAlterTime"])
if table.has_key("lastAlterTime") else None)
schema_file_writer.append(dataset_scehma_record)
for fields in field_detail_list:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
schema_file_writer.flush()
field_file_writer.flush()
self.logger.info("%20s contains %6d tables" %
(one_db_info['database'], i))
schema_file_writer.close()
field_file_writer.close()
def collect_flow_jobs(self, flow_file, job_file, dag_file):
self.logger.info(
"collect flow&jobs [last_execution_unix_time=%s lookback_period=%s]"
% (self.last_execution_unix_time, self.lookback_period))
timezone = "ALTER SESSION SET TIME_ZONE = 'US/Pacific'"
self.aw_cursor.execute(timezone)
schema = "ALTER SESSION SET CURRENT_SCHEMA=APPWORX"
self.aw_cursor.execute(schema)
if self.last_execution_unix_time:
time_filter = "(DATE '1970-01-01' - INTERVAL '8' HOUR) + (%d - 3600) / 86400" % long(
self.last_execution_unix_time)
else:
time_filter = "SYSDATE - %d" % int(self.lookback_period)
flow_query = \
"""SELECT J.SO_JOB_SEQ, J.SO_APPLICATION, J.SO_MODULE, R.LAST_CHAIN_ID
FROM SO_JOB_TABLE J JOIN (
SELECT SO_JOB_SEQ, MAX(SO_CHAIN_ID) as LAST_CHAIN_ID
FROM
( SELECT SO_JOB_SEQ, SO_CHAIN_ID FROM SO_JOB_HISTORY
WHERE SO_JOB_FINISHED >= %s
AND SO_CHILD_COUNT > 0
UNION ALL
SELECT SO_JOB_SEQ, SO_CHAIN_ID FROM SO_JOB_QUEUE
WHERE SO_STATUS_NAME IN ('INITIATED', 'RUNNING', 'FINISHED')
AND SO_CHILD_COUNT > 0
)
GROUP BY SO_JOB_SEQ
) R ON J.SO_JOB_SEQ = R.SO_JOB_SEQ
WHERE SO_COMMAND_TYPE = 'CHAIN'
ORDER BY 2,3
""" % time_filter
job_query = \
"""SELECT d.SO_TASK_NAME, d.SO_CHAIN_ORDER, d.SO_PREDECESSORS as PREDECESSORS, d.SO_DET_SEQ as JOB_ID,
t.* FROM SO_CHAIN_DETAIL d
JOIN SO_JOB_TABLE t ON d.SO_JOB_SEQ = t.SO_JOB_SEQ
WHERE d.SO_CHAIN_SEQ = %d
ORDER BY d.SO_CHAIN_ORDER
"""
self.aw_cursor.execute(flow_query)
rows = DbUtil.dict_cursor(self.aw_cursor)
flow_writer = FileWriter(flow_file)
job_writer = FileWriter(job_file)
dag_writer = FileWriter(dag_file)
row_count = 0
for row in rows:
flow_path = row['SO_APPLICATION'] + ":" + row['SO_MODULE']
flow_record = AppworxFlowRecord(self.app_id,
long(row['SO_JOB_SEQ']),
row['SO_MODULE'],
row['SO_APPLICATION'], flow_path,
0, 0, 0, 'Y', self.wh_exec_id)
flow_writer.append(flow_record)
new_appworx_cursor = self.aw_con.cursor()
new_appworx_cursor.execute(job_query % row['SO_JOB_SEQ'])
job_rows = DbUtil.dict_cursor(new_appworx_cursor)
for job in job_rows:
job_record = AppworxJobRecord(
self.app_id, long(row['SO_JOB_SEQ']), flow_path, 0,
long(job['JOB_ID']), job['SO_TASK_NAME'],
flow_path + '/' + job['SO_TASK_NAME'], job['SO_MODULE'],
'Y', self.wh_exec_id)
command_type = job['SO_COMMAND_TYPE']
if command_type and command_type == 'CHAIN':
job_record.setRefFlowPath(job['SO_APPLICATION'] + ":" +
job['SO_MODULE'])
job_record.setJobType('CHAIN')
job_writer.append(job_record)
predecessors_str = job['PREDECESSORS']
if predecessors_str:
predecessors = re.findall(r"\&\/(.+?)\s\=\sS",
predecessors_str)
if predecessors:
for predecessor in predecessors:
dag_edge = AppworxFlowDagRecord(
self.app_id, long(row['SO_JOB_SEQ']),
flow_path, 0, flow_path + '/' + predecessor,
flow_path + '/' + job['SO_TASK_NAME'],
self.wh_exec_id)
dag_writer.append(dag_edge)
row_count += 1
if row_count % 1000 == 0:
flow_writer.flush()
job_writer.flush()
dag_writer.flush()
flow_writer.close()
job_writer.close()
dag_writer.close()
def __init__(self):
self.logger = LoggerFactory.getLogger('jython script : ' +
self.__class__.__name__)
self.base_url = args[Constant.BASE_URL_KEY]
self.code_search_committer_writer = FileWriter(
args[Constant.DATABASE_SCM_REPO_OUTPUT_KEY])
def transform(self, raw_metadata, metadata_output, field_metadata_output):
# sys.setdefaultencoding("UTF-8")
input_json_file = open(raw_metadata, 'r')
