def test_has_write_access_backend(self):
client = make_logged_in_client(
username="******", groupname="write_access_backend", is_superuser=False
)
grant_access("write_access_backend", "write_access_backend", "metastore")
grant_access("write_access_backend", "write_access_backend", "beeswax")
user = User.objects.get(username="******")
resp = _make_query(
client, "CREATE TABLE test_perm_1 (a int);", database=self.db_name
) # Only fails if we were using Sentry and won't allow SELECT to user
resp = wait_for_query_to_finish(client, resp, max=30.0)
def check(client, http_codes):
resp = client.get("/metastore/tables/drop/%s" % self.db_name)
assert_true(resp.status_code in http_codes, resp.content)
resp = client.post("/metastore/tables/drop/%s" % self.db_name, {u"table_selection": [u"test_perm_1"]})
assert_true(resp.status_code in http_codes, resp.content)
check(client, [301]) # Denied
# Add access
group, created = Group.objects.get_or_create(name="write_access_backend")
perm, created = HuePermission.objects.get_or_create(app="metastore", action="write")
GroupPermission.objects.get_or_create(group=group, hue_permission=perm)
check(client, [200, 302]) # Ok
def test_has_write_access_backend(self):
client = make_logged_in_client(
username="******", groupname="write_access_backend", is_superuser=False
)
grant_access("write_access_backend", "write_access_backend", "metastore")
grant_access("write_access_backend", "write_access_backend", "beeswax")
user = User.objects.get(username="******")
def check(client, http_code):
resp = _make_query(client, "CREATE TABLE test_perm_1 (a int);")
resp = wait_for_query_to_finish(client, resp, max=30.0)
resp = client.get("/metastore/tables/drop/default", follow=True)
# assert_true('want to delete' in resp.content, resp.content)
assert_equal(resp.status_code, http_code, resp.content)
resp = client.post("/metastore/tables/drop/default", {u"table_selection": [u"test_perm_1"]}, follow=True)
assert_equal(resp.status_code, http_code, resp.content)
check(client, 200)
# Remove access
group, created = Group.objects.get_or_create(name="write_access_backend")
perm, created = HuePermission.objects.get_or_create(app="metastore", action="read_only_access")
GroupPermission.objects.get_or_create(group=group, hue_permission=perm)
check(client, 500)
def test_has_write_access_backend(self):
if is_live_cluster():
raise SkipTest('HUE-2900: Needs debugging on live cluster')
client = make_logged_in_client(username='******', groupname='write_access_backend', is_superuser=False)
grant_access("write_access_backend", "write_access_backend", "metastore")
grant_access("write_access_backend", "write_access_backend", "beeswax")
user = User.objects.get(username='******')
resp = _make_query(client, 'CREATE TABLE test_perm_1 (a int);', database=self.db_name) # Only fails if we were using Sentry and won't allow SELECT to user
resp = wait_for_query_to_finish(client, resp, max=30.0)
def check(client, http_codes):
resp = client.get('/metastore/tables/drop/%s' % self.db_name)
assert_true(resp.status_code in http_codes, resp.content)
resp = client.post('/metastore/tables/drop/%s' % self.db_name, {u'table_selection': [u'test_perm_1']})
assert_true(resp.status_code in http_codes, resp.content)
check(client, [301]) # Denied
# Add access
group, created = Group.objects.get_or_create(name='write_access_backend')
perm, created = HuePermission.objects.get_or_create(app='metastore', action='write')
GroupPermission.objects.get_or_create(group=group, hue_permission=perm)
check(client, [200, 302]) # Ok
def setup_class(cls):
if not is_live_cluster():
raise SkipTest()
cls.client = make_logged_in_client(username="******", is_superuser=False)
cls.user = User.objects.get(username="******")
add_to_group("test")
grant_access("test", "test", "libzookeeper")
# Create a ZKNode namespace
cls.namespace = "TestWithZooKeeper"
# Create temporary test directory and file with contents
cls.local_directory = tempfile.mkdtemp()
# Create subdirectory
cls.subdir_name = "subdir"
subdir_path = "%s/%s" % (cls.local_directory, cls.subdir_name)
os.mkdir(subdir_path, 0755)
# Create file
cls.filename = "test.txt"
file_path = "%s/%s" % (subdir_path, cls.filename)
cls.file_contents = "This is a test"
file = open(file_path, "w+")
file.write(cls.file_contents)
file.close()
def test_has_write_access_frontend(self):
# HS2 bug: Proxy user substitution is not supported for unsecure hadoop
raise SkipTest
client = make_logged_in_client(username='******', groupname='write_access_frontend', is_superuser=False)
grant_access("write_access_frontend", "write_access_frontend", "metastore")
user = User.objects.get(username='******')
def check(client, assertz):
response = client.get("/metastore/databases")
assertz("Drop</button>" in response.content, response.content)
assertz("Create a new database" in response.content, response.content)
response = client.get("/metastore/tables/")
assertz("Drop</button>" in response.content, response.content)
assertz("Create a new table" in response.content, response.content)
check(client, assert_true)
# Remove access
group, created = Group.objects.get_or_create(name='write_access_frontend')
perm, created = HuePermission.objects.get_or_create(app='metastore', action='read_only_access')
GroupPermission.objects.get_or_create(group=group, hue_permission=perm)
check(client, assert_false)
def setup_class(cls):
SqoopServerProvider.setup_class()
cls.client = make_logged_in_client(username="******", is_superuser=False)
cls.user = User.objects.get(username="******")
add_to_group("test")
grant_access("test", "test", "sqoop")
def test_impersonation():
from hbased import Hbase as thrift_hbase
c = make_logged_in_client(username='******', is_superuser=False)
grant_access('test_hbase', 'test_hbase', 'hbase')
user = User.objects.get(username='******')
proto = MockProtocol()
client = thrift_hbase.Client(proto)
impersonation_enabled = is_impersonation_enabled()
get_conf()[_CNF_HBASE_IMPERSONATION_ENABLED] = 'FALSE'
try:
client.getTableNames(doas=user.username)
except AttributeError:
pass # We don't mock everything
finally:
get_conf()[_CNF_HBASE_IMPERSONATION_ENABLED] = impersonation_enabled
assert_equal({}, proto.get_headers())
get_conf()[_CNF_HBASE_IMPERSONATION_ENABLED] = 'TRUE'
try:
client.getTableNames(doas=user.username)
except AttributeError:
pass # We don't mock everything
finally:
get_conf()[_CNF_HBASE_IMPERSONATION_ENABLED] = impersonation_enabled
assert_equal({'doAs': u'test_hbase'}, proto.get_headers())
def test_has_write_access_backend(self):
# HS2 bug: Proxy user substitution is not supported for unsecure hadoop
raise SkipTest
client = make_logged_in_client(username='******', groupname='write_access_backend', is_superuser=False)
grant_access("write_access_backend", "write_access_backend", "metastore")
grant_access("write_access_backend", "write_access_backend", "beeswax")
user = User.objects.get(username='******')
def check(client, http_code):
resp = _make_query(client, 'CREATE TABLE test_perm_1 (a int);')
resp = wait_for_query_to_finish(client, resp, max=30.0)
resp = client.get('/metastore/tables/drop/default', follow=True)
#assert_true('want to delete' in resp.content, resp.content)
assert_equal(resp.status_code, http_code, resp.content)
resp = client.post('/metastore/tables/drop/default', {u'table_selection': [u'test_perm_1']}, follow=True)
assert_equal(resp.status_code, http_code, resp.content)
check(client, 200)
# Remove access
group, created = Group.objects.get_or_create(name='write_access_backend')
perm, created = HuePermission.objects.get_or_create(app='metastore', action='read_only_access')
GroupPermission.objects.get_or_create(group=group, hue_permission=perm)
check(client, 500)
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="default", recreate=True, is_superuser=False)
self.client_not_me = make_logged_in_client(username="******", groupname="default", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
self.user_not_me = User.objects.get(username="******")
grant_access("test", "default", "notebook")
grant_access("not_perm_user", "default", "notebook")
self.notebook_json = """
{
"selectedSnippet": "hive",
"showHistory": false,
"description": "Test Hive Query",
"name": "Test Hive Query",
"sessions": [
{
"type": "hive",
"properties": [],
"id": null
}
],
"type": "query-hive",
"id": 50010,
"snippets": [],
"uuid": "5982a274-de78-083c-2efc-74f53dce744c"
}
"""
self.notebook = json.loads(self.notebook_json)
self.doc2 = Document2.objects.create(id=50010, name=self.notebook['name'], type=self.notebook['type'], owner=self.user)
self.doc1 = Document.objects.link(self.doc2, owner=self.user, name=self.doc2.name,
description=self.doc2.description, extra=self.doc2.type)
def test_job_permissions(self):
# Login as ourself
finish = SHARE_JOBS.set_for_testing(True)
try:
response = TestJobBrowserWithHadoop.client.get('/jobbrowser/jobs/?format=json&user='******'/jobbrowser/jobs/?format=json&user='******'not_me', is_superuser=False, groupname='test')
grant_access("not_me", "test", "jobbrowser")
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.get('/jobbrowser/jobs/?format=json&user='******'/jobbrowser/jobs/?format=json&user=')
assert_false(TestJobBrowserWithHadoop.hadoop_job_id_short in response.content)
finally:
finish()
def setUp(self):
self.c = make_logged_in_client(username='******', is_superuser=False)
grant_access('test_search', 'test_search', 'search')
self.user = User.objects.get(username='******')
self.prev_resource = resource.Resource
resource.Resource = MockResource
def setUp(self):
# Beware: Monkey patching
if not hasattr(resource_manager_api, 'old_get_resource_manager_api'):
resource_manager_api.old_get_resource_manager = resource_manager_api.get_resource_manager
if not hasattr(resource_manager_api, 'old_get_mapreduce_api'):
mapreduce_api.old_get_mapreduce_api = mapreduce_api.get_mapreduce_api
if not hasattr(history_server_api, 'old_get_history_server_api'):
history_server_api.old_get_history_server_api = history_server_api.get_history_server_api
self.c = make_logged_in_client(is_superuser=False)
grant_access("test", "test", "jobbrowser")
self.user = User.objects.get(username='******')
self.c2 = make_logged_in_client(is_superuser=False, username="******")
grant_access("test2", "test2", "jobbrowser")
self.user2 = User.objects.get(username='******')
resource_manager_api.get_resource_manager = lambda user: MockResourceManagerApi(user)
mapreduce_api.get_mapreduce_api = lambda: MockMapreduceApi()
history_server_api.get_history_server_api = lambda: HistoryServerApi()
self.finish = [
YARN_CLUSTERS['default'].SUBMIT_TO.set_for_testing(True),
SHARE_JOBS.set_for_testing(False)
]
assert_true(cluster.is_yarn())
def test_dump_config():
c = make_logged_in_client()
CANARY = "abracadabra"
# Depending on the order of the conf.initialize() in settings, the set_for_testing() are not seen in the global settings variable
clear = HIVE_SERVER_HOST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content, response1.content)
response2 = c.get(reverse('desktop.views.dump_config'), dict(private="true"))
assert_true(CANARY in response2.content)
# There are more private variables...
assert_true(len(response1.content) < len(response2.content))
clear()
CANARY = "(localhost|127\.0\.0\.1):(50030|50070|50060|50075)"
clear = proxy.conf.WHITELIST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content)
clear()
# Malformed port per HUE-674
CANARY = "asdfoijaoidfjaosdjffjfjaoojosjfiojdosjoidjfoa"
clear = HIVE_SERVER_HOST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content, response1.content)
clear()
CANARY = '/tmp/spacé.dat'
finish = proxy.conf.WHITELIST.set_for_testing(CANARY)
try:
response = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response.content, response.content)
finally:
finish()
# Not showing some passwords
response = c.get(reverse('desktop.views.dump_config'))
assert_false('bind_password' in response.content)
# Login as someone else
client_not_me = make_logged_in_client(username='******', is_superuser=False, groupname='test')
grant_access("not_me", "test", "desktop")
response = client_not_me.get(reverse('desktop.views.dump_config'))
assert_true("You must be a superuser" in response.content, response.content)
os.environ["HUE_CONF_DIR"] = "/tmp/test_hue_conf_dir"
resp = c.get(reverse('desktop.views.dump_config'))
del os.environ["HUE_CONF_DIR"]
assert_true('/tmp/test_hue_conf_dir' in resp.content, resp)
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="test", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
grant_access("test", "test", "libsentry")
self.api_v1 = MockSentryApiV1()
self.api_v2 = MockSentryApiV2()
self.checker = PrivilegeChecker(user=self.user, api_v1=self.api_v1, api_v2=self.api_v2)
def setUp(self):
super(TestWithHadoop, self).setUp()
