def run_coalescer(cfg: dict,
tables: List[str],
periodstr: str,
run_once: bool,
logger: Logger,
no_sqpoller: bool = False) -> None:
"""Run the coalescer.
Runs it once and returns or periodically depending on the
value of run_once. It also writes out the coalescer records
as a parquet file.
:param cfg: dict, the Suzieq config file read in
:param tables: List[str], list of table names to coalesce
:param periodstr: str, the string of how periodically the poller runs,
Examples are '1h', '1d' etc.
:param run_once: bool, True if you want the poller to run just once
:param logger: logging.Logger, the logger to write logs to
:param no_sqpoller: bool, write records even when there's no sqpoller rec
:returns: Nothing
:rtype: none
"""
try:
schemas = Schema(cfg['schema-directory'])
except Exception as ex:
logger.error(f'Aborting. Unable to load schema: {str(ex)}')
print(f'ERROR: Aborting. Unable to load schema: {str(ex)}')
sys.exit(1)
coalescer_schema = SchemaForTable('sqCoalescer', schemas)
pqdb = get_sqdb_engine(cfg, 'sqCoalescer', None, logger)
status, errmsg = validate_periodstr(periodstr)
if not status:
logger.error(errmsg)
print(f'ERROR: {errmsg}')
sys.exit(1)
while True:
try:
stats = do_coalesce(cfg, tables, periodstr, logger, no_sqpoller)
except Exception:
logger.exception('Coalescer aborted. Continuing')
# Write the selftats
if stats:
df = pd.DataFrame([asdict(x) for x in stats])
if not df.empty:
df['sqvers'] = coalescer_schema.version
df['version'] = SUZIEQ_VERSION
df['active'] = True
df['namespace'] = ''
pqdb.write('sqCoalescer', 'pandas', df, True,
coalescer_schema.get_arrow_schema(), None)
if run_once:
break
sleep_time = get_sleep_time(periodstr)
sleep(sleep_time)
def __init__(self, baseobj):
self.ctxt = baseobj.ctxt
self.iobj = baseobj
self.summary_row_order = []
self._summarize_on_add_field = []
self._summarize_on_add_with_query = []
self._summarize_on_add_list_or_count = []
self._summarize_on_add_stat = []
self._summarize_on_perdevice_stat = []
self._dbeng = get_sqdb_engine(baseobj.ctxt.cfg, baseobj.table, '',
None)
def __init__(self, baseobj):
self.ctxt = baseobj.ctxt
self.iobj = baseobj
self.summary_row_order = []
self.nsgrp = None
self.ns_df = pd.DataFrame()
self.ns = []
self.summary_df = pd.DataFrame()
self._summarize_on_add_field = []
self._summarize_on_add_with_query = []
self._summarize_on_add_list_or_count = []
self._summarize_on_add_stat = []
self._summarize_on_perdevice_stat = []
self._check_empty_col = 'namespace'
self._dbeng = get_sqdb_engine(baseobj.ctxt.cfg, baseobj.table, '',
None)
def test_transform(input_file):
to_transform = Yaml2Class(input_file)
try:
data_directory = to_transform.transform.data_directory
except AttributeError:
print('Invalid transformation file, no data directory')
pytest.fail('AttributeError', pytrace=True)
# Make a copy of the data directory
temp_dir, tmpfile = _coalescer_init(data_directory)
cfg = load_sq_config(config_file=tmpfile.name)
schemas = Schema(cfg['schema-directory'])
for ele in to_transform.transform.transform:
query_str_list = []
# Each transformation has a record => write's happen per record
for record in ele.record:
changed_fields = set()
new_df = pd.DataFrame()
tables = [x for x in dir(record) if not x.startswith('_')]
for table in tables:
# Lets read the data in now that we know the table
tblobj = get_sqobject(table)
pq_db = get_sqdb_engine(cfg, table, None, None)
columns = schemas.fields_for_table(table)
mod_df = tblobj(config_file=tmpfile.name).get(columns=columns)
for key in getattr(record, table):
query_str = key.match
chg_df = pd.DataFrame()
if query_str != "all":
try:
chg_df = mod_df.query(query_str) \
.reset_index(drop=True)
except Exception as ex:
assert (not ex)
query_str_list.append(query_str)
else:
chg_df = mod_df
_process_transform_set(key.set, chg_df, changed_fields)
if new_df.empty:
new_df = chg_df
elif not chg_df.empty:
new_df = pd.concat([new_df, chg_df])
if new_df.empty:
continue
# Write the records now
_write_verify_transform(new_df, table, pq_db,
SchemaForTable(table,
schemas), tmpfile.name,
query_str_list, changed_fields)
# Now we coalesce and verify it works
from suzieq.sqobjects.tables import TablesObj
pre_table_df = TablesObj(config_file=tmpfile.name).get()
do_coalesce(cfg, None)
_verify_coalescing(temp_dir)
post_table_df = TablesObj(config_file=tmpfile.name).get()
assert_df_equal(pre_table_df, post_table_df, None)
# Run additional tests on the coalesced data
for ele in to_transform.transform.verify:
table = [x for x in dir(ele) if not x.startswith('_')][0]
tblobj = get_sqobject(table)
for tst in getattr(ele, table):
start_time = tst.test.get('start-time', '')
end_time = tst.test.get('end-time', '')
columns = tst.test.get('columns', ['default'])
df = tblobj(config_file=tmpfile.name,
start_time=start_time,
end_time=end_time).get(columns=columns)
if not df.empty and 'query' in tst.test:
query_str = tst.test['query']
df = df.query(query_str).reset_index(drop=True)
if 'assertempty' in tst.test:
assert (df.empty)
elif 'shape' in tst.test:
shape = tst.test['shape'].split()
if shape[0] != '*':
assert (int(shape[0]) == df.shape[0])
if shape[1] != '*':
assert (int(shape[1]) == df.shape[1])
else:
assert (not df.empty)
_coalescer_cleanup(temp_dir, tmpfile)