schema_file_writer = FileWriter(metadata_output)
field_file_writer = FileWriter(field_metadata_output)
i = 0
self.sort_id = 0
o_urn = ''
p = ''
def fields_json_to_csv(output_list_, parent_field_path, field_list_):
# string, list, int, optional int
self.sort_id
parent_field_path
parent_id = self.sort_id
for f in field_list_:
self.sort_id += 1
o_field_name = f['name']
o_field_data_type = ''
o_field_data_size = None
o_field_nullable = 'N'
o_field_default = ''
o_field_namespace = ''
o_field_doc = ''
effective_type_index_in_type = -1
if f.has_key('namespace'):
o_field_namespace = f['namespace']
if f.has_key('default') and type(f['default']) != None:
o_field_default = f['default']
if not f.has_key('type'):
o_field_data_type = None
elif type(f['type']) == list:
i = effective_type_index = -1
for data_type in f['type']:
i += 1 # current index
if type(data_type) is None or (data_type == 'null'):
o_field_nullable = 'Y'
elif type(data_type) == dict:
o_field_data_type = data_type['type']
effective_type_index_in_type = i
if data_type.has_key('namespace'):
o_field_namespace = data_type['namespace']
elif data_type.has_key('name'):
o_field_namespace = data_type['name']
if data_type.has_key('size'):
o_field_data_size = data_type['size']
else:
o_field_data_size = None
else:
o_field_data_type = data_type
effective_type_index_in_type = i
elif type(f['type']) == dict:
o_field_data_type = f['type']['type']
else:
o_field_data_type = f['type']
if f.has_key('attributes') and f['attributes'].has_key(
'nullable'):
o_field_nullable = 'Y' if f['attributes'][
'nullable'] else 'N'
if f.has_key('attributes') and f['attributes'].has_key(
'size'):
o_field_data_size = f['attributes']['size']
if f.has_key('doc'):
if len(f['doc']) == 0 and f.has_key('attributes'):
o_field_doc = json.dumps(f['attributes'])
else:
o_field_doc = f['doc']
elif f.has_key('comment'):
o_field_doc = f['comment']
output_list_.append([
o_urn, self.sort_id, parent_id, parent_field_path,
o_field_name, o_field_data_type, o_field_nullable,
o_field_default, o_field_data_size, o_field_namespace,
o_field_doc.replace("\n", ' ')
if o_field_doc is not None else None
])
# check if this field is a nested record
if type(f['type']) == dict and f['type'].has_key('fields'):
current_field_path = o_field_name if parent_field_path == '' else parent_field_path + '.' + o_field_name
fields_json_to_csv(output_list_, current_field_path,
f['type']['fields'])
elif type(f['type']) == dict and f['type'].has_key(
'items') and type(
f['type']['items']
) == dict and f['type']['items'].has_key('fields'):
current_field_path = o_field_name if parent_field_path == '' else parent_field_path + '.' + o_field_name
fields_json_to_csv(output_list_, current_field_path,
f['type']['items']['fields'])
if effective_type_index_in_type >= 0 and type(
f['type'][effective_type_index_in_type]) == dict:
if f['type'][effective_type_index_in_type].has_key(
'items') and type(
f['type'][effective_type_index_in_type]
['items']) == list:
for item in f['type'][effective_type_index_in_type][
'items']:
if type(item) == dict and item.has_key('fields'):
current_field_path = o_field_name if parent_field_path == '' else parent_field_path + '.' + o_field_name
fields_json_to_csv(output_list_,
current_field_path,
item['fields'])
elif f['type'][effective_type_index_in_type].has_key(
'items') and f['type'][
effective_type_index_in_type]['items'].has_key(
'fields'):
# type: [ null, { type: array, items: { name: xxx, type: record, fields: [] } } ]
current_field_path = o_field_name if parent_field_path == '' else parent_field_path + '.' + o_field_name
fields_json_to_csv(
output_list_, current_field_path, f['type']
[effective_type_index_in_type]['items']['fields'])
elif f['type'][effective_type_index_in_type].has_key(