# FIXME (HUE-2562): The tests unfortunately require superuser at the
# moment, but should be rewritten to not need it.
self.c = make_logged_in_client(is_superuser=True)
grant_access("test", "test", "pig")
self.user = User.objects.get(username='******')
self.c.post(reverse('pig:install_examples'))
def setUpClass(cls):
S3TestBase.setUpClass()
if not cls.shouldSkip():
cls.fs = S3FileSystem(cls.s3_connection)
cls.c = make_logged_in_client(username='******', is_superuser=False)
grant_access('test', 'test', 'filebrowser')
add_to_group('test')
def test_upload_file(self):
with tempfile.NamedTemporaryFile() as local_file:
# Make sure we can upload larger than the UPLOAD chunk size
file_size = UPLOAD_CHUNK_SIZE.get() * 2
local_file.write('0' * file_size)
local_file.flush()
prefix = self.cluster.fs_prefix + '/test_upload_file'
self.cluster.fs.mkdir(prefix)
USER_NAME = 'test'
HDFS_DEST_DIR = prefix + "/tmp/fb-upload-test"
LOCAL_FILE = local_file.name
HDFS_FILE = HDFS_DEST_DIR + '/' + os.path.basename(LOCAL_FILE)
self.cluster.fs.do_as_superuser(self.cluster.fs.mkdir, HDFS_DEST_DIR)
self.cluster.fs.do_as_superuser(self.cluster.fs.chown, HDFS_DEST_DIR, USER_NAME, USER_NAME)
self.cluster.fs.do_as_superuser(self.cluster.fs.chmod, HDFS_DEST_DIR, 0700)
stats = self.cluster.fs.stats(HDFS_DEST_DIR)
assert_equal(stats['user'], USER_NAME)
assert_equal(stats['group'], USER_NAME)
# Just upload the current python file
resp = self.c.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR, # GET param avoids infinite looping
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(0, response['status'], response)
stats = self.cluster.fs.stats(HDFS_FILE)
assert_equal(stats['user'], USER_NAME)
assert_equal(stats['group'], USER_NAME)
f = self.cluster.fs.open(HDFS_FILE)
actual = f.read(file_size)
expected = file(LOCAL_FILE).read()
assert_equal(actual, expected, 'files do not match: %s != %s' % (len(actual), len(expected)))
# Upload again and so fails because file already exits
resp = self.c.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR,
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(-1, response['status'], response)
assert_true('already exists' in response['data'], response)
# Upload in / and fails because of missing permissions
not_me = make_logged_in_client("not_me", is_superuser=False)
grant_access("not_me", "not_me", "filebrowser")
try:
resp = not_me.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR,
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(-1, response['status'], response)
assert_true('Permission denied' in response['data'], response)
except AttributeError:
# Seems like a Django bug.
# StopFutureHandlers() does not seem to work in test mode as it continues to MemoryFileUploadHandler after perm issue and so fails.
pass
def test_upload_file():
"""Test file upload"""
cluster = pseudo_hdfs4.shared_cluster()
try:
USER_NAME = 'test'
HDFS_DEST_DIR = "/tmp/fb-upload-test"
LOCAL_FILE = __file__
HDFS_FILE = HDFS_DEST_DIR + '/' + os.path.basename(__file__)
cluster.fs.setuser(USER_NAME)
client = make_logged_in_client(USER_NAME)
cluster.fs.do_as_superuser(cluster.fs.mkdir, HDFS_DEST_DIR)
cluster.fs.do_as_superuser(cluster.fs.chown, HDFS_DEST_DIR, USER_NAME, USER_NAME)
cluster.fs.do_as_superuser(cluster.fs.chmod, HDFS_DEST_DIR, 0700)
stats = cluster.fs.stats(HDFS_DEST_DIR)
assert_equal(stats['user'], USER_NAME)
assert_equal(stats['group'], USER_NAME)
# Just upload the current python file
resp = client.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR, # GET param avoids infinite looping
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(0, response['status'], response)
stats = cluster.fs.stats(HDFS_FILE)
assert_equal(stats['user'], USER_NAME)
assert_equal(stats['group'], USER_NAME)
f = cluster.fs.open(HDFS_FILE)
actual = f.read()
expected = file(LOCAL_FILE).read()
assert_equal(actual, expected)
# Upload again and so fails because file already exits
resp = client.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR,
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(-1, response['status'], response)
assert_true('already exists' in response['data'], response)
# Upload in / and fails because of missing permissions
not_me = make_logged_in_client("not_me", is_superuser=False)
grant_access("not_me", "not_me", "filebrowser")
try:
resp = not_me.post('/filebrowser/upload/file?dest=%s' % HDFS_DEST_DIR,
dict(dest=HDFS_DEST_DIR, hdfs_file=file(LOCAL_FILE)))
response = json.loads(resp.content)
assert_equal(-1, response['status'], response)
assert_true('Permission denied' in response['data'], response)
except AttributeError:
# Seems like a Django bug.
# StopFutureHandlers() does not seem to work in test mode as it continues to MemoryFileUploadHandler after perm issue and so fails.
