openstack/swift
Code Issues Proposed changes

Log error if a local device not identified in replicator

Browse Source 
Example:
* Different port in config and in ring file.
* Running daemon on server not in ring file.
In both cases replication daemon is running but nothing is replicated.
Error log helps to distinguish a local device can't be identified.

Closes-Bug: 1508228
Change-Id: I99351b7d9946f250b7750df91c13d09352a145ce
This commit is contained in:
Peter Lisák 2015-11-06 10:49:09 +01:00
parent bdb7bf4032
commit 16de32f168
4 changed files with 68 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
swift/common/db_replicator.py
View File

@ -620,10 +620,12 @@ class Replicator(Daemon):
self.logger.error(_('ERROR Failed to get my own IPs?'))
return
self._local_device_ids = set()
found_local = False
for node in self.ring.devs:
if node and is_local_device(ips, self.port,
node['replication_ip'],
node['replication_port']):
found_local = True
if self.mount_check and not ismount(
os.path.join(self.root, node['device'])):
self._add_failure_stats(
@ -640,6 +642,10 @@ class Replicator(Daemon):
if os.path.isdir(datadir):
self._local_device_ids.add(node['id'])
dirs.append((datadir, node['id']))
if not found_local:
self.logger.error("Can't find itself %s with port %s in ring "
"file, not replicating",
", ".join(ips), self.port)
self.logger.info(_('Beginning replication run'))
for part, object_file, node_id in roundrobin_datadirs(dirs):
self.cpool.spawn_n(

6
swift/obj/replicator.py
View File

@ -566,6 +566,7 @@ class ObjectReplicator(Daemon):
[(dev['replication_ip'], dev['device'])
for dev in policy.object_ring.devs if dev])
data_dir = get_data_dir(policy)
found_local = False
for local_dev in [dev for dev in policy.object_ring.devs
if (dev
and is_local_device(ips,
@ -574,6 +575,7 @@ class ObjectReplicator(Daemon):
dev['replication_port'])
and (override_devices is None
or dev['device'] in override_devices))]:
found_local = True
dev_path = join(self.devices_dir, local_dev['device'])
obj_path = join(dev_path, data_dir)
tmp_path = join(dev_path, get_tmp_dir(policy))
@ -626,6 +628,10 @@ class ObjectReplicator(Daemon):
for failure_dev in policy.object_ring.devs
if failure_dev])
continue
if not found_local:
self.logger.error("Can't find itself %s with port %s in ring "
"file, not replicating",
", ".join(ips), self.port)
return jobs
def collect_jobs(self, override_devices=None, override_partitions=None,

31
test/unit/common/test_db_replicator.py
View File

@ -266,10 +266,15 @@ class TestDBReplicator(unittest.TestCase):
db_replicator.ring = FakeRing()
self.delete_db_calls = []
self._patchers = []
# recon cache path
self.recon_cache = mkdtemp()
rmtree(self.recon_cache, ignore_errors=1)
os.mkdir(self.recon_cache)
def tearDown(self):
for patcher in self._patchers:
patcher.stop()
rmtree(self.recon_cache, ignore_errors=1)
def _patch(self, patching_fn, *args, **kwargs):
patcher = patching_fn(*args, **kwargs)
@ -463,9 +468,29 @@ class TestDBReplicator(unittest.TestCase):
{'id': 'a', 'point': -1, 'max_row': 10, 'hash': 'd'},
FakeBroker(), -1)), False)
def test_run_once(self):
replicator = TestReplicator({})
replicator.run_once()
def test_run_once_no_local_device_in_ring(self):
logger = unit.debug_logger('test-replicator')
replicator = TestReplicator({'recon_cache_path': self.recon_cache},
logger=logger)
with patch('swift.common.db_replicator.whataremyips',
return_value=['127.0.0.1']):
replicator.run_once()
expected = [
"Can't find itself 127.0.0.1 with port 1000 "
"in ring file, not replicating",
]
self.assertEqual(expected, logger.get_lines_for_level('error'))
def test_run_once_with_local_device_in_ring(self):
logger = unit.debug_logger('test-replicator')
base = 'swift.common.db_replicator.'
with patch(base + 'whataremyips', return_value=['1.1.1.1']), \
patch(base + 'ring', FakeRingWithNodes()):
replicator = TestReplicator({'bind_port': 6000,
'recon_cache_path': self.recon_cache},
logger=logger)
replicator.run_once()
self.assertFalse(logger.get_lines_for_level('error'))
def test_run_once_no_ips(self):
replicator = TestReplicator({}, logger=unit.FakeLogger())

31
test/unit/obj/test_replicator.py
View File

@ -179,6 +179,10 @@ class TestObjectReplicator(unittest.TestCase):
def setUp(self):
utils.HASH_PATH_SUFFIX = 'endcap'
utils.HASH_PATH_PREFIX = ''
# recon cache path
self.recon_cache = tempfile.mkdtemp()
rmtree(self.recon_cache, ignore_errors=1)
os.mkdir(self.recon_cache)
# Setup a test ring (stolen from common/test_ring.py)
self.testdir = tempfile.mkdtemp()
self.devices = os.path.join(self.testdir, 'node')
@ -200,6 +204,7 @@ class TestObjectReplicator(unittest.TestCase):
def tearDown(self):
self.assertFalse(process_errors)
rmtree(self.testdir, ignore_errors=1)
rmtree(self.recon_cache, ignore_errors=1)
def test_handoff_replication_setting_warnings(self):
conf_tests = [
@ -254,11 +259,27 @@ class TestObjectReplicator(unittest.TestCase):
self.replicator.all_devs_info = set()
self.df_mgr = diskfile.DiskFileManager(self.conf, self.logger)
def test_run_once_no_local_device_in_ring(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
bind_ip='1.1.1.1', recon_cache_path=self.recon_cache,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf,
logger=self.logger)
replicator.run_once()
expected = [
"Can't find itself 1.1.1.1 with port 6000 "
"in ring file, not replicating",
"Can't find itself 1.1.1.1 with port 6000 "
"in ring file, not replicating",
]
self.assertEqual(expected, self.logger.get_lines_for_level('error'))
def test_run_once(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
bind_ip=_ips()[0],
bind_ip=_ips()[0], recon_cache_path=self.recon_cache,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
replicator = object_replicator.ObjectReplicator(conf,
logger=self.logger)
was_connector = object_replicator.http_connect
object_replicator.http_connect = mock_http_connect(200)
cur_part = '0'
@ -286,13 +307,16 @@ class TestObjectReplicator(unittest.TestCase):
with _mock_process(process_arg_checker):
replicator.run_once()
self.assertFalse(process_errors)
self.assertFalse(self.logger.get_lines_for_level('error'))
object_replicator.http_connect = was_connector
# policy 1
def test_run_once_1(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
recon_cache_path=self.recon_cache,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
replicator = object_replicator.ObjectReplicator(conf,
logger=self.logger)
was_connector = object_replicator.http_connect
object_replicator.http_connect = mock_http_connect(200)
cur_part = '0'
@ -322,6 +346,7 @@ class TestObjectReplicator(unittest.TestCase):
side_effect=_ips):
replicator.run_once()
self.assertFalse(process_errors)
self.assertFalse(self.logger.get_lines_for_level('error'))
object_replicator.http_connect = was_connector
def test_check_ring(self):