def test_offline_brick_status_when_quorum_not_met(self):
"""
Test Brick status when Quorum is not met after glusterd restart.
1. Create a volume and mount it.
2. Set the quorum type to 'server'.
3. Bring some nodes down such that quorum won't be met.
4. Brick status should be offline in the node which is up.
5. Restart glusterd in this node.
6. The brick status still should be offline as quorum isn't met.
"""
# Set the quorum type to server and validate it.
vol_option = {'cluster.server-quorum-type': 'server'}
ret = set_volume_options(self.mnode, self.volname, vol_option)
self.assertTrue(ret, "gluster volume option set of %s to %s failed"
% ('cluster.server-quorum-type', 'server'))
g.log.info("Cluster quorum set to type server.")
# Get the brick list.
brick_list = get_all_bricks(self.mnode, self.volname)
# Stop glusterd processes.
ret = stop_glusterd(self.servers[1:])
self.assertTrue(ret, "Failed to stop glusterd on specified nodes.")
g.log.info("Glusterd processes stopped in the desired servers.")
# Get the brick status in a node where glusterd is up.
ret = are_bricks_offline(self.mnode, self.volname, brick_list[0:1])
self.assertTrue(ret, "Bricks are online")
g.log.info("Bricks are offline as expected.")
# Restart one of the node which is up.
ret = restart_glusterd(self.servers[0])
self.assertTrue(ret, ("Failed to restart glusterd on desired node."))
g.log.info("Glusterd restarted on the desired node.")
# Wait for glusterd to be online and validate it's running.
self.assertTrue(wait_for_glusterd_to_start(self.servers[0]),
"Glusterd not up on the desired server.")
g.log.info("Glusterd is up in the desired server.")
# Get the brick status from the restarted node.
ret = are_bricks_offline(self.mnode, self.volname, brick_list[0:1])
self.assertTrue(ret, "Bricks are online")
g.log.info("Bricks are offline as expected.")
# Start glusterd on all servers.
ret = start_glusterd(self.servers)
self.assertTrue(ret, "Failed to start glusterd on the specified nodes")
g.log.info("Initiated start of glusterd on all nodes.")
# Wait for glusterd to start.
ret = wait_for_glusterd_to_start(self.servers)
self.assertTrue(ret, "Glusterd not up on all nodes.")
g.log.info("Glusterd is up and running on all nodes.")
# Wait for all volume processes to be online
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname,
timeout=600)
self.assertTrue(ret, ("All volume processes not up."))
g.log.info("All volume processes are up.")
def toggle_bricks_and_perform_io(self, file_list, brick_list):
"""
Kills bricks, does I/O and brings the brick back up.
"""
# Bring down bricks.
g.log.info("Going to bring down the brick process for %s", brick_list)
ret = bring_bricks_offline(self.volname, brick_list)
self.assertTrue(ret, ("Failed to bring down the bricks. Please "
"check the log file for more details."))
g.log.info("Brought down the brick process "
"for %s successfully", brick_list)
ret = are_bricks_offline(self.mnode, self.volname, brick_list)
self.assertTrue(ret, 'Bricks %s are not offline' % brick_list)
# Perform I/O
for filename in file_list:
fpath = self.mounts[0].mountpoint + "/test_gfid_split_brain/" + \
filename
cmd = ("dd if=/dev/urandom of=%s bs=1024 count=1" % fpath)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "Creating %s failed" % fpath)
# Bring up bricks
ret = bring_bricks_online(self.mnode, self.volname, brick_list)
self.assertTrue(ret, 'Failed to bring brick %s online' % brick_list)
g.log.info('Bringing brick %s online is successful', brick_list)
# Waiting for bricks to come online
g.log.info("Waiting for brick process to come online")
timeout = 30
ret = wait_for_bricks_to_be_online(self.mnode, self.volname, timeout)
self.assertTrue(ret, "bricks didn't come online after adding bricks")
g.log.info("Bricks are online")
def _test_brick_down_with_file_rename(self, pfile, rfile, brick):
# Bring brick offline
g.log.info('Bringing brick %s offline', brick)
ret = bring_bricks_offline(self.volname, brick)
self.assertTrue(ret, 'Failed to bring brick %s offline'
% brick)
ret = are_bricks_offline(self.mnode, self.volname,
[brick])
self.assertTrue(ret, 'Brick %s is not offline'
% brick)
g.log.info('Bringing brick %s offline is successful',
brick)
# Rename file
cmd = ("mv %s/%s %s/%s"
% (self.mounts[0].mountpoint, pfile,
self.mounts[0].mountpoint, rfile))
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "rename of file failed")
# Bring brick back online
g.log.info('Bringing brick %s online', brick)
ret = bring_bricks_online(self.mnode, self.volname,
brick)
self.assertTrue(ret, 'Failed to bring brick %s online' %
brick)
g.log.info('Bringing brick %s online is successful', brick)
def _bring_brick_offline_and_check(self, brick):
"""Brings brick offline an checks if it is offline or not"""
ret = bring_bricks_offline(self.volname, [brick])
self.assertTrue(ret, "Unable to bring brick: {} offline".format(brick))
# Validate the brick is offline
ret = are_bricks_offline(self.mnode, self.volname, [brick])
self.assertTrue(ret, "Brick:{} is still online".format(brick))
def _bring_bricks_offline(self):
"""Brings bricks offline and confirms if they are offline"""
# Select bricks to bring offline from a replica set
subvols_dict = get_subvols(self.mnode, self.volname)
subvols = subvols_dict['volume_subvols']
self.bricks_to_bring_offline = []
self.bricks_to_bring_offline.append(choice(subvols[0]))
# Bring bricks offline
ret = bring_bricks_offline(self.volname, self.bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
self.bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
self.bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not offline'
% self.bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
self.bricks_to_bring_offline)
def _bring_redundant_bricks_offline(self, mnode, volname):
"""
Bring redundant bricks offline
"""
brickset_to_offline = []
# List two bricks in each subvol
all_subvols_dict = get_subvols(mnode, volname)
subvols = all_subvols_dict['volume_subvols']
for subvol in subvols:
self.assertTrue(subvol, "List is empty")
brickset_to_offline.extend(sample(subvol, 2))
# Bring two bricks of each subvol down
ret = bring_bricks_offline(volname, brickset_to_offline)
self.assertTrue(ret, "Bricks are still online")
# Validating the bricks are offline
ret = are_bricks_offline(mnode, volname,
brickset_to_offline)
self.assertTrue(ret, "Few of the bricks are still online in"
" {} ".format(brickset_to_offline))
return brickset_to_offline
def test_self_heal_differing_in_file_type(self):
"""
testing self heal of files with different file types
with default configuration
Description:
- create IO
- calculate arequal
- bring down all bricks processes from selected set
- calculate arequal and compare with arequal before
getting bricks offline
- modify the data
- arequal before getting bricks online
- bring bricks online
- check daemons and healing completion
- start healing
- calculate arequal and compare with arequal before bringing bricks
online and after bringing bricks online
"""
# pylint: disable=too-many-locals,too-many-statements
# Creating files on client side
all_mounts_procs = []
test_file_type_differs_self_heal_folder = \
'test_file_type_differs_self_heal'
g.log.info("Generating data for %s:%s",
self.mounts[0].client_system, self.mounts[0].mountpoint)
# Creating files
command = ("cd %s/ ; "
"mkdir %s ;"
"cd %s/ ;"
"for i in `seq 1 10` ; "
"do mkdir l1_dir.$i ; "
"for j in `seq 1 5` ; "
"do mkdir l1_dir.$i/l2_dir.$j ; "
"for k in `seq 1 10` ; "
"do dd if=/dev/urandom of=l1_dir.$i/l2_dir.$j/test.$k "
"bs=1k count=$k ; "
"done ; "
"done ; "
"done ; "
% (self.mounts[0].mountpoint,
test_file_type_differs_self_heal_folder,
test_file_type_differs_self_heal_folder))
proc = g.run_async(self.mounts[0].client_system, command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# wait for io to complete
self.assertTrue(
wait_for_io_to_complete(all_mounts_procs, self.mounts),
"Io failed to complete on some of the clients")
# Get arequal before getting bricks offline
g.log.info('Getting arequal before getting bricks offline...')
ret, result_before_offline = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks offline '
'is successful')
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = bricks_to_bring_offline_dict['volume_bricks']
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Get arequal after getting bricks offline
g.log.info('Getting arequal after getting bricks offline...')
ret, result_after_offline = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks offline '
'is successful')
# Checking arequals before bringing bricks offline
# and after bringing bricks offline
self.assertEqual(sorted(result_before_offline),
sorted(result_after_offline),
'Checksums before and after bringing bricks'
' offline are not equal')
g.log.info('Checksums before and after '
'bringing bricks offline are equal')
# Modify the data
all_mounts_procs = []
g.log.info("Modifying data for %s:%s",
self.mounts[0].client_system, self.mounts[0].mountpoint)
command = ("cd %s/%s/ ; "
"for i in `seq 1 10` ; "
"do for j in `seq 1 5` ; "
"do for k in `seq 1 10` ; "
"do rm -f l1_dir.$i/l2_dir.$j/test.$k ; "
"mkdir l1_dir.$i/l2_dir.$j/test.$k ; "
"done ; "
"done ; "
"done ;"
% (self.mounts[0].mountpoint,
test_file_type_differs_self_heal_folder))
proc = g.run_async(self.mounts[0].client_system, command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# Validate IO
self.assertTrue(
validate_io_procs(all_mounts_procs, self.mounts),
"IO failed on some of the clients"
)
# Get arequal before getting bricks online
g.log.info('Getting arequal before getting bricks online...')
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks online '
'is successful')
# Bring brick online
g.log.info('Bringing bricks %s online', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Wait for volume processes to be online
g.log.info("Wait for volume processes to be online")
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, ("Failed to wait for volume %s processes to "
"be online", self.volname))
g.log.info("Successful in waiting for volume %s processes to be "
"online", self.volname)
# Verify volume's all process are online
g.log.info("Verifying volume's all process are online")
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(ret, ("Volume %s : All process are not online"
% self.volname))
g.log.info("Volume %s : All process are online", self.volname)
# Wait for self-heal-daemons to be online
g.log.info("Waiting for self-heal-daemons to be online")
ret = is_shd_daemonized(self.all_servers)
self.assertTrue(ret, "Either No self heal daemon process found")
g.log.info("All self-heal-daemons are online")
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal after getting bricks online
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks online '
'is successful')
# Checking arequals before bringing bricks online
# and after bringing bricks online
self.assertEqual(sorted(result_before_online),
sorted(result_after_online),
'Checksums before and after bringing bricks'
' online are not equal')
g.log.info('Checksums before and after bringing bricks online '
'are equal')
def test_data_self_heal_algorithm_diff_heal_command(self):
"""
Test Volume Option - 'cluster.data-self-heal-algorithm' : 'diff'
Description:
- set the volume option
"metadata-self-heal": "off"
"entry-self-heal": "off"
"data-self-heal": "off"
"data-self-heal-algorithm": "diff"
"self-heal-daemon": "off"
- create IO
- calculate arequal
- bring down all bricks processes from selected set
- modify the data
- get arequal before getting bricks online
- bring bricks online
- expand volume by adding bricks to the volume
- do rebalance
- set the volume option "self-heal-daemon": "on" and check for daemons
- start healing
- check if heal is completed
- check for split-brain
- calculate arequal and compare with arequal before bringing bricks
offline and after bringing bricks online
"""
# pylint: disable=too-many-branches,too-many-statements
# Setting options
g.log.info('Setting options...')
options = {
"metadata-self-heal": "off",
"entry-self-heal": "off",
"data-self-heal": "off",
"data-self-heal-algorithm": "diff"
}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Options "
"'metadata-self-heal', "
"'entry-self-heal', "
"'data-self-heal', "
"'self-heal-daemon' "
"are set to 'off',"
"'data-self-heal-algorithm' "
"is set to 'diff' successfully")
# Creating files on client side
all_mounts_procs = []
g.log.info("Generating data for %s:%s", self.mounts[0].client_system,
self.mounts[0].mountpoint)
# Creating files
command = "/usr/bin/env python %s create_files -f 100 %s" % (
self.script_upload_path, self.mounts[0].mountpoint)
proc = g.run_async(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# Validate IO
self.assertTrue(validate_io_procs(all_mounts_procs, self.mounts),
"IO failed on some of the clients")
# Setting options
g.log.info('Setting options...')
options = {"self-heal-daemon": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Option 'self-heal-daemon' is set to 'off' successfully")
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = list(
filter(None, (bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks'])))
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s offline' % bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret,
'Bricks %s are not offline' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Modify the data
all_mounts_procs = []
g.log.info("Modifying data for %s:%s", self.mounts[0].client_system,
self.mounts[0].mountpoint)
command = ("/usr/bin/env python %s create_files -f 100 "
"--fixed-file-size 1M %s" %
(self.script_upload_path, self.mounts[0].mountpoint))
proc = g.run_async(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# Validate IO
self.assertTrue(validate_io_procs(all_mounts_procs, self.mounts),
"IO failed on some of the clients")
# Get arequal before getting bricks online
g.log.info('Getting arequal before getting bricks online...')