pass
finally:
cleanup_file(cluster, HDFS_DEST_DIR)
def test_workflow_action_permissions(self):
# Login as someone else
client_not_me = make_logged_in_client(username='******', is_superuser=False, groupname='test')
grant_access("not_me", "test", "oozie")
action1 = Node.objects.get(name='action-name-1')
# Edit
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.get(reverse('oozie:edit_action', args=[action1.id]))
assert_true('Permission denied' in response.content, response.content)
finally:
finish()
# Edit
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.post(reverse('oozie:edit_action', args=[action1.id]))
assert_true('Permission denied' in response.content, response.content)
finally:
finish()
# Delete
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.post(reverse('oozie:delete_action', args=[action1.id]))
assert_true('Permission denied' in response.content, response.content)
finally:
finish()
action1.workflow.is_shared = True
action1.workflow.save()
# Edit
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.get(reverse('oozie:edit_action', args=[action1.id]))
assert_false('Permission denied' in response.content, response.content)
finally:
finish()
# Edit
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.post(reverse('oozie:edit_action', args=[action1.id]))
assert_true('Not allowed' in response.content, response.content)
finally:
finish()
# Delete
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.post(reverse('oozie:delete_action', args=[action1.id]))
assert_true('Not allowed' in response.content, response.content)
finally:
finish()
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="test", recreate=False, is_superuser=False)
self.user = User.objects.get(username='******')
add_to_group('test')
grant_access("test", "test", "notebook")
self.db = dbms.get(self.user, get_query_server_config())
self.api = HS2Api(self.user)
def test_useradmin_ldap_user_group_membership_sync():
settings.MIDDLEWARE_CLASSES.append('useradmin.middleware.LdapSynchronizationMiddleware')
reset_all_users()
reset_all_groups()
# Set up LDAP tests to use a LdapTestConnection instead of an actual LDAP connection
ldap_access.CACHED_LDAP_CONN = LdapTestConnection()
# Make sure LDAP groups exist or they won't sync
import_ldap_groups(ldap_access.CACHED_LDAP_CONN, 'TestUsers', import_members=False, import_members_recursive=False, sync_users=False, import_by_dn=False)
import_ldap_groups(ldap_access.CACHED_LDAP_CONN, 'Test Administrators', import_members=False, import_members_recursive=False, sync_users=False, import_by_dn=False)
reset = []
# Set to nonsensical value just to force new config usage.
# Should continue to use cached connection.
reset.append(desktop.conf.LDAP.LDAP_SERVERS.set_for_testing(get_nonsense_config()))
try:
# Import curly who is part of TestUsers and Test Administrators
import_ldap_users(ldap_access.CACHED_LDAP_CONN, 'curly', sync_groups=False, import_by_dn=False)
# Set a password so that we can login
user = User.objects.get(username='******')
user.set_password('test')
user.save()
# Should have 0 groups
assert_equal(0, user.groups.all().count())
# Make an authenticated request as curly so that we can see call middleware.
c = make_logged_in_client('curly', 'test', is_superuser=False)
grant_access("curly", "test", "useradmin")
response = c.get('/useradmin/users')
# Refresh user groups
user = User.objects.get(username='******')
# Should have 3 groups now. 2 from LDAP and 1 from 'grant_access' call.
assert_equal(3, user.groups.all().count(), user.groups.all())
# Now remove a group and try again.
old_group = ldap_access.CACHED_LDAP_CONN._instance.users['curly']['groups'].pop()
# Make an authenticated request as curly so that we can see call middleware.
response = c.get('/useradmin/users')
# Refresh user groups
user = User.objects.get(username='******')
# Should have 2 groups now. 1 from LDAP and 1 from 'grant_access' call.
assert_equal(3, user.groups.all().count(), user.groups.all())
finally:
settings.MIDDLEWARE_CLASSES.remove('useradmin.middleware.LdapSynchronizationMiddleware')
for finish in reset:
finish()
def setup_class(cls):
cls.client = make_logged_in_client(username='******', is_superuser=False)
cls.user = User.objects.get(username='******')
cls.user = rewrite_user(cls.user)
add_to_group('test')
grant_access("test", "test", "metadata")
if not is_live_cluster() or not has_navigator(cls.user):
raise SkipTest
def setUp(self):
self.client = make_logged_in_client(username="******")
self.client_not_me = make_logged_in_client(username="******")
self.user = User.objects.get(username="******")
self.user_not_me = User.objects.get(username="******")
grant_access(self.user.username, self.user.username, "desktop")
grant_access(self.user_not_me.username, self.user_not_me.username, "desktop")
def setup_class(cls):
cls.db_name = get_db_prefix(name="hive")
cls.cluster, shutdown = get_shared_beeswax_server(cls.db_name)
cls.client = make_logged_in_client(username="******", is_superuser=False)
add_to_group("test")
grant_access("test", "test", "beeswax")
# Weird redirection to avoid binding nonsense.
cls.shutdown = [shutdown]
cls.init_beeswax_db()
def setup_class(cls):
if not is_live_cluster():
raise SkipTest('These tests can only run on a live cluster')
cls.client = make_logged_in_client(username='******', is_superuser=False)
cls.user = User.objects.get(username='******')
add_to_group('test')
grant_access("test", "test", "indexer")
def setUp(self):
self.client = make_logged_in_client(username="******", recreate=True, is_superuser=False)
self.client_not_me = make_logged_in_client(username="******", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
self.user_not_me = User.objects.get(username="******")
grant_access(self.user.username, self.user.username, "desktop")
grant_access(self.user_not_me.username, self.user_not_me.username, "desktop")
def setup_class(cls):
if not is_live_cluster():
raise SkipTest()
cls.client = make_logged_in_client(username='******', is_superuser=False)
cls.user = User.objects.get(username='******')
add_to_group('test')
grant_access("test", "test", "libzookeeper")
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="doc2", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
grant_access("doc2", "doc2", "beeswax")
# Setup Home dir this way currently
response = self.client.get('/desktop/api2/docs/')
data = json.loads(response.content)
assert_equal('/', data['path'], data)
def setup_class(cls):
cls.db_name = get_db_prefix(name='hive')
cls.cluster, shutdown = get_shared_beeswax_server(cls.db_name)
cls.client = make_logged_in_client(username='******', is_superuser=False)
add_to_group('test', 'test')
grant_access('test', 'test', 'beeswax')
grant_access('test', 'test', 'metastore')