'fields'):
# if f['type'][effective_type_index_in_type].has_key('namespace'):
# o_field_namespace = f['type'][effective_type_index_in_type]['namespace']
current_field_path = o_field_name if parent_field_path == '' else parent_field_path + '.' + o_field_name
fields_json_to_csv(
output_list_, current_field_path,
f['type'][effective_type_index_in_type]['fields'])
# End of function
for line in input_json_file:
try:
j = json.loads(line)
except:
self.logger.error(" Invalid JSON:\n%s" % line)
continue
i += 1
o_field_list_ = []
parent_field_path = ''
self.sort_id = 0
if not (j.has_key('attributes_json') or j.has_key('attributes')):
o_properties = {"doc": null}
else:
o_properties = {}
if j.has_key('attributes_json'):
o_properties = json.loads(j['attributes_json'])
del j['attributes_json']
if j.has_key('attributes'):
o_properties = dict(j['attributes'].items() +
o_properties.items())
del j['attributes']
if j.has_key('uri'):
o_urn = j['uri']
elif o_properties.has_key('uri'):
o_urn = o_properties['uri']
else:
self.logger.info('*** Warning: "uri" is not found in %s' %
j['name'])
o_urn = ''
if o_urn.find('hdfs://') == 0:
o_name = o_urn[o_urn.rfind('/') + 1:]
elif o_properties.has_key('table_name'):
o_name = o_properties['table_name']
elif j.has_key('name') and j['name'][0:5] != 'TUPLE':
o_name = j['name']
else:
o_name = o_urn[o_urn.rfind('/') + 1:]
if j.has_key('id') or not j.has_key('fields'): # esWritable schema
o_fields = {}
for k in j:
if not (k == 'uri' or k == 'attributes' or k == 'doc'):
if type(j[k]) == list:
o_fields[k] = {
"name": k,
"type": 'list',
"doc": str(j[k])
}
elif type(j[k]) == dict:
o_fields[k] = {
"name": k,
"type": 'dict',
"doc": str(j[k])
}
else:
o_fields[k] = {
"name": k,
"type": j[k],
"doc": None
}
self.sort_id += 1
o_field_list_.append([
o_urn, self.sort_id, 0, '', k, o_fields[k]['type'],
'', '', '', o_fields[k]['doc'].replace("\n", ' ')
if o_fields[k]['doc'] is not None else None
])
elif j.has_key('fields'):
o_fields = {}
for f in j['fields']:
o_field_name = f['name']
o_fields[o_field_name] = dict(f) # for schema output
if f.has_key('attributes_json'):
f['attributes'] = json.loads(f['attributes_json'])
del f['attributes_json']
fields_json_to_csv(o_field_list_, '', j['fields'])
else:
o_fields = {"doc": None}
if j.has_key('attributes') and not o_properties.has_key('source'):
o_properties['source'] = j['attributes']['source']
if o_urn.startswith(
'hdfs:///') and self.file_regex_source_map is not None:
o_source = self.get_source(o_urn[7:])
else:
self.logger.warn(
"property : " + Constant.HDFS_FILE_SOURCE_MAP_KEY +
" is None, will use default source for all dataset")
o_source = 'Hdfs'
self.logger.info(
"%4i (%6i): %4i fields, %4i total fields(including nested) found in [%s]@%s with source %s"
% (i, len(j), len(o_fields), len(o_field_list_), o_name, o_urn,
o_source))
dataset_schema_record = DatasetSchemaRecord(
o_name, json.dumps(j, sort_keys=True),
json.dumps(o_properties, sort_keys=True), json.dumps(o_fields),
o_urn, o_source, None, None, None)
schema_file_writer.append(dataset_schema_record)
for fields in o_field_list_:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
schema_file_writer.close()
field_file_writer.close()
input_json_file.close()
def collect_flow_jobs(self, flow_file, job_file, dag_file):
self.logger.info("collect flow&jobs")
query = "SELECT distinct f.*, p.name as project_name FROM project_flows f inner join projects p on f.project_id = p.id and f.version = p.version where p.active = 1"
self.az_cursor.execute(query)
rows = DbUtil.dict_cursor(self.az_cursor)
flow_writer = FileWriter(flow_file)
job_writer = FileWriter(job_file)
dag_writer = FileWriter(dag_file)
row_count = 0
for row in rows:
row['version'] = 0 if (row["version"] is None) else row["version"]
json_column = 'json'
unzipped_content = gzip.GzipFile(
mode='r',
fileobj=StringIO.StringIO(row[json_column].tostring())).read()
try:
row[json_column] = json.loads(unzipped_content)
except:
pass
flow_path = row['project_name'] + ":" + row['flow_id']
flow_record = AzkabanFlowRecord(self.app_id, row['flow_id'],
row['project_name'], flow_path, 0,
row['modified_time'] / 1000,
row["version"], 'Y',
self.wh_exec_id)
flow_writer.append(flow_record)
# get flow jobs
nodes = row[json_column]['nodes']
for node in nodes:
job_record = AzkabanJobRecord(self.app_id, flow_path,
row["version"], node['id'],
flow_path + '/' + node['id'],
node['jobType'], 'Y',
self.wh_exec_id)
if node['jobType'] == 'flow':
job_record.setRefFlowPath(row['project_name'] + ":" +
node['embeddedFlowId'])
job_writer.append(job_record)
# job dag
edges = row[json_column]['edges']
for edge in edges:
dag_edge = AzkabanFlowDagRecord(
self.app_id, flow_path, row['version'],
flow_path + '/' + edge['source'],
flow_path + '/' + edge['target'], self.wh_exec_id)
dag_writer.append(dag_edge)
row_count += 1
if row_count % 1000 == 0:
flow_writer.flush()
job_writer.flush()
dag_writer.flush()
flow_writer.close()
job_writer.close()
dag_writer.close()
def transform(self, input, hive_instance, hive_metadata,
hive_field_metadata):
"""
convert from json to csv
:param input: input json file
:param hive_metadata: output data file for hive table metadata
:param hive_field_metadata: output data file for hive field metadata
:return:
"""
all_data = []
with open(input) as input_file:
for line in input_file:
all_data.append(json.loads(line))
dataset_idx = -1
instance_file_writer = FileWriter(hive_instance)
schema_file_writer = FileWriter(hive_metadata)
field_file_writer = FileWriter(hive_field_metadata)
lineageInfo = LineageInfo()
depends_sql = """
SELECT d.NAME DB_NAME, case when t.TBL_NAME regexp '_[0-9]+_[0-9]+_[0-9]+$'
then concat(substring(t.TBL_NAME, 1, length(t.TBL_NAME) - length(substring_index(t.TBL_NAME, '_', -3)) - 1),'_{version}')
else t.TBL_NAME
end dataset_name,
concat('/', d.NAME, '/', t.TBL_NAME) object_name,
case when (d.NAME like '%\_mp' or d.NAME like '%\_mp\_versioned') and d.NAME not like 'dalitest%' and t.TBL_TYPE = 'VIRTUAL_VIEW'
then 'dalids'
else 'hive'
end object_type,
case when (d.NAME like '%\_mp' or d.NAME like '%\_mp\_versioned') and d.NAME not like 'dalitest%' and t.TBL_TYPE = 'VIRTUAL_VIEW'
then 'View'
else
case when LOCATE('view', LOWER(t.TBL_TYPE)) > 0 then 'View'
when LOCATE('index', LOWER(t.TBL_TYPE)) > 0 then 'Index'
else 'Table'
end
end object_sub_type,
case when (d.NAME like '%\_mp' or d.NAME like '%\_mp\_versioned') and t.TBL_TYPE = 'VIRTUAL_VIEW'
then 'dalids'
else 'hive'
end prefix
FROM TBLS t JOIN DBS d on t.DB_ID = d.DB_ID
WHERE d.NAME = '{db_name}' and t.TBL_NAME = '{table_name}'
"""
# one db info : 'type', 'database', 'tables'
# one table info : required : 'name' , 'type', 'serializationFormat' ,'createTime', 'DB_ID', 'TBL_ID', 'SD_ID'
# optional : 'schemaLiteral', 'schemaUrl', 'fieldDelimiter', 'fieldList'
for one_db_info in all_data:
i = 0
for table in one_db_info['tables']:
i += 1
schema_json = {}
prop_json = {} # set the prop json
for prop_name in TableInfo.optional_prop:
if prop_name in table and table[prop_name] is not None:
prop_json[prop_name] = table[prop_name]
view_expanded_text = ''
if TableInfo.view_expended_text in prop_json:
view_expanded_text = prop_json[
TableInfo.view_expended_text]
text = prop_json[TableInfo.view_expended_text].replace(
'`',
'') # this will be fixed after switching to Hive AST
array = []
try:
array = HiveViewDependency.getViewDependency(text)
except:
self.logger.error(
"HiveViewDependency.getViewDependency(%s) failed!"