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks online '
'is successful')
# Bring brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s online' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Expand volume by adding bricks to the volume
g.log.info("Start adding bricks to volume...")
ret = expand_volume(self.mnode, self.volname, self.servers,
self.all_servers_info)
self.assertTrue(ret, ("Failed to expand the volume when IO in "
"progress on volume %s", self.volname))
g.log.info("Expanding volume is successful on volume %s", self.volname)
# Do rebalance
ret, _, _ = rebalance_start(self.mnode, self.volname)
self.assertEqual(ret, 0, 'Failed to start rebalance')
g.log.info('Rebalance is started')
ret = wait_for_rebalance_to_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Rebalance is not completed')
g.log.info('Rebalance is completed successfully')
# Setting options
g.log.info('Setting options...')
options = {"self-heal-daemon": "on"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Option 'self-heal-daemon' is set to 'on' successfully")
# Wait for self-heal-daemons to be online
g.log.info("Waiting for self-heal-daemons to be online")
ret = is_shd_daemonized(self.all_servers)
self.assertTrue(ret, "Either No self heal daemon process found")
g.log.info("All self-heal-daemons are online")
# Start healing
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not started')
g.log.info('Healing is started')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal after getting bricks online
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks online '
'is successful')
# Checking arequals before bringing bricks offline
# and after bringing bricks online
self.assertItemsEqual(result_before_online, result_after_online,
'Checksums are not equal')
g.log.info('Checksums are equal')
def test_dist_to_repl_automatic_heal_should_be_triggered(self):
"""
- create a single brick volume
- add some files and directories
- get arequal from mountpoint
- add-brick such that this brick makes the volume a replica vol 1x2
- make sure heal is completed
- get arequals from all bricks and compare with arequal from mountpoint
- bring down brick 0
- create new files and validate IO
- bring brick 0 up
- make sure heal is completed
"""
# pylint: disable=too-many-statements,too-many-locals
# Start IO on mounts
g.log.info("Starting IO on all mounts...")
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Starting IO on %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
cmd = ("/usr/bin/env python %s create_deep_dirs_with_files "
"--dir-length 1 "
"--dir-depth 1 "
"--max-num-of-dirs 1 "
"--num-of-files 10 %s" %
(self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
g.log.info("IO on %s:%s is started successfully",
mount_obj.client_system, mount_obj.mountpoint)
self.io_validation_complete = False
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Get arequal for mount before adding bricks
g.log.info('Getting arequal before adding bricks...')
ret, arequals = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after healing is successful')
mount_point_total = arequals[0].splitlines()[-1].split(':')[-1]
# Form brick list to add
g.log.info('Forming brick list to add...')
bricks_to_add = form_bricks_list(self.mnode, self.volname, 1,
self.servers, self.all_servers_info)
g.log.info('Brick list to add: %s', bricks_to_add)
# Add bricks
g.log.info("Start adding bricks to volume...")
ret, _, _ = add_brick(self.mnode,
self.volname,
bricks_to_add,
force=True,
replica_count=2)
self.assertFalse(ret, "Failed to add bricks %s" % bricks_to_add)
g.log.info("Adding bricks is successful on volume %s", self.volname)
# Make sure the newly added bricks are available in the volume
# get the bricks for the volume
g.log.info("Fetching bricks for the volume: %s", self.volname)
bricks_list = get_all_bricks(self.mnode, self.volname)
g.log.info("Brick list: %s", bricks_list)
for brick in bricks_to_add:
self.assertIn(brick, bricks_list,
'Brick %s is not in brick list' % brick)
g.log.info('New bricks are present in the volume')
# Make sure volume change from distribute to replicate volume
vol_info_dict = get_volume_type_info(self.mnode, self.volname)
vol_type = vol_info_dict['volume_type_info']['typeStr']
self.assertEqual(
'Replicate', vol_type, 'Volume type is not converted to Replicate '
'after adding bricks')
g.log.info('Volume type is successfully converted to Replicate '
'after adding bricks')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal on bricks and compare with mount_point_total
# It should be the same
g.log.info('Getting arequal on bricks...')
arequals_after_heal = {}
for brick in bricks_list:
g.log.info('Getting arequal on bricks %s...', brick)
node, brick_path = brick.split(':')
command = ('arequal-checksum -p %s '
'-i .glusterfs -i .landfill -i .trashcan' % brick_path)
ret, arequal, _ = g.run(node, command)
self.assertFalse(ret, 'Failed to get arequal on brick %s' % brick)
g.log.info('Getting arequal for %s is successful', brick)
brick_total = arequal.splitlines()[-1].split(':')[-1]
arequals_after_heal[brick] = brick_total
self.assertEqual(
mount_point_total, brick_total,
'Arequals for mountpoint and %s are not equal' % brick)
g.log.info('Arequals for mountpoint and %s are equal', brick)
g.log.info('All arequals are equal for replicated')
# Bring brick 0 offline
g.log.info('Bringing bricks %s offline...', bricks_list[0])
ret = bring_bricks_offline(self.volname, [bricks_list[0]])
self.assertTrue(ret,
'Failed to bring bricks %s offline' % bricks_list[0])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[0]])
self.assertTrue(ret, 'Bricks %s are not offline' % bricks_list[0])
g.log.info('Bringing bricks %s offline is successful', bricks_list[0])
# Start IO on mounts
g.log.info("Starting IO on all mounts...")
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Starting IO on %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
cmd = ("/usr/bin/env python %s create_files -f 100 "
"--fixed-file-size 1k %s" %
(self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
g.log.info("IO on %s:%s is started successfully",
mount_obj.client_system, mount_obj.mountpoint)
self.io_validation_complete = False
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Bring brick 0 online
g.log.info('Bringing bricks %s online...', bricks_list[0])
ret = bring_bricks_online(self.mnode, self.volname, [bricks_list[0]])
self.assertTrue(ret,
'Failed to bring bricks %s online' % bricks_list[0])
g.log.info('Bringing bricks %s online is successful', bricks_list[0])
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
def test_add_brick_when_quorum_not_met(self):
# pylint: disable=too-many-statements
# create and start a volume
ret = setup_volume(self.mnode, self.all_servers_info, self.volume)
self.assertTrue(ret, ("Failed to create "
"and start volume %s" % self.volname))
g.log.info("Volume is created and started successfully")
# set cluster.server-quorum-type as server
ret = set_volume_options(self.mnode, self.volname,
{'cluster.server-quorum-type': 'server'})
self.assertTrue(ret, ("Failed to set the quorum type as a server"
" on volume %s", self.volname))
g.log.info("Able to set server quorum successfully on volume %s",
self.volname)
# Setting quorum ratio to 95%
ret = set_volume_options(self.mnode, 'all',
{'cluster.server-quorum-ratio': '95%'})
self.assertTrue(
ret, "Failed to set server quorum ratio on %s" % self.volname)
g.log.info("Able to set server quorum ratio successfully on %s",
self.servers)
# bring down glusterd of half nodes
num_of_servers = len(self.servers)
num_of_nodes_to_bring_down = num_of_servers / 2
for node in range(num_of_nodes_to_bring_down, num_of_servers):
ret = stop_glusterd(self.servers[node])
self.assertTrue(
ret, ("Failed to stop glusterd on %s" % self.servers[node]))
g.log.info("Glusterd stopped successfully on server %s",
self.servers[node])
for node in range(num_of_nodes_to_bring_down, num_of_servers):
count = 0
while count < 80:
ret = is_glusterd_running(self.servers[node])
if ret:
break
sleep(2)
count += 1
self.assertNotEqual(
ret, 0, "glusterd is still running on %s" % self.servers[node])
# Verifying node count in volume status after glusterd stopped
# on half of the servers, Its not possible to check the brick status
# immediately in volume status after glusterd stop
count = 0
while count < 100:
vol_status = get_volume_status(self.mnode, self.volname)
servers_count = len(vol_status[self.volname])
if servers_count == (num_of_servers - num_of_nodes_to_bring_down):
break
sleep(2)
count += 1
# confirm that quorum is not met, brick process should be down
bricks_list = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, "Failed to get the brick list")
bricks_to_check = bricks_list[0:num_of_nodes_to_bring_down]
ret = are_bricks_offline(self.mnode, self.volname, bricks_to_check)
self.assertTrue(
ret, "Unexpected: Server quorum is not met, "
"Bricks are up")
g.log.info("Server quorum is not met, bricks are down as expected")
# try add brick operation, which should fail
num_bricks_to_add = 1
brick = form_bricks_list(self.mnode, self.volname, num_bricks_to_add,
self.servers, self.all_servers_info)
ret, _, _ = add_brick(self.mnode, self.volname, brick)
self.assertNotEqual(ret, 0, ("Unexpected: add brick is success, "
"when quorum is not met"))
g.log.info("Add brick is failed as expected, when quorum is not met")
# confirm that, newly added brick is not part of volume
bricks_list = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, "Failed to get the brick list")
if brick in bricks_list:
ret = False
self.assertTrue(ret, ("Unexpected: add brick is success, "
"when quorum is not met"))
g.log.info("Add brick is failed as expected, when quorum is not met")
# set cluster.server-quorum-type as none
ret = set_volume_options(self.mnode, self.volname,
{'cluster.server-quorum-type': 'none'})
self.assertTrue(ret, ("Failed to set the quorum type as a server"
" on volume %s", self.volname))
g.log.info("Able to set server quorum successfully on volume %s",
self.volname)
def test_self_heal_daemon(self):
"""
Test Data-Self-Heal(heal command)
Description:
- Create directory test_hardlink_self_heal
- Create directory test_data_self_heal
- Creating files for hardlinks and data files
- Get arequal before getting bricks offline
- Select bricks to bring offline
- Bring brick offline
- Create hardlinks and append data to data files
- Bring brick online
- Wait for volume processes to be online
- Verify volume's all process are online
- Monitor heal completion
- Check for split-brain
- Get arequal after getting bricks online
- Select bricks to bring offline
- Bring brick offline
- Truncate data to data files and verify hardlinks
- Bring brick online
- Wait for volume processes to be online
- Verify volume's all process are online
- Monitor heal completion
- Check for split-brain
- Get arequal again
"""
# pylint: disable=too-many-branches,too-many-statements,too-many-locals
# Creating directory test_hardlink_self_heal
ret = mkdir(
self.mounts[0].client_system,
"{}/test_hardlink_self_heal".format(self.mounts[0].mountpoint))
self.assertTrue(ret, "Failed to create directory")
g.log.info(
"Directory 'test_hardlink_self_heal' on %s created "
"successfully", self.mounts[0])
# Creating directory test_data_self_heal
ret = mkdir(self.mounts[0].client_system,
"{}/test_data_self_heal".format(self.mounts[0].mountpoint))
self.assertTrue(ret, "Failed to create directory")
g.log.info(
"Directory test_hardlink_self_heal on %s created "
"successfully", self.mounts[0])
# Creating files for hardlinks and data files
cmd = ('cd %s/test_hardlink_self_heal;for i in `seq 1 5`;'
'do mkdir dir.$i ; for j in `seq 1 10` ; do dd if='
'/dev/urandom of=dir.$i/file.$j bs=1k count=$j;done; done;'
'cd ..' % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Failed to create file on mountpoint")
g.log.info("Successfully created files on mountpoint")
cmd = ('cd %s/test_data_self_heal;for i in `seq 1 100`;'
'do dd if=/dev/urandom of=file.$i bs=128K count=$i;done;'
'cd ..' % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Failed to create file on mountpoint")
g.log.info("Successfully created files on mountpoint")
# Get arequal before getting bricks offline
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Arequal before getting bricks online-%s',
result_before_online)
# Select bricks to bring offline
bricks_to_bring_offline = select_volume_bricks_to_bring_offline(
self.mnode, self.volname)
self.assertIsNotNone(bricks_to_bring_offline, "List is empty")
# Bring brick offline
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks {} offline'.format(
bricks_to_bring_offline))
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret, 'Bricks {} are not offline'.format(bricks_to_bring_offline))
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Append data to data files and create hardlinks
cmd = ('cd %s/test_data_self_heal;for i in `seq 1 100`;'
'do dd if=/dev/urandom of=file.$i bs=512K count=$i ; done ;'
'cd .. ' % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Failed to modify data files.")
g.log.info("Successfully modified data files")
cmd = ('cd %s/test_hardlink_self_heal;for i in `seq 1 5` ;do '
'for j in `seq 1 10`;do ln dir.$i/file.$j dir.$i/link_file.$j;'
'done ; done ; cd .. ' % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Hardlinks creation failed")
g.log.info("Successfully created hardlinks of files")
# Bring bricks online
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret,
'Failed to bring bricks {} online'.format(bricks_to_bring_offline))
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Wait for volume processes to be online
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, ("Failed to wait for volume {} processes to "
"be online".format(self.volname)))
g.log.info(
"Successful in waiting for volume %s processes to be "
"online", self.volname)
# Verify volume's all process are online
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(
ret,
("Volume {} : All process are not online".format(self.volname)))
g.log.info("Volume %s : All process are online", self.volname)
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal after getting bricks online
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Arequal after getting bricks online '
'is %s', result_after_online)
# Select bricks to bring offline
bricks_to_bring_offline = select_volume_bricks_to_bring_offline(
self.mnode, self.volname)
self.assertIsNotNone(bricks_to_bring_offline, "List is empty")
# Bring brick offline
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks {} offline'.format(
bricks_to_bring_offline))
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret, 'Bricks {} are not offline'.format(bricks_to_bring_offline))
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Truncate data to data files and verify hardlinks
cmd = ('cd %s/test_data_self_heal ; for i in `seq 1 100` ;'
'do truncate -s $(( $i * 128)) file.$i ; done ; cd ..' %
self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Failed to truncate files")
g.log.info("Successfully truncated files on mountpoint")
file_path = ('%s/test_hardlink_self_heal/dir{1..5}/file{1..10}' %
(self.mounts[0].mountpoint))
link_path = ('%s/test_hardlink_self_heal/dir{1..5}/link_file{1..10}' %
(self.mounts[0].mountpoint))
file_stat = get_file_stat(self.mounts[0], file_path)
link_stat = get_file_stat(self.mounts[0], link_path)
self.assertEqual(file_stat, link_stat, "Verification of hardlinks "
"failed")
g.log.info("Successfully verified hardlinks")
# Bring brick online
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret,
'Failed to bring bricks {} online'.format(bricks_to_bring_offline))
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Wait for volume processes to be online
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, ("Failed to wait for volume {} processes to "
"be online".format(self.volname)))
g.log.info(
"Successful in waiting for volume %s processes to be "
"online", self.volname)
# Verify volume's all process are online
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(
ret,
("Volume {} : All process are not online".format(self.volname)))
g.log.info("Volume %s : All process are online", self.volname)
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
def test_self_heal_when_io_in_progress(self):
"""Test self-heal is successful when IO is in progress.