# Weird redirection to avoid binding nonsense.
cls.shutdown = [ shutdown ]
cls.init_beeswax_db()
def setUp(self):
self.c = make_logged_in_client(username='******', is_superuser=False)
grant_access('test', 'test', 'filebrowser')
add_to_group('test')
self.user = User.objects.get(username='******')
self.cluster = pseudo_hdfs4.shared_cluster()
self.cluster.fs.setuser('test')
self.prefix = self.cluster.fs_prefix + '/filebrowser'
self.cluster.fs.do_as_user('test', self.cluster.fs.create_home_dir, '/user/test')
def setUp(self):
self.cluster = pseudo_hdfs4.shared_cluster()
self.cli = make_logged_in_client(username='******', is_superuser=True)
grant_access('admin', 'admin', 'filebrowser')
self.cluster.fs.setuser('admin')
def test_dump_config():
c = make_logged_in_client()
CANARY = "abracadabra"
# Depending on the order of the conf.initialize() in settings, the set_for_testing() are not seen in the global settings variable
clear = HIVE_SERVER_HOST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content, response1.content)
response2 = c.get(reverse('desktop.views.dump_config'),
dict(private="true"))
assert_true(CANARY in response2.content)
# There are more private variables...
assert_true(len(response1.content) < len(response2.content))
clear()
CANARY = "(localhost|127\.0\.0\.1):(50030|50070|50060|50075)"
clear = proxy.conf.WHITELIST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content)
clear()
# Malformed port per HUE-674
CANARY = "asdfoijaoidfjaosdjffjfjaoojosjfiojdosjoidjfoa"
clear = HIVE_SERVER_HOST.set_for_testing(CANARY)
response1 = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response1.content, response1.content)
clear()
CANARY = '/tmp/spacé.dat'
finish = proxy.conf.WHITELIST.set_for_testing(CANARY)
try:
response = c.get(reverse('desktop.views.dump_config'))
assert_true(CANARY in response.content, response.content)
finally:
finish()
# Not showing some passwords
response = c.get(reverse('desktop.views.dump_config'))
assert_false('bind_password' in response.content)
# Login as someone else
client_not_me = make_logged_in_client(username='******',
is_superuser=False,
groupname='test')
grant_access("not_me", "test", "desktop")
response = client_not_me.get(reverse('desktop.views.dump_config'))
assert_true("You must be a superuser" in response.content,
response.content)
os.environ["HUE_CONF_DIR"] = "/tmp/test_hue_conf_dir"
resp = c.get(reverse('desktop.views.dump_config'))
del os.environ["HUE_CONF_DIR"]
assert_true('/tmp/test_hue_conf_dir' in resp.content, resp)
def test_job(self):
"""
Test new job views.
The status of the jobs should be the same as the status reported back by oozie.
In this case, all jobs should succeed.
"""
# Clone design
assert_equal(0, OozieDesign.objects.filter(owner__username=self.username).count())
self.client.post('/jobsub/clone_design/%d' % self.sleep_design_id)
assert_equal(1, OozieDesign.objects.filter(owner__username=self.username).count())
# Run the sleep example, since it doesn't require user home directory
design_id = OozieDesign.objects.get(owner__username=self.username).id
response = self.client.post("/jobsub/submit_design/%d" % (design_id,),
dict(map_sleep_time=1,
num_maps=1,
num_reduces=1,
reduce_sleep_time=1),
follow=True)
oozie_jobid = response.context['jobid']
job = OozieServerProvider.wait_until_completion(oozie_jobid, timeout=120, step=1)
hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
# All jobs page and fetch job ID
# Taking advantage of the fact new jobs are at the top of the list!
response = self.client.get('/jobbrowser/jobs/')
assert_true(hadoop_job_id in response.content)
# Make sure job succeeded
response = self.client.get('/jobbrowser/jobs/?state=completed')
assert_true(hadoop_job_id in response.content)
response = self.client.get('/jobbrowser/jobs/?state=failed')
assert_false(hadoop_job_id in response.content)
response = self.client.get('/jobbrowser/jobs/?state=running')
assert_false(hadoop_job_id in response.content)
response = self.client.get('/jobbrowser/jobs/?state=killed')
assert_false(hadoop_job_id in response.content)
# Check sharing permissions
# Login as ourself
finish = SHARE_JOBS.set_for_testing(True)
try:
response = self.client.get('/jobbrowser/jobs/?user='******'/jobbrowser/jobs/?user='******'not_me', is_superuser=False, groupname='test')
grant_access("not_me", "test", "jobbrowser")
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.get('/jobbrowser/jobs/?user='******'/jobbrowser/jobs/?user='******'/jobbrowser/jobs/%s' % hadoop_job_id)
# Check some counters for single job.
counters = response.context['job'].counters
counters_file_bytes_written = counters['org.apache.hadoop.mapreduce.FileSystemCounter']['counters']['FILE_BYTES_WRITTEN']
assert_true(counters_file_bytes_written['map'] > 0)
assert_true(counters_file_bytes_written['reduce'] > 0)
# We can't just check the complete contents of the python map because the
# SLOTS_MILLIS_* entries have a variable number of milliseconds from
# run-to-run.
assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_MAPS']['total'], 1)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['TOTAL_LAUNCHED_REDUCES']['total'], 1)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0)
assert_equal(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0)
assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
assert_true(response.context['job'].counters['org.apache.hadoop.mapreduce.JobCounter']['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)
# There should be 4 tasks for this job: cleanup, setup, map, reduce
response = self.client.get('/jobbrowser/jobs/%s/tasks' % (hadoop_job_id,))
assert_true(len(response.context['page'].object_list), 4)
# Select by tasktype
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id,))
assert_true(len(response.context['page'].object_list), 1)
# Select by taskstate
response = self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
assert_true(len(response.context['page'].object_list), 4)
# Select by text
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktext=clean' % (hadoop_job_id,))
assert_true(len(response.context['page'].object_list), 1)
# Test job single logs page
response = self.client.get('/jobbrowser/jobs/%s/single_logs' % (hadoop_job_id))
assert_true('syslog' in response.content)
def setUp(self):
self.c = make_logged_in_client(is_superuser=False)
grant_access("test", "test", "indexer")
add_to_group("test")
self.finish = ENABLE_NEW_INDEXER.set_for_testing(True)
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="empty", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
grant_access("test", "empty", "impala")
def test_useradmin_ldap_user_group_membership_sync(self):
settings.MIDDLEWARE_CLASSES.append(
'useradmin.middleware.LdapSynchronizationMiddleware')
try:
# Set up LDAP tests to use a LdapTestConnection instead of an actual LDAP connection
ldap_access.CACHED_LDAP_CONN = LdapTestConnection()
# Make sure LDAP groups exist or they won't sync
import_ldap_groups(ldap_access.CACHED_LDAP_CONN,
'TestUsers',
import_members=False,
import_members_recursive=False,
sync_users=False,
import_by_dn=False)
import_ldap_groups(ldap_access.CACHED_LDAP_CONN,
'Test Administrators',
import_members=False,
import_members_recursive=False,
sync_users=False,
import_by_dn=False)
# Import curly who is part of TestUsers and Test Administrators
import_ldap_users(ldap_access.CACHED_LDAP_CONN,
'curly',
sync_groups=False,
import_by_dn=False)
# Set a password so that we can login
user = User.objects.get(username='******')
user.set_password('test')
user.save()
# Should have 0 groups
assert_equal(0, user.groups.all().count())