% (table['name']))
l = []
for a in array:
l.append(a)
names = str(a).split('.')
if names and len(names) >= 2:
db_name = names[0].lower()
table_name = names[1].lower()
if db_name and table_name:
self.curs.execute(
depends_sql.format(db_name=db_name,
table_name=table_name,
version='{version}'))
rows = self.curs.fetchall()
self.conn_hms.commit()
if rows and len(rows) > 0:
for row_index, row_value in enumerate(
rows):
dependent_record = HiveDependencyInstanceRecord(
one_db_info['type'], table['type'],
"/%s/%s" %
(one_db_info['database'],
table['name']), 'dalids:///' +
one_db_info['database'] + '/' +
table['dataset_name']
if one_db_info['type'].lower()
== 'dalids' else 'hive:///' +
one_db_info['database'] + '/' +
table['dataset_name'],
'depends on', 'Y', row_value[3],
row_value[4], row_value[2],
row_value[5] + ':///' +
row_value[0] + '/' + row_value[1],
'')
self.instance_writer.append(
dependent_record)
prop_json['view_depends_on'] = l
self.instance_writer.flush()
# process either schema
flds = {}
field_detail_list = []
if TableInfo.schema_literal in table and \
table[TableInfo.schema_literal] is not None and \
table[TableInfo.schema_literal].startswith('{'):
sort_id = 0
urn = "hive:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
self.logger.info("Getting schema literal for: %s" % (urn))
try:
schema_data = json.loads(
table[TableInfo.schema_literal])
schema_json = schema_data
acp = AvroColumnParser(schema_data, urn=urn)
result = acp.get_column_list_result()
field_detail_list += result
except ValueError:
self.logger.error(
"Schema Literal JSON error for table: " +
str(table))
elif TableInfo.field_list in table:
# Convert to avro
uri = "hive:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
if one_db_info['type'].lower() == 'dalids':
uri = "dalids:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
else:
uri = "hive:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
self.logger.info("Getting column definition for: %s" %
(uri))
try:
hcp = HiveColumnParser(table, urn=uri)
schema_json = {
'fields': hcp.column_type_dict['fields'],
'type': 'record',
'name': table['name'],
'uri': uri
}
field_detail_list += hcp.column_type_list
except:
self.logger.error("HiveColumnParser(%s) failed!" %
(uri))
schema_json = {
'fields': {},
'type': 'record',
'name': table['name'],
'uri': uri
}
if one_db_info['type'].lower() == 'dalids':
dataset_urn = "dalids:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
else:
dataset_urn = "hive:///%s/%s" % (one_db_info['database'],
table['dataset_name'])
dataset_instance_record = DatasetInstanceRecord(
'dalids:///' + one_db_info['database'] + '/' +
table['name'] if one_db_info['type'].lower() == 'dalids'
else 'hive:///' + one_db_info['database'] + '/' +
table['name'], 'grid', '', '', '*', 0,
table['native_name'], table['logical_name'],
table['version'], table['create_time'],
json.dumps(schema_json), json.dumps(view_expanded_text),
dataset_urn)
instance_file_writer.append(dataset_instance_record)
if dataset_urn not in self.dataset_dict:
dataset_scehma_record = DatasetSchemaRecord(
table['dataset_name'], json.dumps(schema_json),
json.dumps(prop_json), json.dumps(flds), dataset_urn,
'Hive', one_db_info['type'], table['type'], '',
table.get(TableInfo.create_time),
(int(table.get(TableInfo.source_modified_time, "0"))))
schema_file_writer.append(dataset_scehma_record)
dataset_idx += 1
self.dataset_dict[dataset_urn] = dataset_idx
for fields in field_detail_list:
field_record = DatasetFieldRecord(fields)
field_file_writer.append(field_record)
instance_file_writer.flush()
schema_file_writer.flush()
field_file_writer.flush()
self.logger.info("%20s contains %6d tables" %
(one_db_info['database'], i))
instance_file_writer.close()
schema_file_writer.close()
field_file_writer.close()