Description:
- simulate brick down.
- bring bricks online
- wait for heal to complete
- validate IO
"""
# Log Volume Info and Status before simulating brick failure
g.log.info(
"Logging volume info and Status before bringing bricks "
"offlien from the volume %s", self.volname)
ret = log_volume_info_and_status(self.mnode, self.volname)
self.assertTrue(ret, ("Logging volume info and status failed on "
"volume %s", self.volname))
g.log.info("Successful in logging volume info and status of volume %s",
self.volname)
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = filter(
None, (bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks']))
# Bring bricks offline
g.log.info("Bringing bricks: %s offline", bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret,
("Failed to bring bricks: %s offline", bricks_to_bring_offline))
g.log.info("Successful in bringing bricks: %s offline",
bricks_to_bring_offline)
# Wait for gluster processes to be offline
time.sleep(10)
# Log Volume Info and Status
g.log.info(
"Logging volume info and Status after bringing bricks "
"offline from the volume %s", self.volname)
ret = log_volume_info_and_status(self.mnode, self.volname)
self.assertTrue(ret, ("Logging volume info and status failed on "
"volume %s", self.volname))
g.log.info("Successful in logging volume info and status of volume %s",
self.volname)
# Validate if bricks are offline
g.log.info("Validating if bricks: %s are offline",
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, "Not all the bricks in list:%s are offline")
g.log.info("Successfully validated that bricks: %s are all offline")
# Add delay before bringing bricks online
time.sleep(40)
# Bring bricks online
g.log.info("Bring bricks: %s online", bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret,
("Failed to bring bricks: %s online", bricks_to_bring_offline))
g.log.info("Successfully brought all bricks:%s online",
bricks_to_bring_offline)
# Wait for gluster processes to be online
time.sleep(10)
# Log Volume Info and Status
g.log.info(
"Logging volume info and Status after bringing bricks "
"online from the volume %s", self.volname)
ret = log_volume_info_and_status(self.mnode, self.volname)
self.assertTrue(ret, ("Logging volume info and status failed on "
"volume %s", self.volname))
g.log.info("Successful in logging volume info and status of volume %s",
self.volname)
# Verify volume's all process are online
g.log.info("Verifying volume's all process are online")
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(
ret, ("Volume %s : All process are not online", self.volname))
g.log.info("Volume %s : All process are online", self.volname)
# Wait for self-heal to complete
g.log.info("Wait for self-heal to complete")
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(
ret, "Self heal didn't complete even after waiting "
"for 20 minutes. 20 minutes is too much a time for "
"current test workload")
g.log.info("self-heal is successful after replace-brick operation")
# Validate IO
g.log.info("Wait for IO to complete and validate IO ...")
ret = validate_io_procs(self.all_mounts_procs, self.mounts)
self.io_validation_complete = True
self.assertTrue(ret, "IO failed on some of the clients")
g.log.info("IO is successful on all mounts")
# List all files and dirs created
g.log.info("List all files and directories:")
ret = list_all_files_and_dirs_mounts(self.mounts)
self.assertTrue(ret, "Failed to list all files and dirs")
g.log.info("Listing all files and directories is successful")
def test_data_split_brain_resolution(self):
# Setting options
g.log.info('Setting options...')
options = {
"metadata-self-heal": "off",
"entry-self-heal": "off",
"data-self-heal": "off"
}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Successfully set %s for volume %s", options, self.volname)
# Creating files and directories on client side
g.log.info('Creating files and directories...')
cmd = ("for i in `seq 1 10`; do mkdir %s/dir.$i; for j in `seq 1 5`;"
"do dd if=/dev/urandom of=%s/dir.$i/file.$j bs=1K count=1;"
"done; dd if=/dev/urandom of=%s/file.$i bs=1K count=1; done" %
(self.mounts[0].mountpoint, self.mounts[0].mountpoint,
self.mounts[0].mountpoint))
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Creating files and directories failed")
g.log.info("Files & directories created successfully")
# Check arequals for all the bricks
g.log.info('Getting arequal before getting bricks offline...')
self.verify_brick_arequals()
g.log.info('Getting arequal before getting bricks offline '
'is successful')
# Set option self-heal-daemon to OFF
g.log.info('Setting option self-heal-daemon to off...')
options = {"self-heal-daemon": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Option 'self-heal-daemon' is set to 'off' successfully")
bricks_list = get_all_bricks(self.mnode, self.volname)
# Bring brick1 offline
g.log.info('Bringing brick %s offline', bricks_list[0])
ret = bring_bricks_offline(self.volname, bricks_list[0])
self.assertTrue(ret,
'Failed to bring bricks %s offline' % bricks_list[0])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[0]])
self.assertTrue(ret, 'Brick %s is not offline' % bricks_list[0])
g.log.info('Bringing brick %s offline is successful', bricks_list[0])
# Modify the contents of the files
cmd = ("for i in `seq 1 10`; do for j in `seq 1 5`;"
"do dd if=/dev/urandom of=%s/dir.$i/file.$j bs=1M count=1;"
"done; dd if=/dev/urandom of=%s/file.$i bs=1K count=1; done" %
(self.mounts[0].mountpoint, self.mounts[0].mountpoint))
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Updating file contents failed")
g.log.info("File contents updated successfully")
# Bricng brick1 online and check the status
g.log.info('Bringing brick %s online', bricks_list[0])
ret = bring_bricks_online(self.mnode, self.volname, [bricks_list[0]])
self.assertTrue(ret,
'Failed to bring brick %s online' % bricks_list[0])
g.log.info('Bringing brick %s online is successful', bricks_list[0])
g.log.info("Verifying if brick %s is online", bricks_list[0])
ret = are_bricks_online(self.mnode, self.volname, bricks_list)
self.assertTrue(ret, ("Brick %s did not come up", bricks_list[0]))
g.log.info("Brick %s has come online.", bricks_list[0])
# Bring brick2 offline
g.log.info('Bringing brick %s offline', bricks_list[1])
ret = bring_bricks_offline(self.volname, bricks_list[1])
self.assertTrue(ret,
'Failed to bring bricks %s offline' % bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[1]])
self.assertTrue(ret, 'Brick %s is not offline' % bricks_list[1])
g.log.info('Bringing brick %s offline is successful', bricks_list[1])
# Modify the contents of the files
cmd = ("for i in `seq 1 10`; do for j in `seq 1 5`;"
"do dd if=/dev/urandom of=%s/dir.$i/file.$j bs=1M count=2;"
"done; dd if=/dev/urandom of=%s/file.$i bs=1K count=2; done" %
(self.mounts[0].mountpoint, self.mounts[0].mountpoint))
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Updating file contents failed")
g.log.info("File contents updated successfully")
# Bricng brick2 online and check the status
g.log.info('Bringing brick %s online', bricks_list[1])
ret = bring_bricks_online(self.mnode, self.volname, [bricks_list[1]])
self.assertTrue(ret,
'Failed to bring brick %s online' % bricks_list[1])
g.log.info('Bringing brick %s online is successful', bricks_list[1])
g.log.info("Verifying if brick %s is online", bricks_list[1])
ret = are_bricks_online(self.mnode, self.volname, bricks_list)
self.assertTrue(ret, ("Brick %s did not come up", bricks_list[1]))
g.log.info("Brick %s has come online.", bricks_list[1])
# Set option self-heal-daemon to ON
g.log.info('Setting option self-heal-daemon to on...')
options = {"self-heal-daemon": "on"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Option 'self-heal-daemon' is set to 'on' successfully")
g.log.info("Checking if files are in split-brain")
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertTrue(ret, "Unable to create split-brain scenario")
g.log.info("Successfully created split brain scenario")
g.log.info("Resolving split-brain by using the source-brick option "
"by choosing second brick as source for all the files")
node, _ = bricks_list[1].split(':')
command = ("gluster v heal " + self.volname + " split-brain "
"source-brick " + bricks_list[1])
ret, _, _ = g.run(node, command)
self.assertEqual(ret, 0, "Command execution not successful")
# triggering heal
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, "Heal not triggered")
# waiting for heal to complete
ret = monitor_heal_completion(self.mnode,
self.volname,
timeout_period=120)
self.assertTrue(ret, "Heal not completed")
# Try accessing the file content from the mount
cmd = ("for i in `seq 1 10`; do cat %s/file.$i > /dev/null;"
"for j in `seq 1 5` ; do cat %s/dir.$i/file.$j > /dev/null;"
"done ; done" %
(self.mounts[0].mountpoint, self.mounts[0].mountpoint))
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Unable to access the file contents")
g.log.info("File contents are accessible")
# checking if file is in split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, "File still in split-brain")
g.log.info("Successfully resolved split brain situation using "
"CLI based resolution")
# Check arequals for all the bricks
g.log.info('Getting arequal for all the bricks after heal...')
self.verify_brick_arequals()
g.log.info('Getting arequal after heal is successful')
def test_metadata_self_heal(self):
"""
Test MetaData Self-Heal (heal command)
Description:
- set the volume option
"metadata-self-heal": "off"
"entry-self-heal": "off"
"data-self-heal": "off"
- create IO
- set the volume option
"self-heal-daemon": "off"
- bring down all bricks processes from selected set
- Change the permissions, ownership and the group
of the files under "test_meta_data_self_heal" folder
- get arequal before getting bricks online
- bring bricks online
- set the volume option
"self-heal-daemon": "on"
- check daemons and start healing
- check is heal is completed
- check for split-brain
- get arequal after getting bricks online and compare with
arequal before getting bricks online
- check group and user are 'qa'
"""
# pylint: disable=too-many-locals,too-many-statements
# Setting options
g.log.info('Setting options...')
options = {"metadata-self-heal": "off",
"entry-self-heal": "off",
"data-self-heal": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Options "
"'metadata-self-heal', "
"'entry-self-heal', "
"'data-self-heal', "
"are set to 'off' successfully")
# Creating files on client side
all_mounts_procs = []
test_meta_data_self_heal_folder = 'test_meta_data_self_heal'
g.log.info("Generating data for %s:%s",
self.mounts[0].client_system, self.mounts[0].mountpoint)
# Create files
g.log.info('Creating files...')
command = ("cd %s/ ; "
"mkdir %s ;"
"cd %s/ ;"
"for i in `seq 1 50` ; "
"do dd if=/dev/urandom of=test.$i bs=10k count=1 ; "
"done ;"
% (self.mounts[0].mountpoint,
test_meta_data_self_heal_folder,
test_meta_data_self_heal_folder))
proc = g.run_async(self.mounts[0].client_system, command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# wait for io to complete
self.assertTrue(
wait_for_io_to_complete(all_mounts_procs, self.mounts),
"Io failed to complete on some of the clients")
# Setting options
g.log.info('Setting options...')
options = {"self-heal-daemon": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Option 'self-heal-daemon' is set to 'off' successfully")
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = list(filter(None, (
bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks'])))
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Changing the permissions, ownership and the group
# of the files under "test_meta_data_self_heal" folder
g.log.info("Modifying data for %s:%s",
self.mounts[0].client_system, self.mounts[0].mountpoint)
# Change permissions to 444
g.log.info('Changing permissions...')
command = ("cd %s/%s/ ; "
"chmod -R 444 *"
% (self.mounts[0].mountpoint,
test_meta_data_self_heal_folder))
ret, out, err = g.run(self.mounts[0].client_system, command)
self.assertEqual(ret, 0, err)
g.log.info('Permissions are changed successfully')
# Change the ownership to qa
g.log.info('Changing the ownership...')
command = ("cd %s/%s/ ; "
"chown -R qa *"
% (self.mounts[0].mountpoint,
test_meta_data_self_heal_folder))
ret, out, err = g.run(self.mounts[0].client_system, command)
self.assertEqual(ret, 0, err)
g.log.info('Ownership is changed successfully')
# Change the group to qa
g.log.info('Changing the group...')
command = ("cd %s/%s/ ; "
"chgrp -R qa *"
% (self.mounts[0].mountpoint,
test_meta_data_self_heal_folder))
ret, out, err = g.run(self.mounts[0].client_system, command)
self.assertEqual(ret, 0, err)
g.log.info('Group is changed successfully')
# Get arequal before getting bricks online
g.log.info('Getting arequal before getting bricks online...')