# Make an authenticated request as curly so that we can see call middleware.
c = make_logged_in_client('curly', 'test', is_superuser=False)
grant_access("curly", "test", "useradmin")
response = c.get('/useradmin/users')
# Refresh user groups
user = User.objects.get(username='******')
# Should have 3 groups now. 2 from LDAP and 1 from 'grant_access' call.
assert_equal(3, user.groups.all().count(), user.groups.all())
# Now remove a group and try again.
old_group = ldap_access.CACHED_LDAP_CONN._instance.users['curly'][
'groups'].pop()
# Make an authenticated request as curly so that we can see call middleware.
response = c.get('/useradmin/users')
# Refresh user groups
user = User.objects.get(username='******')
# Should have 2 groups now. 1 from LDAP and 1 from 'grant_access' call.
assert_equal(3, user.groups.all().count(), user.groups.all())
finally:
settings.MIDDLEWARE_CLASSES.remove(
'useradmin.middleware.LdapSynchronizationMiddleware')
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="default", recreate=True, is_superuser=False)
self.user = rewrite_user(User.objects.get(username="******"))
grant_access("test", "default", "notebook")
def setUp(self):
self.client = make_logged_in_client(username="******",
groupname="default",
recreate=True,
is_superuser=False)
self.client_not_me = make_logged_in_client(username="******",
groupname="default",
recreate=True,
is_superuser=False)
self.user = User.objects.get(username="******")
self.user_not_me = User.objects.get(username="******")
# Beware: Monkey patch HS2API Mock API
if not hasattr(
notebook.connectors.hiveserver2,
'original_HS2Api'): # Could not monkey patch base.get_api
notebook.connectors.hiveserver2.original_HS2Api = notebook.connectors.hiveserver2.HS2Api
notebook.connectors.hiveserver2.HS2Api = MockedApi
originalCluster.get_hdfs()
self.original_fs = originalCluster.FS_CACHE["default"]
originalCluster.FS_CACHE["default"] = MockFs()
grant_access("test", "default", "notebook")
grant_access("test", "default", "beeswax")
grant_access("test", "default", "hive")
grant_access("not_perm_user", "default", "notebook")
grant_access("not_perm_user", "default", "beeswax")
grant_access("not_perm_user", "default", "hive")
add_permission('test',
'has_adls',
permname='adls_access',
appname='filebrowser')
def test_job_design_cycle():
"""
Tests for the "job design" CMS.
Submission requires a cluster, so that's separate.
"""
c = make_logged_in_client()
# New should give us a form.
response = c.get('/jobsub/new_design/java')
assert_equal(
1,
response.content.count(
'action="/jobsub/new_design/java" method="POST"'))
# Streaming also:
response = c.get('/jobsub/new_design/streaming')
assert_equal(
1,
response.content.count(
'action="/jobsub/new_design/streaming" method="POST"'))
# Post back to create a new submission
design_count = OozieDesign.objects.count()
response = c.post(
'/jobsub/new_design/java', {
u'wf-name': [u'name-1'],
u'wf-description': [u'description name-1'],
u'action-args': [u'x y z'],
u'action-main_class': [u'MyClass'],
u'action-jar_path': [u'myfile.jar'],
u'action-java_opts': [u''],
u'action-archives': [u'[]'],
u'action-job_properties': [u'[]'],
u'action-files': [u'[]']
})
assert_equal(design_count + 1, OozieDesign.objects.count())
job_id = OozieDesign.objects.get(name='name-1').id
response = c.post(
'/jobsub/new_design/mapreduce', {
u'wf-name': [u'name-2'],
u'wf-description': [u'description name-2'],
u'action-args': [u'x y z'],
u'action-jar_path': [u'myfile.jar'],
u'action-archives': [u'[]'],
u'action-job_properties': [u'[]'],
u'action-files': [u'[]']
})
# Follow it
edit_url = '/jobsub/edit_design/%d' % job_id
response = c.get(edit_url)
assert_true('x y z' in response.content, response.content)
# Make an edit
response = c.post(
edit_url, {
u'wf-name': [u'name-1'],
u'wf-description': [u'description name-1'],
u'action-args': [u'a b c'],
u'action-main_class': [u'MyClass'],
u'action-jar_path': [u'myfile.jar'],
u'action-java_opts': [u''],
u'action-archives': [u'[]'],
u'action-job_properties': [u'[]'],
u'action-files': [u'[]']
})
assert_true('a b c' in c.get(edit_url).content)
# Try to post
response = c.post(
'/jobsub/new_design/java',
dict(name='test2',
jarfile='myfile.jar',
arguments='x y z',
submit='Save'))
assert_false('This field is required' in response)
# Now check list
response = c.get('/jobsub/')
for design in OozieDesign.objects.all():
assert_true(design.name in response.content, response.content)
# With some filters
response = c.get('/jobsub/', dict(name='name-1'))
assert_true('name-1' in response.content, response.content)
assert_false('name-2' in response.content, response.content)
response = c.get('/jobsub/', dict(owner='doesnotexist'))
assert_false('doesnotexist' in response.content)
response = c.get('/jobsub/', dict(owner='test', name='name-1'))
assert_true('name-1' in response.content, response.content)
assert_false('name-2' in response.content, response.content)
response = c.get('/jobsub/', dict(name="name"))
assert_true('name-1' in response.content, response.content)
assert_true('name-2' in response.content, response.content)
assert_false('doesnotexist' in response.content, response.content)
# Combined filters
response = c.get('/jobsub/', dict(owner="test", name="name-2"))
assert_false('name-1' in response.content, response.content)
assert_true('name-2' in response.content, response.content)
assert_false('doesnotexist' in response.content, response.content)
# Try delete
job_id = OozieDesign.objects.get(name='name-1').id
response = c.post('/jobsub/delete_design/%d' % job_id)
assert_raises(OozieDesign.DoesNotExist, OozieDesign.objects.get, id=job_id)
