def test_hive_udfs_missing_jar(self, vector):
""" IMPALA-2365: Impalad shouldn't crash if the udf jar isn't present
on HDFS"""
# Copy hive-exec.jar to a temporary file
jar_path = "tmp/" + get_random_id(5) + ".jar"
self.hdfs_client.copy('test-warehouse/hive-exec.jar', jar_path)
drop_fn_stmt = "drop function if exists default.pi_missing_jar()"
create_fn_stmt = "create function default.pi_missing_jar() returns double \
location '/%s' symbol='org.apache.hadoop.hive.ql.udf.UDFPI'" % jar_path
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
client = impalad.service.create_beeswax_client()
# Create and drop functions with sync_ddl to make sure they are reflected
# in every impalad.
exec_option = vector.get_value('exec_option')
exec_option['sync_ddl'] = 1
self.execute_query_expect_success(client, drop_fn_stmt, exec_option)
self.execute_query_expect_success(client, create_fn_stmt, exec_option)
# Delete the udf jar
self.hdfs_client.delete_file_dir(jar_path)
different_impalad = cluster.get_different_impalad(impalad)
client = different_impalad.service.create_beeswax_client()
# Run a query using the udf from an impalad other than the one
# we used to create the function. This is to bypass loading from
# the cache
try:
self.execute_query_using_client(client,
"select default.pi_missing_jar()", vector)
assert False, "Query expected to fail"
except ImpalaBeeswaxException, e:
assert "Failed to get file info" in str(e)
def test_hive_udfs_missing_jar(self, vector):
""" IMPALA-2365: Impalad shouldn't crash if the udf jar isn't present
on HDFS"""
# Copy hive-exec.jar to a temporary file
jar_path = "tmp/" + get_random_id(5) + ".jar"
self.hdfs_client.copy('test-warehouse/hive-exec.jar', jar_path)
drop_fn_stmt = "drop function if exists default.pi_missing_jar()"
create_fn_stmt = "create function default.pi_missing_jar() returns double \
location '/%s' symbol='org.apache.hadoop.hive.ql.udf.UDFPI'" % jar_path
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
client = impalad.service.create_beeswax_client()
# Create and drop functions with sync_ddl to make sure they are reflected
# in every impalad.
exec_option = vector.get_value('exec_option')
exec_option['sync_ddl'] = 1
self.execute_query_expect_success(client, drop_fn_stmt, exec_option)
self.execute_query_expect_success(client, create_fn_stmt, exec_option)
# Delete the udf jar
self.hdfs_client.delete_file_dir(jar_path)
different_impalad = cluster.get_different_impalad(impalad)
client = different_impalad.service.create_beeswax_client()
# Run a query using the udf from an impalad other than the one
# we used to create the function. This is to bypass loading from
# the cache
try:
self.execute_query_using_client(client,
"select default.pi_missing_jar()",
vector)
assert False, "Query expected to fail"
except ImpalaBeeswaxException, e:
assert "Failed to get file info" in str(e)
def test_query_profile_encoded_unknown_query_id(self):
"""Test that /query_profile_encoded error message starts with the expected line in
case of missing query and does not contain any leading whitespace.
"""
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
result = impalad.service.read_debug_webpage("query_profile_encoded?query_id=123")
assert result.startswith("Could not obtain runtime profile: Query id")
def test_query_profile_encoded_unknown_query_id(self):
"""Test that /query_profile_encoded error message starts with the expected line in
case of missing query and does not contain any leading whitespace.
"""
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
result = impalad.service.read_debug_webpage("query_profile_encoded?query_id=123")
assert result.startswith("Could not obtain runtime profile: Query id")
def test_hive_udfs_missing_jar(self, vector, unique_database):
""" IMPALA-2365: Impalad shouldn't crash if the udf jar isn't present
on HDFS"""
# Copy hive-exec.jar to a temporary file
jar_path = get_fs_path(
"/test-warehouse/{0}.db/".format(unique_database) +
get_random_id(5) + ".jar")
hive_jar = get_fs_path("/test-warehouse/hive-exec.jar")
check_call(["hadoop", "fs", "-cp", hive_jar, jar_path])
drop_fn_stmt = ("drop function if exists "
"`{0}`.`pi_missing_jar`()".format(unique_database))
create_fn_stmt = (
"create function `{0}`.`pi_missing_jar`() returns double location '{1}' "
"symbol='org.apache.hadoop.hive.ql.udf.UDFPI'".format(
unique_database, jar_path))
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
client = impalad.service.create_beeswax_client()
# Create and drop functions with sync_ddl to make sure they are reflected
# in every impalad.
exec_option = copy(vector.get_value('exec_option'))
exec_option['sync_ddl'] = 1
self.execute_query_expect_success(client, drop_fn_stmt, exec_option)
self.execute_query_expect_success(client, create_fn_stmt, exec_option)
# Delete the udf jar
check_call(["hadoop", "fs", "-rm", jar_path])
different_impalad = cluster.get_different_impalad(impalad)
client = different_impalad.service.create_beeswax_client()
# Run a query using the udf from an impalad other than the one
# we used to create the function. This is to bypass loading from
# the cache
try:
self.execute_query_using_client(
client,
"select `{0}`.`pi_missing_jar`()".format(unique_database),
vector)
assert False, "Query expected to fail"
except ImpalaBeeswaxException, e:
assert "Failed to get file info" in str(e)
def test_hive_udfs_missing_jar(self, vector, unique_database):
""" IMPALA-2365: Impalad shouldn't crash if the udf jar isn't present
on HDFS"""
# Copy hive-exec.jar to a temporary file
jar_path = get_fs_path("/test-warehouse/{0}.db/".format(unique_database)
+ get_random_id(5) + ".jar")
hive_jar = get_fs_path("/test-warehouse/hive-exec.jar")
check_call(["hadoop", "fs", "-cp", hive_jar, jar_path])
drop_fn_stmt = (
"drop function if exists "
"`{0}`.`pi_missing_jar`()".format(unique_database))
create_fn_stmt = (
"create function `{0}`.`pi_missing_jar`() returns double location '{1}' "
"symbol='org.apache.hadoop.hive.ql.udf.UDFPI'".format(unique_database, jar_path))
cluster = ImpalaCluster()
impalad = cluster.get_any_impalad()
client = impalad.service.create_beeswax_client()
# Create and drop functions with sync_ddl to make sure they are reflected
# in every impalad.
exec_option = copy(vector.get_value('exec_option'))
exec_option['sync_ddl'] = 1
self.execute_query_expect_success(client, drop_fn_stmt, exec_option)
self.execute_query_expect_success(client, create_fn_stmt, exec_option)
# Delete the udf jar
check_call(["hadoop", "fs", "-rm", jar_path])
different_impalad = cluster.get_different_impalad(impalad)
client = different_impalad.service.create_beeswax_client()
# Run a query using the udf from an impalad other than the one
# we used to create the function. This is to bypass loading from
# the cache
try:
self.execute_query_using_client(
client, "select `{0}`.`pi_missing_jar`()".format(unique_database), vector)
assert False, "Query expected to fail"
except ImpalaBeeswaxException, e:
assert "Failed to get file info" in str(e)