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks online '
'is successful')
# Bring brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Setting options
g.log.info('Setting options...')
options = {"self-heal-daemon": "on"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Option 'self-heal-daemon' is set to 'on' successfully")
# Wait for volume processes to be online
g.log.info("Wait for volume processes to be online")
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, ("Volume process %s not online "
"despite waiting for 5 minutes", self.volname))
g.log.info("Successful in waiting for volume %s processes to be "
"online", self.volname)
# Verify volume's all process are online
g.log.info("Verifying volume's all process are online")
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(ret, ("Volume %s : All process are not online"
% self.volname))
g.log.info("Volume %s : All process are online", self.volname)
# Wait for self-heal-daemons to be online
g.log.info("Waiting for self-heal-daemons to be online")
ret = is_shd_daemonized(self.all_servers)
self.assertTrue(ret, "Either No self heal daemon process found")
g.log.info("All self-heal-daemons are online")
# Start healing
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not started')
g.log.info('Healing is started')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal after getting bricks online
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks online '
'is successful')
# Checking arequals before bringing bricks online
# and after bringing bricks online
self.assertItemsEqual(result_before_online, result_after_online,
'Checksums are not equal')
g.log.info('Checksums before bringing bricks online '
'and after bringing bricks online are equal')
# Adding servers and client in single dict to check permissions
nodes_to_check = {}
all_bricks = get_all_bricks(self.mnode, self.volname)
for brick in all_bricks:
node, brick_path = brick.split(':')
nodes_to_check[node] = brick_path
nodes_to_check[self.mounts[0].client_system] = \
self.mounts[0].mountpoint
# Checking for user and group
for node in nodes_to_check:
# Get file list
command = ("cd %s/%s/ ; "
"ls"
% (nodes_to_check[node],
test_meta_data_self_heal_folder))
ret, out, err = g.run(node, command)
file_list = out.split()
for file_name in file_list:
file_to_check = '%s/%s/%s' % (nodes_to_check[node],
test_meta_data_self_heal_folder,
file_name)
g.log.info('Checking for permissions, user and group for %s',
file_name)
# Check for permissions
cmd = ("stat -c '%a %n' {} | awk '{{print $1}}'"
.format(file_to_check))
ret, permissions, _ = g.run(node, cmd)
self.assertEqual(permissions.split('\n')[0], '444',
'Permissions %s is not equal to 444'
% permissions)
g.log.info("Permissions are '444' for %s", file_name)
# Check for user
cmd = ("ls -ld {} | awk '{{print $3}}'"
.format(file_to_check))
ret, username, _ = g.run(node, cmd)
self.assertEqual(username.split('\n')[0],
'qa', 'User %s is not equal qa'
% username)
g.log.info("User is 'qa' for %s", file_name)
# Check for group
cmd = ("ls -ld {} | awk '{{print $4}}'"
.format(file_to_check))
ret, groupname, _ = g.run(node, cmd)
self.assertEqual(groupname.split('\n')[0],
'qa', 'Group %s is not equal qa'
% groupname)
g.log.info("Group is 'qa' for %s", file_name)
def test_self_heal(self):
"""
Description:-
- Create files on mount point
- Kill one brick from volume
- rm -rfv on mount point
- bring bricks online
- wait for heals
- list
"""
# pylint: disable=too-many-statements
# IO on the mount point
g.log.info("Starting IO on all mounts...")
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Starting IO on %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
cmd = ("/usr/bin/env python %s create_deep_dirs_with_files "
"--dirname-start-num %d "
"--dir-depth 2 "
"--dir-length 35 "
"--max-num-of-dirs 5 "
"--num-of-files 5 %s" % (
self.script_upload_path,
self.counter, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system, cmd,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.counter = self.counter + 10
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = list(filter(None, (
bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks'])))
# Killing one brick from the volume set
g.log.info("Bringing bricks: %s offline", bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, ("Failed to bring bricks: %s offline",
bricks_to_bring_offline))
g.log.info("Successful in bringing bricks: %s offline",
bricks_to_bring_offline)
# Validate if bricks are offline
g.log.info("Validating if bricks: %s are offline",
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, "Not all the bricks in list: %s are offline" %
bricks_to_bring_offline)
g.log.info("Successfully validated that bricks: %s are all offline",
bricks_to_bring_offline)
# Validate IO
self.assertTrue(
validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients"
)
self.io_validation_complete = True
# Checking volume status
g.log.info("Logging volume info and Status after bringing bricks "
"offline from the volume %s", self.volname)
ret = log_volume_info_and_status(self.mnode, self.volname)
self.assertTrue(ret, ("Logging volume info and status failed on "
"volume %s", self.volname))
g.log.info("Successful in logging volume info and status of volume %s",
self.volname)
# Removing files from the mount point when one brick is down
g.log.info("Removing files from the mount point")
mountpoint = self.mounts[0].mountpoint
client = self.mounts[0].client_system
cmd = "rm -rfv %s/*" % mountpoint
ret, _, _ = g.run(client, cmd)
if ret != 0:
raise ExecutionError("failed to delete the files")
# Bringing bricks online
g.log.info('Bringing bricks %s online', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_to_bring_offline)
g.log.info('Bricks %s are online', bricks_to_bring_offline)
# Check if bricks are online
g.log.info("Checking bricks are online or not")
ret = are_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not online' %
bricks_to_bring_offline)
g.log.info('Bricks %s are online', bricks_to_bring_offline)
# Monitoring heals on the volume
g.log.info("Wait for heal completion...")
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, "Self heal didn't complete even after waiting "
"for 20 minutes.")
g.log.info("self-heal is successful after changing the volume type "
"from replicated to arbitered volume")
# List all files and dirs created
g.log.info("List all files and directories:")
ret = list_all_files_and_dirs_mounts(self.mounts)
self.assertTrue(ret, "Failed to list all files and dirs")
g.log.info("Listing all files and directories is successful")
def test_ec_lookup_and_move_operations_few_bricks_are_offline(self):
"""
Test Steps:
1. Mount this volume on 3 mount point, c1, c2, and c3
2. Bring down two bricks offline in each subvol.
3. On client1: under dir1 create files f{1..10000} run in background
4. On client2: under root dir of mountpoint touch x{1..1000}
5. On client3: after step 4 action completed, start creating
x{1001..10000}
6. Bring bricks online which were offline(brought up all the bricks
which were down (2 in each of the two subvols)
7. While IO on Client1 and Client3 were happening, On client2 move all
the x* files into dir1
8. Perform lookup from client 3
"""
# List two bricks in each subvol
all_subvols_dict = get_subvols(self.mnode, self.volname)
subvols = all_subvols_dict['volume_subvols']
bricks_to_bring_offline = []
for subvol in subvols:
self.assertTrue(subvol, "List is empty")
bricks_to_bring_offline.extend(sample(subvol, 2))
# Bring two bricks of each subvol offline
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, "Bricks are still online")
g.log.info("Bricks are offline %s", bricks_to_bring_offline)
# Validating the bricks are offline or not
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, "Few of the bricks are still online in"
" {} in".format(bricks_to_bring_offline))
g.log.info("%s bricks are offline as expected",
bricks_to_bring_offline)
# Create directory on client1
dir_on_mount = self.mounts[0].mountpoint + '/dir1'
ret = mkdir(self.mounts[0].client_system, dir_on_mount)
self.assertTrue(ret, "unable to create directory on client"
" 1 {}".format(self.mounts[0].client_system))
g.log.info("Dir1 created on %s successfully",
self.mounts[0].client_system)
# Next IO to be ran in the background so using mount_procs
# and run_async.
self.mount_procs = []
# On client1: under dir1 create files f{1..10000} run in background
self._run_create_files(file_count=10000, base_name="f_",
mpoint=dir_on_mount,
client=self.mounts[0].client_system)
# On client2: under root dir of the mountpoint touch x{1..1000}
cmd = ("/usr/bin/env python {} create_files -f 1000 --fixed-file-size"
" 10k --base-file-name x {}".format(self.script_upload_path,
self.mounts[1].mountpoint))
ret, _, err = g.run(self.mounts[1].client_system, cmd)
self.assertEqual(ret, 0, "File creation failed on {} with {}".
format(self.mounts[1].client_system, err))
g.log.info("File creation successful on %s",
self.mounts[1].client_system)
# On client3: start creating x{1001..10000}
cmd = ("cd {}; for i in `seq 1000 10000`; do touch x$i; done; "
"cd -".format(self.mounts[2].mountpoint))
proc = g.run_async(self.mounts[2].client_system, cmd)
self.mount_procs.append(proc)
# Bring bricks online with volume start force
ret, _, err = volume_start(self.mnode, self.volname, force=True)
self.assertEqual(ret, 0, err)
g.log.info("Volume: %s started successfully", self.volname)
# Check whether bricks are online or not
ret = are_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, "Bricks {} are still offline".
format(bricks_to_bring_offline))
g.log.info("Bricks %s are online now", bricks_to_bring_offline)
# From client2 move all the files with name starting with x into dir1
cmd = ("for i in `seq 0 999`; do mv {}/x$i.txt {}; "
"done".format(self.mounts[1].mountpoint, dir_on_mount))
proc = g.run_async(self.mounts[1].client_system, cmd)
self.mount_procs.append(proc)
# Perform a lookup in loop from client3 for 20 iterations
cmd = ("ls -R {}".format(self.mounts[2].mountpoint))
counter = 20
while counter:
ret, _, err = g.run(self.mounts[2].client_system, cmd)
self.assertEqual(ret, 0, "ls while mv operation being carried"
" failed with {}".format(err))
g.log.debug("ls successful for the %s time", 21-counter)
counter -= 1
self.assertTrue(validate_io_procs(self.mount_procs, self.mounts),
"IO failed on the clients")
# Emptying mount_procs for not validating IO in tearDown
self.mount_procs *= 0
# Wait for heal to complete
ret = monitor_heal_completion(self.mnode, self.volname,)
self.assertTrue(ret, "Heal didn't completed in the expected time")
g.log.info("Heal completed successfully on %s volume", self.volname)
def test_multiple_clients_dd_on_same_file_default(self):
"""
- Create 2GB file
- While creating file, start reading file
- Bring down brick1
- Bring back the brick brick1
- Start healing
- Bring down brick1
- Wait for IO to complete
- Wait for reading to complete
- Bring back the brick brick1
- Start healing
- Wait for heal to complete
- Check for split-brain
- Calculate arequals on all the bricks and compare with mountpoint
"""
# pylint: disable=too-many-statements,too-many-locals
bricks_list = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, 'Brick list is None')
# Creating files on client side
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("cd %s ; "
"dd if=/dev/urandom of=test_file bs=1M count=2020"
% mount_obj.mountpoint)
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Reading files on client side
all_mounts_procs_read = []
for mount_obj in self.mounts:
g.log.info("Reading data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Create files
g.log.info('Reading files...')
command = ("python %s read %s"
% (self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
all_mounts_procs_read.append(proc)
# Bring brick1 offline
g.log.info('Bringing bricks %s offline...', bricks_list[1])
ret = bring_bricks_offline(self.volname, [bricks_list[1]])
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_list[1])
g.log.info('Bringing bricks %s offline is successful',
bricks_list[1])
# Bring brick1 online
g.log.info('Bringing bricks %s online...', bricks_list[1])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_list[1])
g.log.info('Bringing bricks %s online is successful',
bricks_list[1])
# Start healing
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not started')
g.log.info('Healing is started')
# Bring brick1 offline
g.log.info('Bringing bricks %s offline...', bricks_list[1])
ret = bring_bricks_offline(self.volname, [bricks_list[1]])
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_list[1])
g.log.info('Bringing bricks %s offline is successful',
bricks_list[1])
# Validate IO
self.assertTrue(
validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients"
)
# Validate reading
self.assertTrue(
validate_io_procs(all_mounts_procs_read, self.mounts),
"Reading failed on some of the clients"
)
self.io_validation_complete = True
# Bring brick1 online
g.log.info('Bringing bricks %s online...', bricks_list[1])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_list[1])
g.log.info('Bringing bricks %s online is successful',
bricks_list[1])
# Start healing
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not started')
g.log.info('Healing is started')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal for mount
g.log.info('Getting arequal...')
ret, arequals = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after healing is successful')
mount_point_total = arequals[0].splitlines()[-1].split(':')[-1]
# Get arequal on bricks and compare with mount_point_total
# It should be the same
g.log.info('Getting arequal on bricks...')
arequals_after_heal = {}
for brick in bricks_list:
g.log.info('Getting arequal on bricks %s...', brick)
node, brick_path = brick.split(':')
command = ('arequal-checksum -p %s '
'-i .glusterfs -i .landfill -i .trashcan'
% brick_path)
ret, arequal, _ = g.run(node, command)
self.assertFalse(ret, 'Failed to get arequal on brick %s'
% brick)
g.log.info('Getting arequal for %s is successful', brick)
brick_total = arequal.splitlines()[-1].split(':')[-1]
arequals_after_heal[brick] = brick_total
self.assertEqual(mount_point_total, brick_total,
'Arequals for mountpoint and %s are not equal'
% brick)
g.log.info('Arequals for mountpoint and %s are equal', brick)
g.log.info('All arequals are equal')
def test_entry_heal_with_quota(self):
"""
- Create a 1x3 volume
- Set quota object limit
- Create files less than the limit
- Bring down a brick and create more files until limit is hit
- Delete one file so that we are below the limit, and create one more
file
- Bring the brick back up and launch heal
- Verify that after heal is complete, the deleted file does not
re-appear in any of the bricks.