# Let's make sure we can't delete other people's designs.
c.logout()
c = make_logged_in_client('test2', is_superuser=False)
grant_access('test2', 'test-grp', 'jobsub')
not_mine = OozieDesign.objects.get(name='name-2')
response = c.post('/jobsub/delete_design/%d' % not_mine.id)
assert_true('Permission denied.' in response.content, response.content)
def setUp(self):
self.client = make_logged_in_client(username="******", groupname="default", recreate=True, is_superuser=False)
self.user = User.objects.get(username="******")
grant_access(self.user.username, self.user.username, "desktop")
def test_job(self):
"""
Test new job views.
The status of the jobs should be the same as the status reported back by oozie.
In this case, all jobs should succeed.
"""
# Run the sleep example, since it doesn't require user home directory
design_id = self.design.id
response = self.client.post(reverse('oozie:submit_workflow',
args=[design_id]),
data={
u'form-MAX_NUM_FORMS': [u''],
u'form-INITIAL_FORMS': [u'1'],
u'form-0-name':
[u'REDUCER_SLEEP_TIME'],
u'form-0-value': [u'1'],
u'form-TOTAL_FORMS': [u'1']
},
follow=True)
oozie_jobid = response.context['oozie_workflow'].id
OozieServerProvider.wait_until_completion(oozie_jobid,
timeout=120,
step=1)
hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
hadoop_job_id_short = views.get_shorter_id(hadoop_job_id)
# All jobs page and fetch job ID
# Taking advantage of the fact new jobs are at the top of the list!
response = self.client.get('/jobbrowser/jobs/?format=json')
assert_true(hadoop_job_id_short in response.content, response.content)
# Make sure job succeeded
response = self.client.get(
'/jobbrowser/jobs/?format=json&state=completed')
assert_true(hadoop_job_id_short in response.content)
response = self.client.get(
'/jobbrowser/jobs/?format=json&state=failed')
assert_false(hadoop_job_id_short in response.content)
response = self.client.get(
'/jobbrowser/jobs/?format=json&state=running')
assert_false(hadoop_job_id_short in response.content)
response = self.client.get(
'/jobbrowser/jobs/?format=json&state=killed')
assert_false(hadoop_job_id_short in response.content)
# Test tracker page
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' %
(hadoop_job_id, early_task_id))
tracker_url = re.search('<a href="(/jobbrowser/trackers/.+?)"',
response.content).group(1)
response = self.client.get(tracker_url)
assert_true('Tracker at' in response.content)
# Check sharing permissions
# Login as ourself
finish = SHARE_JOBS.set_for_testing(True)
try:
response = self.client.get('/jobbrowser/jobs/?format=json&user='******'/jobbrowser/jobs/?format=json&user='******'not_me',
is_superuser=False,
groupname='test')
grant_access("not_me", "test", "jobbrowser")
finish = SHARE_JOBS.set_for_testing(True)
try:
response = client_not_me.get('/jobbrowser/jobs/?format=json&user='******'/jobbrowser/jobs/?format=json&user='******'/jobbrowser/jobs/%s' % hadoop_job_id)
# Check some counters for single job.
counters = response.context['job'].counters
counters_file_bytes_written = counters[
'org.apache.hadoop.mapreduce.FileSystemCounter']['counters'][
'FILE_BYTES_WRITTEN']
assert_true(counters_file_bytes_written['map'] > 0)
assert_true(counters_file_bytes_written['reduce'] > 0)
# We can't just check the complete contents of the python map because the
# SLOTS_MILLIS_* entries have a variable number of milliseconds from
# run-to-run.
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']['counters']
['TOTAL_LAUNCHED_MAPS']['total'], 2L)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']['counters']
['TOTAL_LAUNCHED_REDUCES']['total'], 1L)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']['counters']
['FALLOW_SLOTS_MILLIS_MAPS']['total'], 0L)
assert_equal(
response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']['counters']
['FALLOW_SLOTS_MILLIS_REDUCES']['total'], 0L)
assert_true(response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']
['counters']['SLOTS_MILLIS_MAPS']['total'] > 0)
assert_true(response.context['job'].
counters['org.apache.hadoop.mapreduce.JobCounter']
['counters']['SLOTS_MILLIS_REDUCES']['total'] > 0)
# There should be 4 tasks for this job: cleanup, setup, map, reduce
response = self.client.get('/jobbrowser/jobs/%s/tasks' %
(hadoop_job_id, ))
assert_true(len(response.context['page'].object_list), 4)
# Select by tasktype
response = self.client.get(
'/jobbrowser/jobs/%s/tasks?tasktype=reduce' % (hadoop_job_id, ))
assert_true(len(response.context['page'].object_list), 1)
# Select by taskstate
response = self.client.get(
'/jobbrowser/jobs/%s/tasks?taskstate=succeeded' %
(hadoop_job_id, ))
assert_true(len(response.context['page'].object_list), 4)
# Select by text
response = self.client.get('/jobbrowser/jobs/%s/tasks?tasktext=clean' %
(hadoop_job_id, ))
assert_true(len(response.context['page'].object_list), 1)
# Test job single logs page
response = self.client.get('/jobbrowser/jobs/%s/single_logs' %
(hadoop_job_id))
assert_true('syslog' in response.content)
assert_true(
'<div class="tab-pane active" id="logsSysLog">' in response.content
or '<div class="tab-pane active" id="logsStdErr">'
in response.content or # Depending on Hadoop
'<div class="tab-pane active" id="logsStdOut">'
in response.content, # For jenkins
response.content)
def test_kill_job(self):
"""
Test job in kill state.