"""
# pylint: disable=too-many-statements
# Enable Quota
g.log.info("Enabling quota on the volume %s", self.volname)
ret, _, _ = quota_enable(self.mnode, self.volname)
self.assertEqual(
ret, 0, ("Failed to enable quota on the volume %s", self.volname))
g.log.info("Successfully enabled quota on the volume %s", self.volname)
# Check if quota is enabled
g.log.info("Validate Quota is enabled on the volume %s", self.volname)
ret = is_quota_enabled(self.mnode, self.volname)
self.assertTrue(
ret, ("Quota is not enabled on the volume %s", self.volname))
g.log.info("Successfully Validated quota is enabled on volume %s",
self.volname)
# Set quota related options
options = {
"quota-deem-statfs": "on",
"soft-timeout": "0",
"hard-timeout": "0"
}
g.log.info("setting quota volume options %s", options)
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, ("Unable to set volume option %s for "
"volume %s" % (options, self.volname)))
g.log.info("Successfully set %s for volume %s", options, self.volname)
# Create directory on mount
ret = mkdir(self.mounts[0].client_system,
"%s/dir" % self.mounts[0].mountpoint)
self.assertTrue(ret, "mkdir failed")
# Set Quota limit on the directory
path = "/dir"
g.log.info(
"Setting Quota Limit object on the path %s of the "
"volume %s", path, self.volname)
ret, _, _ = quota_limit_objects(self.mnode,
self.volname,
path=path,
limit="10")
self.assertEqual(ret, 0,
("Failed to set quota limit object "
"on path %s of the volume %s", path, self.volname))
g.log.info(
"Successfully set the Quota limit object on %s of the "
"volume %s", path, self.volname)
cmd = ("touch %s/dir/file{1..5}" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "file creation failed")
# Bring brick3 offline
bricks_list = get_all_bricks(self.mnode, self.volname)
g.log.info('Bringing brick %s offline', bricks_list[2])
ret = bring_bricks_offline(self.volname, bricks_list[2])
self.assertTrue(ret,
'Failed to bring brick %s offline' % bricks_list[2])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[2]])
self.assertTrue(ret, 'Brick %s is not offline' % bricks_list[2])
g.log.info('Bringing brick %s offline was successful', bricks_list[2])
# Create files until quota object limit
cmd = ("touch %s/dir/file{6..9}" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "file creation failed")
# The next create must fail
cmd = ("touch %s/dir/file10" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(
ret, 1, ("Creation of %s/dir/file10 succeeded while "
"it was not supposed to." % self.mounts[0].mountpoint))
g.log.info(
"Creation of %s/dir/file10 failed as expected due to "
"quota object limit.", self.mounts[0].mountpoint)
# Delete one file and re-try the create to succeed.
cmd = ("rm %s/dir/file1" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "File deletion failed")
cmd = ("touch %s/dir/file10" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.clients[0], cmd)
self.assertEqual(ret, 0, "File creation failed")
# Bring brick3 online and check status
g.log.info('Bringing brick %s online...', bricks_list[2])
ret = bring_bricks_online(self.mnode, self.volname, [bricks_list[2]])
self.assertTrue(ret,
'Failed to bring brick %s online' % bricks_list[2])
g.log.info('Bringing brick %s online is successful', bricks_list[2])
g.log.info("Verifying if brick3 is online....")
ret = are_bricks_online(self.mnode, self.volname, bricks_list)
self.assertTrue(ret, ("brick3 did not come up"))
g.log.info("brick3 has come online.")
# Trigger heal
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Starting heal failed')
g.log.info('Index heal launched')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Verify that file10 did not get recreated on the down brick by an
# accidental conservative merge.
for brick in bricks_list:
node, brick_path = brick.split(':')
ret, _, _ = g.run(node, 'stat %s/dir/file10' % brick_path)
self.assertFalse(ret, 'File present!')
def test_heal_info_no_hang(self):
"""
Testcase steps:
1. Start kernel untar on the mount
2. While untar is going on, kill a brick of the replica.
3. Wait for the untar to be over, resulting in pending heals.
4. Get the approx. number of pending heals and save it
5. Bring the brick back online.
6. Trigger heal
7. Run more I/Os with dd command
8. Run heal info command and check that it completes successfully under
a timeout that is based on the no. of heals in step 4.
"""
self.list_of_io_processes = []
self.linux_untar_dir = "{}/{}".format(self.mounts[0].mountpoint,
"linuxuntar")
ret = mkdir(self.clients[0], self.linux_untar_dir)
self.assertTrue(ret, "Failed to create dir linuxuntar for untar")
# Start linux untar on dir linuxuntar
ret = run_linux_untar(self.clients[0], self.mounts[0].mountpoint,
dirs=tuple(['linuxuntar']))
self.list_of_io_processes += ret
self.is_io_running = True
# Kill brick resulting in heal backlog.
brick_to_bring_offline = random.choice(self.bricks_list)
ret = bring_bricks_offline(self.volname, brick_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline'
% brick_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
[brick_to_bring_offline])
self.assertTrue(ret, 'Bricks %s are not offline'
% brick_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
brick_to_bring_offline)
ret = self._wait_for_untar_completion()
self.assertFalse(ret, "IO didn't complete or failed on client")
self.is_io_running = False
# Get approx. no. of entries to be healed.
cmd = ("gluster volume heal %s statistics heal-count | grep Number "
"| awk '{sum+=$4} END {print sum/2}'" % self.volname)
ret, self.num_entries, _ = g.run(self.mnode, cmd)
self.assertEqual(ret, 0, "Failed to get heal-count statistics")
# Restart the down bricks
ret = bring_bricks_online(self.mnode, self.volname,
[brick_to_bring_offline])
self.assertTrue(ret, 'Failed to bring brick %s online' %
brick_to_bring_offline)
g.log.info('Bringing brick %s online is successful',
brick_to_bring_offline)
# Trigger heal
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Starting heal failed')
g.log.info('Index heal launched')
# Run more I/O
cmd = ("for i in `seq 1 10`; do dd if=/dev/urandom of=%s/file_$i "
"bs=1M count=100; done" % self.mounts[0].mountpoint)
ret = g.run_async(self.mounts[0].client_system, cmd,
user=self.mounts[0].user)
# Get heal info
ret = self._does_heal_info_complete_within_timeout()
self.assertTrue(ret, 'Heal info timed out')
g.log.info('Heal info completed succesfully')
def test_handling_data_split_brain(self):
"""
- create IO
- calculate arequal from mountpoint
- set volume option 'self-heal-daemon' to value "off"
- kill data brick1
- calculate arequal checksum and compare it
- modify files and directories
- bring back all bricks processes online
- kill data brick3
- modify files and directories
- calculate arequal from mountpoint
- bring back all bricks processes online
- run the find command to trigger heal from mountpoint
- set volume option 'self-heal-daemon' to value "on"
- check if heal is completed
- check for split-brain
- read files
- calculate arequal checksum and compare it
"""
# pylint: disable=too-many-locals,too-many-statements
# Creating files on client side
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("cd %s ; "
"for i in `seq 1 10` ; "
"do mkdir dir.$i ; "
"for j in `seq 1 5` ; "
"do dd if=/dev/urandom of=dir.$i/file.$j "
"bs=1K count=1 ; "
"done ; "
"dd if=/dev/urandom of=file.$i bs=1k count=1 ; "
"done"
% mount_obj.mountpoint)
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
g.log.info("Wait for IO to complete and validate IO ...")
ret = validate_io_procs(self.all_mounts_procs, self.mounts)
self.assertTrue(ret, "IO failed on some of the clients")
self.io_validation_complete = True
g.log.info("IO is successful on all mounts")
# Get arequal before getting bricks offline
g.log.info('Getting arequal before getting bricks offline...')
ret, result_before_offline = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks offline '
'is successful')
# Setting options
options = {"self-heal-daemon": "off"}
g.log.info('Setting options %s for volume %s',
options, self.volname)
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Option 'self-heal-daemon' is set to 'off' successfully")
# get the bricks for the volume
g.log.info("Fetching bricks for the volume: %s", self.volname)
bricks_list = get_all_bricks(self.mnode, self.volname)
g.log.info("Brick list: %s", bricks_list)
# Bring brick 1 offline
bricks_to_bring_offline = [bricks_list[0]]
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Get arequal after getting bricks offline
g.log.info('Getting arequal after getting bricks offline...')
ret, result_after_offline = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks offline '
'is successful')
# Comparing arequals before getting bricks offline
# and after getting bricks offline
self.assertEqual(result_before_offline, result_after_offline,
'Arequals before getting bricks offline '
'and after getting bricks offline are not equal')
g.log.info('Arequals before getting bricks offline '
'and after getting bricks offline are equal')
# Modify the data
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Modifying data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Modify files
g.log.info('Modifying files...')
command = ("cd %s ; "
"for i in `seq 1 10` ; "
"do for j in `seq 1 5` ; "
"do dd if=/dev/urandom of=dir.$i/file.$j "
"bs=1M count=1 ; "
"done ; "
"dd if=/dev/urandom of=file.$i bs=1M count=1 ; "
"done"
% mount_obj.mountpoint)
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
g.log.info("Wait for IO to complete and validate IO ...")
ret = validate_io_procs(self.all_mounts_procs, self.mounts)
self.assertTrue(ret, "IO failed on some of the clients")
self.io_validation_complete = True
g.log.info("IO is successful on all mounts")
# Bring 1-st brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Bring brick 3rd offline
bricks_to_bring_offline = [bricks_list[-1]]
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Modify the data
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Modifying data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Create files
g.log.info('Modifying files...')
command = ("cd %s ; "
"for i in `seq 1 10` ; "
"do for j in `seq 1 5` ; "
"do dd if=/dev/urandom of=dir.$i/file.$j "
"bs=1M count=1 ; "
"done ; "
"dd if=/dev/urandom of=file.$i bs=1M count=1 ; "
"done"
% mount_obj.mountpoint)
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
g.log.info("Wait for IO to complete and validate IO ...")
ret = validate_io_procs(self.all_mounts_procs, self.mounts)
self.assertTrue(ret, "IO failed on some of the clients")
self.io_validation_complete = True
g.log.info("IO is successful on all mounts")
# Get arequal before getting bricks online
g.log.info('Getting arequal before getting bricks online...')
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks online '
'is successful')
# Bring 3rd brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret, 'Failed to bring bricks %s online' %
bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Mount and unmount mounts
ret = self.unmount_volume(self.mounts)
self.assertTrue(ret, 'Failed to unmount %s' % self.volname)
ret = self.mount_volume(self.mounts)
self.assertTrue(ret, 'Unable to mount %s' % self.volname)
# Start heal from mount point
g.log.info('Starting heal from mount point...')
for mount_obj in self.mounts:
g.log.info("Start heal for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
command = "/usr/bin/env python %s read %s" % (
self.script_upload_path,
self.mounts[0].mountpoint)
ret, _, err = g.run(mount_obj.client_system, command)
self.assertFalse(ret, err)
g.log.info("Heal triggered for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
g.log.info('Heal triggered for all mountpoints')
# Enable self-heal daemon
ret = enable_self_heal_daemon(self.mnode, self.volname)
self.assertTrue(ret, 'Successfully started self heal daemon')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Reading files
g.log.info('Reading files...')
for mount_obj in self.mounts:
g.log.info("Start reading files for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
command = ('cd %s/ ; '
'for i in `seq 1 10` ; '
'do cat file.$i > /dev/null ; '
'for j in `seq 1 5` ; '
'do cat dir.$i/file.$j > /dev/null ; '
'done ; done'
% mount_obj.mountpoint)
ret, _, err = g.run(mount_obj.client_system, command)
self.assertFalse(ret, err)
g.log.info("Reading files successfully for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
g.log.info('Reading files successfully for all mountpoints')
# Get arequal after getting bricks online
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks online '
'is successful')
# Comparing arequals before getting bricks online
# and after getting bricks online
self.assertEqual(result_before_online, result_after_online,
'Arequals before getting bricks online '
'and after getting bricks online are not equal')
g.log.info('Arequals before getting bricks online '
'and after getting bricks online are equal')
def test_brick_process_not_started_on_read_only_node_disks(self):
"""
* create volume and start
* kill one brick
* start IO
* unmount the brick directory from node
* remount the brick directory with read-only option
* start the volume with "force" option
* check for error 'posix: initializing translator failed' in log file
* remount the brick directory with read-write option
* start the volume with "force" option
* validate IO
"""
# pylint: disable=too-many-locals,too-many-statements
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = filter(
None, (bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks']))
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s offline' % bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret,
'Bricks %s are not offline' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Creating files for all volumes
for mount_obj in self.mounts:
g.log.info("Starting IO on %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
cmd = ("python %s create_files -f 100 %s/%s/test_dir" %
(self.script_upload_path, mount_obj.mountpoint,
mount_obj.client_system))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
# umount brick
brick_node, volume_brick = bricks_to_bring_offline[0].split(':')
node_brick = '/'.join(volume_brick.split('/')[0:3])
g.log.info('Start umount brick %s...', node_brick)
ret, _, _ = g.run(brick_node, 'umount -l %s' % node_brick)
self.assertFalse(ret, 'Failed to umount brick %s' % node_brick)
g.log.info('Successfully umounted %s', node_brick)
# get time before remount the directory and checking logs for error
g.log.info('Getting time before remount the directory and '
'checking logs for error...')
_, time_before_checking_logs, _ = g.run(brick_node, 'date -u +%s')
g.log.info('Time before remount the directory and checking logs - %s',
time_before_checking_logs)
# remount the directory with read-only option
g.log.info('Start remount brick %s with read-only option...',
node_brick)
ret, _, _ = g.run(brick_node, 'mount -o ro %s' % node_brick)
self.assertFalse(ret, 'Failed to remount brick %s' % node_brick)
g.log.info('Successfully remounted %s with read-only option',
node_brick)
# start volume with "force" option
g.log.info('starting volume with "force" option...')
ret, _, _ = volume_start(self.mnode, self.volname, force=True)
self.assertFalse(
ret,
'Failed to start volume %s with "force" option' % self.volname)
g.log.info('Successfully started volume %s with "force" option',
self.volname)
# check logs for an 'initializing translator failed' error
g.log.info(
"Checking logs for an 'initializing translator failed' "
"error for %s brick...", node_brick)
error_msg = 'posix: initializing translator failed'
cmd = ("cat /var/log/glusterfs/bricks/%s-%s-%s.log | "
"grep '%s'" %
(volume_brick.split('/')[-3], volume_brick.split('/')[-2],
volume_brick.split('/')[-1], error_msg))
ret, log_msgs, _ = g.run(brick_node, cmd)
log_msg = log_msgs.rstrip().split('\n')[-1]
self.assertTrue(error_msg in log_msg, 'No errors in logs')
g.log.info('EXPECTED: %s', error_msg)
# get time from log message
log_time_msg = log_msg.split('E')[0][1:-2].split('.')[0]
log_time_msg_converted = calendar.timegm(
time.strptime(log_time_msg, '%Y-%m-%d %H:%M:%S'))
g.log.info('Time_msg from logs - %s ', log_time_msg)
g.log.info('Time from logs - %s ', log_time_msg_converted)
# get time after remount the directory checking logs for error
g.log.info('Getting time after remount the directory and '
'checking logs for error...')
_, time_after_checking_logs, _ = g.run(brick_node, 'date -u +%s')
g.log.info('Time after remount the directory and checking logs - %s',
time_after_checking_logs)
# check time periods
g.log.info('Checking if an error is in right time period...')
self.assertTrue(
int(time_before_checking_logs) <= int(log_time_msg_converted) <=
int(time_after_checking_logs),
'Expected error is not in right time period')
g.log.info('Expected error is in right time period')
# umount brick
g.log.info('Start umount brick %s...', node_brick)
ret, _, _ = g.run(brick_node, 'umount -l %s' % node_brick)
self.assertFalse(ret, 'Failed to umount brick %s' % node_brick)
g.log.info('Successfully umounted %s', node_brick)
# remount the directory with read-write option
g.log.info('Start remount brick %s with read-write option...',
node_brick)
ret, _, _ = g.run(brick_node, 'mount %s' % node_brick)
self.assertFalse(ret, 'Failed to remount brick %s' % node_brick)
g.log.info('Successfully remounted %s with read-write option',
node_brick)
# start volume with "force" option
g.log.info('starting volume with "force" option...')
ret, _, _ = volume_start(self.mnode, self.volname, force=True)
self.assertFalse(
ret,
'Failed to start volume %s with "force" option' % self.volname)
g.log.info('Successfully started volume %s with "force" option',
self.volname)
# Validate IO
g.log.info('Validating IO on all mounts')
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
g.log.info('Successfully Validated IO on all mounts')
self.io_validation_complete = True
def test_heal_info_should_have_fixed_fields(self):
"""
- Create IO
- While IO is creating - bring down a couple of bricks
- Wait for IO to complete
- Bring up the down bricks
- Wait for heal to complete
- Check for fields 'Brick', 'Status', 'Number of entries' in heal info
"""
# Creating files on client side
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("/usr/bin/env python %s create_deep_dirs_with_files "
"-d 2 -l 2 -f 50 %s" %
(self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = list(
filter(None, (bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks'])))
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s offline' % bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret,
'Bricks %s are not offline' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Bring brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s online' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get heal info
g.log.info('Getting heal info...')
heal_info_dicts = get_heal_info_summary(self.mnode, self.volname)
self.assertFalse(ret, 'Failed to get heal info')
g.log.info(heal_info_dicts)
bricks_list = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, 'Brick list is None')
# Check all fields in heal info dict
g.log.info('Checking for all the fields in heal info...')
for brick in bricks_list:
g.log.info('Checking fields for %s', brick)
self.assertEqual(heal_info_dicts[brick]['status'], 'Connected',
'Status is not Connected for brick %s' % brick)
self.assertEqual(heal_info_dicts[brick]['numberOfEntries'], '0',
'numberOfEntries is not 0 for brick %s' % brick)
g.log.info('Successfully checked for all the fields in heal info')
def test_heal_command_unsuccessful_as_bricks_down(self):
"""
- write 2 Gb file on mount
- while write is in progress, kill brick b0
- start heal on the volume (should fail and have error message)
- bring up the brick which was down (b0)
- bring down another brick (b1)
- start heal on the volume (should fail and have error message)
- bring bricks up
- wait for heal to complete
"""
# pylint: disable=too-many-statements
bricks_list = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, 'Brick list is None')
# Creating files on client side
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s",
mount_obj.client_system, mount_obj.mountpoint)
# Create 2 Gb file
g.log.info('Creating files...')
command = ("cd %s ; dd if=/dev/zero of=file1 bs=10M count=200"
% mount_obj.mountpoint)
proc = g.run_async(mount_obj.client_system, command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Bring brick0 offline
g.log.info('Bringing bricks %s offline...', bricks_list[0])
ret = bring_bricks_offline(self.volname, [bricks_list[0]])
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_list[0])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[0]])
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_list[0])
g.log.info('Bringing bricks %s offline is successful',
bricks_list[0])
# Start healing
# Need to use 'gluster volume heal' command to check error message
# after g.run
cmd = "gluster volume heal %s" % self.volname
ret, _, err = g.run(self.mnode, cmd)
self.assertTrue(ret, 'Heal is started')
# Check for error message
self.assertIn("Launching heal operation to perform index self heal on "
"volume %s has been unsuccessful" % self.volname,
err,
"Error message is not present or not valid")
g.log.info('Expected: Healing is not started')
# Bring brick0 online
g.log.info("Bring bricks: %s online", bricks_list[0])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[0]])
self.assertTrue(ret, "Failed to bring bricks: %s online"
% bricks_list[0])
g.log.info("Successfully brought all bricks:%s online",
bricks_list[0])
# Bring brick1 offline
g.log.info('Bringing bricks %s offline...', bricks_list[1])
ret = bring_bricks_offline(self.volname, [bricks_list[1]])
self.assertTrue(ret, 'Failed to bring bricks %s offline' %
bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Bricks %s are not offline'
% bricks_list[1])
g.log.info('Bringing bricks %s offline is successful',
bricks_list[1])
# Start healing
# Need to use 'gluster volume heal' command to check error message
# after g.run
cmd = "gluster volume heal %s" % self.volname
ret, _, err = g.run(self.mnode, cmd)
self.assertTrue(ret, 'Heal is started')
# Check for error message
self.assertIn("Launching heal operation to perform index self heal on "
"volume %s has been unsuccessful" % self.volname,
err,
"Error message is not present or not valid")
g.log.info('Expected: Healing is not started')
# Bring brick 1 online
g.log.info("Bring bricks: %s online", bricks_list[1])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, "Failed to bring bricks: %s online"
% bricks_list[1])
g.log.info("Successfully brought all bricks:%s online",
bricks_list[1])
# Start healing
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not started')
g.log.info('Healing is started')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Validate IO
self.assertTrue(
validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients"
)
self.io_validation_complete = True
def test_remove_brick_status(self):
'''
-> Create volume
-> Enable server quorum on volume
-> Stop glusterd on all nodes except first node
-> Verify brick status of nodes where glusterd is running with
default quorum ratio(51%)
-> Change the cluster.server-quorum-ratio from default to 95%
-> Start glusterd on all servers except last node
-> Verify the brick status again
'''
# Enabling server quorum
ret = set_volume_options(self.mnode, self.volname,
{'cluster.server-quorum-type': 'server'})
self.assertTrue(
ret, "Failed to set server quorum on volume %s" % self.volname)
g.log.info("Able to set server quorum on volume %s successfully",
self.volname)
# Getting brick list
brick_list = get_all_bricks(self.mnode, self.volname)
# Stopping glusterd on remaining servers except first node
ret = stop_glusterd(self.servers[1:])
self.assertTrue(
ret, "Failed to stop gluterd on some of the servers "
"%s" % self.servers[1:])
g.log.info("Glusterd stopped successfully on servers %s",
self.servers[1:])
# Checking brick status for glusterd running nodes with
# default quorum ratio(51%)
ret = are_bricks_offline(self.mnode, self.volname, brick_list[0:1])
self.assertTrue(
ret, "Bricks are online when quorum is in not "
"met condition for %s" % self.volname)
g.log.info(
"Bricks are offline when quorum is in not met "
"condition for %s", self.volname)
# Setting quorum ratio to 95%
ret = set_volume_options(self.mnode, 'all',
{'cluster.server-quorum-ratio': '95%'})
self.assertTrue(
ret, "Failed to set quorum ratio to 95 percentage on "
"servers %s" % self.servers)
g.log.info(
"Able to set server quorum ratio to 95 percentage "
"on servers %s", self.servers)
# Starting glusterd on remaining servers except last node
ret = start_glusterd(self.servers[1:5])
self.assertTrue(
ret, "Failed to start glusterd on some of the servers"
" %s" % self.servers[1:5])
g.log.info(
"Glusterd started successfully on all servers except "
"last node %s", self.servers[1:5])
# Verfiying node count in volume status after glusterd
# started on servers, Its not possible to check the brick status
# immediately after glusterd start, that's why verifying that all
# glusterd started nodes available in gluster volume status or not
count = 0
while count < 50:
vol_status = get_volume_status(self.mnode, self.volname)
servers_count = len(vol_status[self.volname].keys())
if servers_count == 5:
break
sleep(2)
count += 1
# Checking brick status with quorum ratio(95%)
ret = are_bricks_offline(self.mnode, self.volname, brick_list[0:5])
self.assertTrue(
ret, "Bricks are online when quorum is in not "
"met condition for %s" % self.volname)
g.log.info(
"Bricks are offline when quorum is in not met "
"condition for %s", self.volname)
def _perform_brick_ops_and_enable_self_heal(self, op_type):
'''Refactor of steps common to all tests: Brick down and perform
metadata/data operations'''
# First brick in the subvol will always be online and used for self
# heal, so make keys match brick index
self.op_cmd = {
# The operation with key `4` in every op_type will be used for
# final data consistency check
# Metadata Operations (owner and permission changes)
'metadata': {
2:
'''cd {0}; for i in `seq 1 3`; do chown -R qa_all:qa_func \