"""
# Run the sleep example, since it doesn't require user home directory
design_id = self.design.id
response = self.client.post(reverse('oozie:submit_workflow',
args=[self.design.id]),
data={
u'form-MAX_NUM_FORMS': [u''],
u'form-INITIAL_FORMS': [u'1'],
u'form-0-name':
[u'REDUCER_SLEEP_TIME'],
u'form-0-value': [u'1'],
u'form-TOTAL_FORMS': [u'1']
},
follow=True)
oozie_jobid = response.context['oozie_workflow'].id
# Wait for a job to be created and fetch job ID
hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
client2 = make_logged_in_client('test_non_superuser',
is_superuser=False,
groupname='test')
grant_access('test_non_superuser', 'test', 'jobbrowser')
response = client2.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id, ))
assert_equal(
"Permission denied. User test_non_superuser cannot delete user %s's job."
% self.username, response.context["error"])
# Make sure that the first map task succeeds before moving on
# This will keep us from hitting timing-related failures
first_mapper = 'm_000000'
start = time.time()
timeout_sec = 60
while first_mapper not in \
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,)).content:
time.sleep(1)
# If this assert fails, something has probably really failed
assert_true(time.time() - start < timeout_sec,
"Timed out waiting for first mapper to complete")
# Kill task
self.client.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id, ))
# It should say killed at some point
response = self.client.get('/jobbrowser/jobs/%s?format=json' %
(hadoop_job_id, ))
html = response.content.lower()
i = 0
while 'killed' not in html and i < 10:
time.sleep(5)
response = self.client.get('/jobbrowser/jobs/%s?format=json' %
(hadoop_job_id, ))
html = response.content.lower()
i += 1
assert_true(views.get_shorter_id(hadoop_job_id) in html)
assert_true('killed' in html, html)
# Exercise select by taskstate
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=running' %
(hadoop_job_id, ))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=killed' %
(hadoop_job_id, ))
# Test single task page
late_task_id = hadoop_job_id.replace('job', 'task') + '_r_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' %
(hadoop_job_id, late_task_id))
assert_false('succeed' in response.content)
assert_true('killed' in response.content)
# The first task should've succeeded
# We use a different method of checking success for this one
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' %
(hadoop_job_id, early_task_id))
assert_true('succeed' in response.content)
assert_false('failed' in response.content)
# Test single attempt page
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
attempt_id = early_task_id.replace('task', 'attempt') + '_0'
response = self.client.get(
'/jobbrowser/jobs/%s/tasks/%s/attempts/%s/logs' %
(hadoop_job_id, early_task_id, attempt_id))
assert_true('syslog' in response.content)
# Test dock jobs
response = self.client.get('/jobbrowser/dock_jobs/')
assert_false('completed' in response.content)
assert_false('failed' in response.content)
def test_kill_job(self):
"""
Test job in kill state.
"""
# Clone design
assert_equal(0, OozieDesign.objects.filter(owner__username=self.username).count())
self.client.post('/jobsub/clone_design/%d' % self.sleep_design_id)
assert_equal(1, OozieDesign.objects.filter(owner__username=self.username).count())
# Run the sleep example, since it doesn't require user home directory
design_id = OozieDesign.objects.get(owner__username=self.username).id
response = self.client.post("/jobsub/submit_design/%d" % (design_id,),
dict(map_sleep_time=1,
num_maps=1,
num_reduces=1,
reduce_sleep_time=1),
follow=True)
oozie_jobid = response.context['jobid']
# Wait for a job to be created and fetch job ID
hadoop_job_id = get_hadoop_job_id(self.oozie, oozie_jobid, 1)
client2 = make_logged_in_client('test_non_superuser', is_superuser=False, groupname='test')
grant_access('test_non_superuser', 'test', 'jobbrowser')
response = client2.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id,))
assert_equal("Permission denied. User test_non_superuser cannot delete user %s's job." % self.username, response.context["error"])
# Make sure that the first map task succeeds before moving on
# This will keep us from hitting timing-related failures
first_mapper = hadoop_job_id.replace('job', 'task') + '_m_000000'
start = time.time()
timeout_sec = 60
while first_mapper not in \
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,)).content:
time.sleep(1)
# If this assert fails, something has probably really failed
assert_true(time.time() - start < timeout_sec,
"Timed out waiting for first mapper to complete")
# Kill task
self.client.post('/jobbrowser/jobs/%s/kill' % (hadoop_job_id,))
# It should say killed
response = self.client.get('/jobbrowser/jobs/%s' % (hadoop_job_id,))
html = response.content.lower()
assert_true(hadoop_job_id in html)
assert_true('killed' in html)
# Exercise select by taskstate
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=failed' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=succeeded' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=running' % (hadoop_job_id,))
self.client.get('/jobbrowser/jobs/%s/tasks?taskstate=killed' % (hadoop_job_id,))
# Test single task page
late_task_id = hadoop_job_id.replace('job', 'task') + '_r_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, late_task_id))
assert_false('succeed' in response.content)
assert_true('killed' in response.content)
# The first task should've succeeded
# We use a different method of checking success for this one
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s' % (hadoop_job_id, early_task_id))
assert_true('succeed' in response.content)
assert_false('failed' in response.content)
# Test single attempt page
early_task_id = hadoop_job_id.replace('job', 'task') + '_m_000000'
attempt_id = early_task_id.replace('task', 'attempt') + '_0'
response = self.client.get('/jobbrowser/jobs/%s/tasks/%s/attempts/%s/logs' %
(hadoop_job_id, early_task_id, attempt_id))
assert_true('syslog' in response.content)
# Test dock jobs
response = self.client.get('/jobbrowser/dock_jobs/')
assert_false('completed' in response.content)
assert_false('failed' in response.content)
def setUp(self):
self.c = make_logged_in_client(is_superuser=False)
grant_access("test", "test", "pig")
self.user = User.objects.get(username='******')