dir.$i file.$i; chmod -R 555 dir.$i file.$i; done;''',
3:
'''cd {0}; for i in `seq 1 3`; do chown -R :qa_system \
dir.$i file.$i; chmod -R 777 dir.$i file.$i; done;''',
4:
'''cd {0}; for i in `seq 1 6`; do chown -R qa_all:qa_system \
dir.$i file.$i; chmod -R 777 dir.$i file.$i; done;''',
},
# Data Operations (append data to the files)
'data': {
2:
'''cd {0}; for i in `seq 1 3`;
do {1} 2K >> file.$i;
for j in `seq 1 3`;
do {1} 2K >> dir.$i/file.$j; done;
done;''',
3:
'''cd {0}; for i in `seq 1 3`;
do {1} 3K >> file.$i;
for j in `seq 1 3`;
do {1} 3K >> dir.$i/file.$j; done;
done;''',
4:
'''cd {0}; for i in `seq 1 6`;
do {1} 4K >> file.$i;
for j in `seq 1 6`;
do {1} 4K >> dir.$i/file.$j; done;
done;''',
},
# Create files and directories when brick is down with no
# initial IO
'gfid': {
2:
'''cd {0}; for i in `seq 1 3`;
do {1} 2K > file.2.$i; mkdir dir.2.$i;
for j in `seq 1 3`;
do {1} 2K > dir.2.$i/file.2.$j; done;
done;''',
3:
'''cd {0}; for i in `seq 1 3`;
do {1} 2K > file.3.$i; mkdir dir.3.$i;
for j in `seq 1 3`;
do {1} 2K > dir.3.$i/file.3.$j; done;
done;''',
4:
'''cd {0}; for i in `seq 4 6`;
do {1} 2K > file.$i; mkdir dir.$i;
for j in `seq 4 6`;
do {1} 2K > dir.$i/file.$j; done;
done;''',
},
# Create different file type with same name while a brick was down
# with no initial IO and validate failure
'file_type': {
2:
'cd {0}; for i in `seq 1 6`; do {1} 2K > notype.$i; done;',
3:
'cd {0}; for i in `seq 1 6`; do mkdir -p notype.$i; done;',
4:
'''cd {0}; for i in `seq 1 6`;
do {1} 2K > file.$i;
for j in `seq 1 6`;
do mkdir -p dir.$i; {1} 2K > dir.$i/file.$j; done;
done;''',
},
# Create symlinks for files and directories while a brick was down
# Out of 6 files, 6 dirs and 6 files in each dir, symlink
# outer 2 files, inner 2 files in each dir, 2 dirs and
# verify it's a symlink(-L) and linking file exists(-e)
'symlink': {
2:
'''cd {0}; for i in `seq 1 2`;
do ln -sr file.$i sl_file.2.$i;
[ -L sl_file.2.$i ] && [ -e sl_file.2.$i ] || exit -1;
for j in `seq 1 2`;
do ln -sr dir.$i/file.$j dir.$i/sl_file.2.$j; done;
[ -L dir.$i/sl_file.2.$j ] && [ -e dir.$i/sl_file.2.$j ] \
|| exit -1;
done; for k in `seq 3 4`; do ln -sr dir.$k sl_dir.2.$k;
[ -L sl_dir.2.$k ] && [ -e sl_dir.2.$k ] || exit -1;
done;''',
3:
'''cd {0}; for i in `seq 1 2`;
do ln -sr file.$i sl_file.3.$i;
[ -L sl_file.3.$i ] && [ -e sl_file.3.$i ] || exit -1;
for j in `seq 1 2`;
do ln -sr dir.$i/file.$j dir.$i/sl_file.3.$j; done;
[ -L dir.$i/sl_file.3.$j ] && [ -e dir.$i/sl_file.3.$j ] \
|| exit -1;
done; for k in `seq 3 4`; do ln -sr dir.$k sl_dir.3.$k;
[ -L sl_dir.3.$k ] && [ -e sl_dir.3.$k ] || exit -1;
done;''',
4:
'''cd {0}; ln -sr dir.4 sl_dir_new.4; mkdir sl_dir_new.4/dir.1;
{1} 4K >> sl_dir_new.4/dir.1/test_file;
{1} 4K >> sl_dir_new.4/test_file;
''',
},
}
bricks = get_online_bricks_list(self.mnode, self.volname)
self.assertIsNotNone(bricks,
'Not able to get list of bricks in the volume')
# Make first brick always online and start operations from second brick
for index, brick in enumerate(bricks[1:], start=2):
# Bring brick offline
ret = bring_bricks_offline(self.volname, brick)
self.assertTrue(ret, 'Unable to bring {} offline'.format(brick))
self.assertTrue(
are_bricks_offline(self.mnode, self.volname, [brick]),
'Brick {} is not offline'.format(brick))
# Perform file/dir operation
cmd = self.op_cmd[op_type][index].format(self.fqpath, self.io_cmd)
ret, _, err = g.run(self.client, cmd)
if op_type == 'file_type' and index == 3:
# Should fail with ENOTCONN as one brick is down, lookupt can't
# happen and quorum is not met
self.assertNotEqual(
ret, 0, '{0} should fail as lookup fails, quorum is not '
'met'.format(cmd))
self.assertIn(
'Transport', err, '{0} should fail with ENOTCONN '
'error'.format(cmd))
else:
self.assertEqual(ret, 0,
'{0} failed with {1}'.format(cmd, err))
self.assertFalse(err, '{0} failed with {1}'.format(cmd, err))
# Bring brick online
ret = bring_bricks_online(
self.mnode,
self.volname,
brick,
bring_bricks_online_methods='volume_start_force')
self.assertTrue(
are_bricks_online(self.mnode, self.volname, [brick]),
'Brick {} is not online'.format(brick))
# Assert metadata/data operations resulted in pending heals
self.assertFalse(is_heal_complete(self.mnode, self.volname))
# Enable and wait self heal daemon to be online
self.assertTrue(enable_self_heal_daemon(self.mnode, self.volname),
'Enabling self heal daemon failed')
self.assertTrue(
wait_for_self_heal_daemons_to_be_online(self.mnode, self.volname),
'Not all self heal daemons are online')
def test_afr_reset_brick_heal_full(self):
"""
1. Create files/dirs from mount point
2. With IO in progress execute reset-brick start
3. Now format the disk from back-end, using rm -rf <brick path>
4. Execute reset brick commit and check for the brick is online.
5. Issue volume heal using "gluster vol heal <volname> full"
6. Check arequal for all bricks to verify all backend bricks
including the resetted brick have same data
"""
self.all_mounts_procs = []
for count, mount_obj in enumerate(self.mounts):
cmd = ("/usr/bin/env python %s create_deep_dirs_with_files "
"--dirname-start-num %d --dir-depth 3 --dir-length 5 "
"--max-num-of-dirs 5 --num-of-files 5 %s" %
(self.script_upload_path, count, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
all_bricks = get_all_bricks(self.mnode, self.volname)
self.assertIsNotNone(all_bricks, "Unable to fetch bricks of volume")
brick_to_reset = choice(all_bricks)
# Start reset brick
ret, _, err = reset_brick(self.mnode,
self.volname,
src_brick=brick_to_reset,
option="start")
self.assertEqual(ret, 0, err)
g.log.info("Reset brick: %s started", brick_to_reset)
# Validate the brick is offline
ret = are_bricks_offline(self.mnode, self.volname, [brick_to_reset])
self.assertTrue(ret, "Brick:{} is still online".format(brick_to_reset))
# rm -rf of the brick directory
node, brick_path = brick_to_reset.split(":")
ret = rmdir(node, brick_path, force=True)
self.assertTrue(
ret, "Unable to delete the brick {} on "
"node {}".format(brick_path, node))
# Reset brick commit
ret, _, err = reset_brick(self.mnode,
self.volname,
src_brick=brick_to_reset,
option="commit")
self.assertEqual(ret, 0, err)
g.log.info("Reset brick committed successfully")
# Check the brick is online
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(
ret, "Few volume processess are offline for the "
"volume: {}".format(self.volname))
# Trigger full heal
ret = trigger_heal_full(self.mnode, self.volname)
self.assertTrue(ret, "Unable to trigger the heal full command")
# Wait for the heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, "Heal didn't complete in 20 mins time")
# Validate io on the clients
ret = validate_io_procs(self.all_mounts_procs, self.mounts)
self.assertTrue(ret, "IO failed on the mounts")
self.all_mounts_procs *= 0
# Check arequal of the back-end bricks after heal completion
all_subvols = get_subvols(self.mnode, self.volname)['volume_subvols']
for subvol in all_subvols:
ret, arequal_from_subvol = collect_bricks_arequal(subvol)
self.assertTrue(
ret, "Arequal is collected successfully across the"
" bricks in the subvol {}".format(subvol))
self.assertEqual(
len(set(arequal_from_subvol)), 1, "Arequal is "
"same on all the bricks in the subvol")
def test_heal_on_file_appends(self):
"""
Test steps:
- create and mount EC volume 4+2
- start append to a file from client
- bring down one of the bricks (say b1)
- wait for ~minute and bring down another brick (say b2)
- after ~minute bring up first brick (b1)
- check the xattrs 'ec.size', 'ec.version'
- xattrs of online bricks should be same as an indication to heal
"""
# Get bricks list
bricks_list = get_online_bricks_list(self.mnode, self.volname)
self.assertIsNotNone(bricks_list, 'Not able to get bricks list')
# Creating a file, generate and append data to the file
self.file_name = 'test_file'
cmd = ("cd %s ;"
"while true; do "
"cat /dev/urandom | tr -dc [:space:][:print:] "
"| head -c 4K >> %s; sleep 2; "
"done;" % (self.mount_obj.mountpoint, self.file_name))
ret = g.run_async(self.client, cmd, user=self.mount_obj.user)
self.assertIsNotNone(ret, "Not able to start IO on client")
g.log.info('Started generating and appending data to the file')
self.is_io_started = True
# Select 3 bricks, 2 need to be offline and 1 will be healthy
brick_1, brick_2, brick_3 = sample(bricks_list, 3)
# Wait for IO to fill the bricks
sleep(30)
# Bring first brick offline and validate
ret = bring_bricks_offline(self.volname, [brick_1])
self.assertTrue(ret,
'Failed to bring brick {} offline'.format(brick_1))
ret = are_bricks_offline(self.mnode, self.volname, [brick_1])
self.assertTrue(
ret, 'Not able to validate brick {} being '
'offline'.format(brick_1))
g.log.info("Brick %s is brought offline successfully", brick_1)
self.offline_bricks.append(brick_1)
# Wait for IO to fill the bricks
sleep(30)
# Bring second brick offline and validate
ret = bring_bricks_offline(self.volname, [brick_2])
self.assertTrue(ret,
'Failed to bring brick {} offline'.format(brick_2))
ret = are_bricks_offline(self.mnode, self.volname, [brick_2])
self.assertTrue(
ret, 'Not able to validate brick {} being '
'offline'.format(brick_2))
g.log.info("Brick %s is brought offline successfully", brick_2)
self.offline_bricks.append(brick_2)
# Wait for IO to fill the bricks
sleep(30)
# Bring first brick online and validate peer status
ret = bring_bricks_online(
self.mnode,
self.volname, [brick_1],
bring_bricks_online_methods=['glusterd_restart'])
self.assertTrue(ret, 'Not able to bring brick {} '
'online'.format(brick_1))
g.log.info("Offlined brick %s is brought online successfully", brick_1)
ret = self.validate_peers_are_connected()
self.assertTrue(
ret, "Peers are not in connected state after bringing "
"an offline brick to online via `glusterd restart`")
g.log.info("Successfully validated peers are in connected state")
# To catchup onlined brick with healthy bricks
sleep(30)
# Validate the xattr to be same on onlined and healthy bric
online_bricks = get_online_bricks_list(self.mnode, self.volname)
self.assertIsNotNone(online_bricks, 'Unable to fetch online bricks')
g.log.info('All online bricks are fetched successfully')
for xattr in ('trusted.ec.size', 'trusted.ec.version'):
ret = validate_xattr_on_all_bricks([brick_1, brick_3],
self.file_name, xattr)
self.assertTrue(
ret, "{} is not same on all online "
"bricks".format(xattr))
# Get epoch time on the client
ret, prev_ctime, _ = g.run(self.client, 'date +%s')
self.assertEqual(ret, 0, 'Not able to get epoch time from client')
# Headroom for file ctime to get updated
sleep(5)
# Validate file was being apended while checking for xattrs
ret = get_file_stat(
self.client, '{}/{}'.format(self.mount_obj.mountpoint,
self.file_name))
self.assertIsNotNone(ret, "Not able to get stats of the file")
curr_ctime = ret['epoch_ctime']
self.assertGreater(
int(curr_ctime), int(prev_ctime), "Not able "
"to validate data is appended to the file "
"while checking for xaatrs")
g.log.info("Data on all online bricks is healed and consistent")
def test_shd_should_not_crash_executed_heal_info(self):
"""
- set "entry-self-heal", "metadata-self-heal", "data-self-heal" to off
- write a few files
- bring down brick0
- add IO
- do a heal info and check for files pending heal on last 2 bricks
- set "performance.enable-least-priority" to "enable"
- bring down brick1
- set the "quorum-type" to "fixed"
- add IO
- do a heal info and check for files pending heal on the last brick
"""
# pylint: disable=too-many-statements
bricks_list = get_all_bricks(self.mnode, self.volname)
# Setting options
g.log.info('Setting options...')
options = {
"metadata-self-heal": "off",
"entry-self-heal": "off",
"data-self-heal": "off"
}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Successfully set %s for volume %s", options, self.volname)
# Creating files on client side
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("/usr/bin/env python %s create_files -f 10 "
"--fixed-file-size 1M %s" %
(self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Bring brick0 offline
g.log.info('Bringing bricks %s offline', bricks_list[0])
ret = bring_bricks_offline(self.volname, bricks_list[0])
self.assertTrue(ret,
'Failed to bring bricks %s offline' % bricks_list[0])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[0]])
self.assertTrue(ret, 'Bricks %s are not offline' % bricks_list[0])
g.log.info('Bringing bricks %s offline is successful', bricks_list[0])
# Creating files on client side
number_of_files_one_brick_off = '1000'
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("/usr/bin/env python %s create_files "
"-f %s "
"--fixed-file-size 1k "
"--base-file-name new_file "
"%s" %
(self.script_upload_path, number_of_files_one_brick_off,
mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Get heal info
g.log.info("Getting heal info...")
heal_info_data = get_heal_info_summary(self.mnode, self.volname)
self.assertIsNotNone(heal_info_data, 'Failed to get heal info.')
g.log.info('Success in getting heal info')
# Check quantity of file pending heal
for brick in bricks_list[1:]:
self.assertEqual(heal_info_data[brick]['numberOfEntries'],
str(int(number_of_files_one_brick_off) + 1),
'Number of files pending heal is not correct')
# Setting options
g.log.info('Setting options...')
options = {"performance.enable-least-priority": "enable"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Successfully set %s for volume %s", options, self.volname)
# Bring brick1 offline
g.log.info('Bringing bricks %s offline', bricks_list[1])
ret = bring_bricks_offline(self.volname, bricks_list[1])
self.assertTrue(ret,
'Failed to bring bricks %s offline' % bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname, [bricks_list[1]])
self.assertTrue(ret, 'Bricks %s are not offline' % bricks_list[1])
g.log.info('Bringing bricks %s offline is successful', bricks_list[1])
# Setting options
g.log.info('Setting options...')
options = {"quorum-type": "fixed"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Successfully set %s for volume %s", options, self.volname)
# Creating files on client side
number_of_files_two_brick_off = '100'
self.all_mounts_procs = []
for mount_obj in self.mounts:
g.log.info("Generating data for %s:%s", mount_obj.client_system,
mount_obj.mountpoint)
# Create files
g.log.info('Creating files...')
command = ("/usr/bin/env python %s create_files "
"-f %s "
"--fixed-file-size 1k "
"--base-file-name new_new_file "
"%s" %
(self.script_upload_path, number_of_files_two_brick_off,
mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
command,
user=mount_obj.user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
self.assertTrue(validate_io_procs(self.all_mounts_procs, self.mounts),
"IO failed on some of the clients")
self.io_validation_complete = True
# Get heal info
g.log.info("Getting heal info...")
heal_info_data = get_heal_info_summary(self.mnode, self.volname)
self.assertIsNotNone(heal_info_data, 'Failed to get heal info.')
g.log.info('Success in getting heal info')
# Check quantity of file pending heal
number_of_files_to_check = str(
int(number_of_files_one_brick_off) +
int(number_of_files_two_brick_off) + 1)
self.assertEqual(heal_info_data[bricks_list[-1]]['numberOfEntries'],
number_of_files_to_check,
'Number of files pending heal is not correct')
def test_data_self_heal_algorithm_full_default(self):
"""
Test Volume Option - 'cluster.data-self-heal-algorithm' : 'full'
Description:
- set the volume option "data-self-heal-algorithm" to value "full"
- create IO
- bring down all bricks processes from selected set
- modify the data
- calculate arequal
- bring bricks online
- start healing
- calculate arequal and compare with arequal before bringing bricks
offline and after bringing bricks online
"""
# pylint: disable=too-many-locals,too-many-statements
# Setting options
g.log.info('Setting options "data-self-heal-algorithm": "full"...')
options = {"data-self-heal-algorithm": "full"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options')
g.log.info("Option 'data-self-heal-algorithm' is set to 'full' "
"successfully")
# Creating files on client side
all_mounts_procs = []
g.log.info("Generating data for %s:%s", self.mounts[0].client_system,
self.mounts[0].mountpoint)
# Creating files
command = "/usr/bin/env python %s create_files -f 100 %s" % (
self.script_upload_path, self.mounts[0].mountpoint)
proc = g.run_async(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# Validate IO
self.assertTrue(validate_io_procs(all_mounts_procs, self.mounts),
"IO failed on some of the clients")
# Select bricks to bring offline
bricks_to_bring_offline_dict = (select_bricks_to_bring_offline(
self.mnode, self.volname))
bricks_to_bring_offline = list(
filter(None, (bricks_to_bring_offline_dict['hot_tier_bricks'] +
bricks_to_bring_offline_dict['cold_tier_bricks'] +
bricks_to_bring_offline_dict['volume_bricks'])))
# Bring brick offline
g.log.info('Bringing bricks %s offline...', bricks_to_bring_offline)
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s offline' % bricks_to_bring_offline)
ret = are_bricks_offline(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(ret,
'Bricks %s are not offline' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s offline is successful',
bricks_to_bring_offline)
# Modify the data
all_mounts_procs = []
g.log.info("Modifying data for %s:%s", self.mounts[0].client_system,
self.mounts[0].mountpoint)
command = ("/usr/bin/env python %s create_files -f 100 "
"--fixed-file-size 1M %s" %
(self.script_upload_path, self.mounts[0].mountpoint))
proc = g.run_async(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
all_mounts_procs.append(proc)
# Validate IO
self.assertTrue(validate_io_procs(all_mounts_procs, self.mounts),
"IO failed on some of the clients")
# Get arequal before getting bricks online
g.log.info('Getting arequal before getting bricks online...')
ret, result_before_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks online '
'is successful')
# Bring brick online
g.log.info('Bringing bricks %s online...', bricks_to_bring_offline)
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline)
self.assertTrue(
ret, 'Failed to bring bricks %s online' % bricks_to_bring_offline)
g.log.info('Bringing bricks %s online is successful',
bricks_to_bring_offline)
# Wait for volume processes to be online
g.log.info("Wait for volume processes to be online")
ret = wait_for_volume_process_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, ("Failed to wait for volume %s processes to "
"be online", self.volname))
g.log.info(
"Successful in waiting for volume %s processes to be "
"online", self.volname)
# Verify volume's all process are online
g.log.info("Verifying volume's all process are online")
ret = verify_all_process_of_volume_are_online(self.mnode, self.volname)
self.assertTrue(
ret, ("Volume %s : All process are not online" % self.volname))
g.log.info("Volume %s : All process are online", self.volname)
# Wait for self-heal-daemons to be online
g.log.info("Waiting for self-heal-daemons to be online")
ret = is_shd_daemonized(self.all_servers)
self.assertTrue(ret, "Either No self heal daemon process found")
g.log.info("All self-heal-daemons are online")
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, 'Heal has not yet completed')
# Check if heal is completed
ret = is_heal_complete(self.mnode, self.volname)
self.assertTrue(ret, 'Heal is not complete')
g.log.info('Heal is completed successfully')
# Check for split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, 'Volume is in split-brain state')
g.log.info('Volume is not in split-brain state')
# Get arequal after getting bricks online
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_online = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after getting bricks online '
'is successful')
# Checking arequals before bringing bricks online
# and after bringing bricks online
self.assertItemsEqual(result_before_online, result_after_online,
'Checksums are not equal')
g.log.info('Checksums before bringing bricks online '
'and after bringing bricks online are equal')
def test_metadata_split_brain_resolution(self):
# Setting options
g.log.info('Setting options...')
options = {"metadata-self-heal": "off",
"entry-self-heal": "off",
"data-self-heal": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Successfully set %s for volume %s",
options, self.volname)
# Creating files and directories on client side
g.log.info('Creating files and directories...')
cmd = ("mkdir %s/test_metadata_sb && cd %s/test_metadata_sb &&"
"for i in `seq 1 3`; do mkdir dir.$i; for j in `seq 1 5`;"
"do dd if=/dev/urandom of=dir.$i/file.$j bs=1K count=1;"
"done; dd if=/dev/urandom of=file.$i bs=1K count=1; done"
% (self.mounts[0].mountpoint, self.mounts[0].mountpoint))
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Creating files and directories failed")
g.log.info("Files & directories created successfully")
# Check arequals for all the bricks
g.log.info('Getting arequal before getting bricks offline...')
self.verify_brick_arequals()
g.log.info('Getting arequal before getting bricks offline '
'is successful')
# Set option self-heal-daemon to OFF
g.log.info('Setting option self-heal-daemon to off...')
options = {"self-heal-daemon": "off"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Option 'self-heal-daemon' is set to 'off' successfully")
bricks_list = get_all_bricks(self.mnode, self.volname)
# Bring brick1 offline
g.log.info('Bringing brick %s offline', bricks_list[0])
ret = bring_bricks_offline(self.volname, bricks_list[0])
self.assertTrue(ret, 'Failed to bring bricks %s offline'
% bricks_list[0])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[0]])
self.assertTrue(ret, 'Brick %s is not offline'
% bricks_list[0])
g.log.info('Bringing brick %s offline is successful',
bricks_list[0])
# Change metadata of some files & directories
cmd = ("cd %s/test_metadata_sb &&"
"for i in `seq 1 2`; do chmod -R 0555 dir.$i file.$i ; done"
% self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Updating file permissions failed")
g.log.info("File permissions updated successfully")
# Bricng brick1 online and check the status
# Bring brick3 online and check status
g.log.info('Bringing brick %s online', bricks_list[0])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[0]])
self.assertTrue(ret, 'Failed to bring brick %s online' %
bricks_list[0])
g.log.info('Bringing brick %s online is successful', bricks_list[0])
g.log.info("Verifying if brick %s is online", bricks_list[0])
ret = are_bricks_online(self.mnode, self.volname, bricks_list)
self.assertTrue(ret, ("Brick %s did not come up", bricks_list[0]))
g.log.info("Brick %s has come online.", bricks_list[0])
# Bring brick2 offline
g.log.info('Bringing brick %s offline', bricks_list[1])
ret = bring_bricks_offline(self.volname, bricks_list[1])
self.assertTrue(ret, 'Failed to bring bricks %s offline'
% bricks_list[1])
ret = are_bricks_offline(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Brick %s is not offline'
% bricks_list[1])
g.log.info('Bringing brick %s offline is successful',
bricks_list[1])
# Change metadata of same files & directories as before
cmd = ("cd %s/test_metadata_sb &&"
"for i in `seq 1 2` ; do chmod -R 0777 dir.$i file.$i ; done"
% self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Updating file permissions failed")
g.log.info("File permissions updated successfully")
# Bricng brick2 online and check the status
g.log.info('Bringing brick %s online', bricks_list[1])
ret = bring_bricks_online(self.mnode, self.volname,
[bricks_list[1]])
self.assertTrue(ret, 'Failed to bring brick %s online' %
bricks_list[1])
g.log.info('Bringing brick %s online is successful', bricks_list[1])
g.log.info("Verifying if brick %s is online", bricks_list[1])
ret = are_bricks_online(self.mnode, self.volname, bricks_list)
self.assertTrue(ret, ("Brick %s did not come up", bricks_list[1]))
g.log.info("Brick %s has come online.", bricks_list[1])
# Set option self-heal-daemon to ON
g.log.info('Setting option self-heal-daemon to on...')
options = {"self-heal-daemon": "on"}
ret = set_volume_options(self.mnode, self.volname, options)
self.assertTrue(ret, 'Failed to set options %s' % options)
g.log.info("Option 'self-heal-daemon' is set to 'on' successfully")
g.log.info("Checking if files are in split-brain")
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertTrue(ret, "Unable to create split-brain scenario")
g.log.info("Successfully created split brain scenario")
g.log.info("Resolving split-brain by using the source-brick option "
"by choosing second brick as source for all the files")
node, _ = bricks_list[1].split(':')
command = ("gluster v heal " + self.volname + " split-brain "
"source-brick " + bricks_list[1])
ret, _, _ = g.run(node, command)
self.assertEqual(ret, 0, "Command execution not successful")
# waiting for heal to complete
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, "Heal not completed")
# Do lookup on the files from mount
cmd = ("ls -lR %s/test_metadata_sb"
% self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Failed to lookup")
g.log.info("Lookup successful")
# Checking if files are still in split-brain
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertFalse(ret, "File still in split-brain")
g.log.info("Successfully resolved split brain situation using "
"CLI based resolution")
# Check arequals for all the bricks
g.log.info('Getting arequal for all the bricks after heal...')
self.verify_brick_arequals()
g.log.info('Getting arequal after heal is successful')
# Change metadata of same files & directories as before
cmd = ("cd %s/test_metadata_sb &&"
"for i in `seq 1 2` ; do chmod -R 0555 dir.$i file.$i ; done"
% self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Updating file permissions failed")
g.log.info("File permissions updated successfully")
# Do lookup on the mount
cmd = ("find %s | xargs stat" % self.mounts[0].mountpoint)
ret, _, _ = g.run(self.mounts[0].client_system, cmd)
self.assertEqual(ret, 0, "Lookup on the mount failed")
g.log.info("Lookup on the mount is successful")
# Check arequals for all the bricks
g.log.info('Getting arequal for all the bricks...')
self.verify_brick_arequals()
g.log.info('Getting arequal is successful')
