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_gfid_split_brain_resolution(self):
"""
Description: Simulates gfid split brain on multiple files in a dir and
resolve them via `bigger-file`, `mtime` and `source-brick` methods
Steps:
- Create and mount a replicated volume, create a dir and ~10 data files
- Simulate gfid splits in 9 of the files
- Resolve each 3 set of files using `bigger-file`, `mtime` and
`source-bricks` split-brain resoultion methods
- Trigger and monitor for heal completion
- Validate all the files are healed and arequal matches for bricks in
subvols
"""
io_cmd = 'cat /dev/urandom | tr -dc [:space:][:print:] | head -c '
client, m_point = (self.mounts[0].client_system,
self.mounts[0].mountpoint)
arbiter = self.volume_type.find('arbiter') >= 0
# Disable self-heal daemon and set `quorum-type` option to `none`
ret = set_volume_options(self.mnode, self.volname, {
'self-heal-daemon': 'off',
'cluster.quorum-type': 'none'
})
self.assertTrue(
ret, 'Not able to disable `quorum-type` and '
'`self-heal` daemon volume options')
# Create required dir and files from the mount
split_dir = 'gfid_split_dir'
file_io = ('cd %s; for i in {1..10}; do ' + io_cmd +
' 1M > %s/file$i; done;')
ret = mkdir(client, '{}/{}'.format(m_point, split_dir))
self.assertTrue(ret, 'Unable to create a directory from mount point')
ret, _, _ = g.run(client, file_io % (m_point, split_dir))
# `file{4,5,6}` are re-created every time to be used in `bigger-file`
# resolution method
cmd = 'rm -rf {0}/file{1} && {2} {3}M > {0}/file{1}'
split_cmds = {
1:
';'.join(cmd.format(split_dir, i, io_cmd, 2) for i in range(1, 7)),
2:
';'.join(cmd.format(split_dir, i, io_cmd, 3) for i in range(4, 7)),
3: ';'.join(
cmd.format(split_dir, i, io_cmd, 1) for i in range(4, 10)),
4: ';'.join(
cmd.format(split_dir, i, io_cmd, 1) for i in range(7, 10)),
}
# Get subvols and simulate entry split brain
subvols = get_subvols(self.mnode, self.volname)['volume_subvols']
self.assertTrue(subvols, 'Not able to get list of subvols')
msg = ('Unable to bring files under {} dir to entry split brain while '
'{} are down')
for index, bricks in enumerate(self._get_two_bricks(subvols, arbiter),
1):
# Bring down two bricks from each subvol
ret = bring_bricks_offline(self.volname, list(bricks))
self.assertTrue(ret, 'Unable to bring {} offline'.format(bricks))
ret, _, _ = g.run(client,
'cd {}; {}'.format(m_point, split_cmds[index]))
self.assertEqual(ret, 0, msg.format(split_dir, bricks))
# Bricks will be brought down only two times in case of arbiter and
# bringing remaining files into split brain for `latest-mtime` heal
if arbiter and index == 2:
ret, _, _ = g.run(client,
'cd {}; {}'.format(m_point, split_cmds[4]))
self.assertEqual(ret, 0, msg.format(split_dir, bricks))
# Bring offline bricks online
ret = bring_bricks_online(
self.mnode,
self.volname,
bricks,
bring_bricks_online_methods='volume_start_force')
self.assertTrue(ret, 'Unable to bring {} online'.format(bricks))
# Enable self-heal daemon, trigger heal and assert volume is in split
# brain condition
ret = enable_self_heal_daemon(self.mnode, self.volname)
self.assertTrue(ret, 'Failed to enable self heal daemon')
ret = wait_for_self_heal_daemons_to_be_online(self.mnode, self.volname)
self.assertTrue(ret, 'Not all self heal daemons are online')
ret = trigger_heal(self.mnode, self.volname)
self.assertTrue(ret, 'Unable to trigger index heal on the volume')
ret = is_volume_in_split_brain(self.mnode, self.volname)
self.assertTrue(ret, 'Volume should be in split brain condition')
# Select source brick and take note of files in source brick
stop = len(subvols[0]) - 1 if arbiter else len(subvols[0])
source_bricks = [choice(subvol[0:stop]) for subvol in subvols]
files = [
self._get_files_in_brick(path, split_dir) for path in source_bricks
]
# Resolve `file1, file2, file3` gfid split files using `source-brick`
cmd = ('gluster volume heal ' + self.volname + ' split-brain '
'source-brick {} /' + split_dir + '/{}')
for index, source_brick in enumerate(source_bricks):
for each_file in files[index]:
run_cmd = cmd.format(source_brick, each_file)
self._run_cmd_and_assert(run_cmd)
# Resolve `file4, file5, file6` gfid split files using `bigger-file`
cmd = ('gluster volume heal ' + self.volname +
' split-brain bigger-file /' + split_dir + '/{}')
for each_file in ('file4', 'file5', 'file6'):
run_cmd = cmd.format(each_file)
self._run_cmd_and_assert(run_cmd)
# Resolve `file7, file8, file9` gfid split files using `latest-mtime`
cmd = ('gluster volume heal ' + self.volname +
' split-brain latest-mtime /' + split_dir + '/{}')
for each_file in ('file7', 'file8', 'file9'):
run_cmd = cmd.format(each_file)
self._run_cmd_and_assert(run_cmd)
# Unless `shd` is triggered manually/automatically files will still
# appear in `heal info`
ret = trigger_heal_full(self.mnode, self.volname)
self.assertTrue(ret, 'Unable to trigger full self heal')
# Monitor heal completion
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(
ret, 'All files in volume should be healed after healing files via'
' `source-brick`, `bigger-file`, `latest-mtime` methods manually')
# Validate normal file `file10` and healed files don't differ in
# subvols via an `arequal`
for subvol in subvols:
# Disregard last brick if volume is of arbiter type
ret, arequal = collect_bricks_arequal(subvol[0:stop])
self.assertTrue(
ret, 'Unable to get `arequal` checksum on '
'{}'.format(subvol[0:stop]))
self.assertEqual(
len(set(arequal)), 1, 'Mismatch of `arequal` '
'checksum among {} is identified'.format(subvol[0:stop]))
g.log.info('Pass: Resolution of gfid split-brain via `source-brick`, '
'`bigger-file` and `latest-mtime` methods is complete')
def test_server_side_healing_happens_only_when_glustershd_running(self):
"""
Test Script which verifies that the server side healing must happen
only if the heal daemon is running on the node where source brick
resides.
* Create and start the Replicate volume
* Check the glustershd processes - Only 1 glustershd should be listed
* Bring down the bricks without affecting the cluster
* Create files on volume
* kill the glustershd on node where bricks is running
* bring the bricks up which was killed in previous steps
* check the heal info - heal info must show pending heal info, heal
shouldn't happen since glustershd is down on source node
* issue heal
* trigger client side heal
* heal should complete successfully
"""
# pylint: disable=too-many-locals,too-many-statements,too-many-lines
# Setting Volume options
options = {
"metadata-self-heal": "on",
"entry-self-heal": "on",
"data-self-heal": "on"
}
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)
# Check the self-heal daemon process
ret, pids = get_self_heal_daemon_pid(self.servers)
self.assertTrue(ret, ("Either No self heal daemon process found or "
"more than One self heal daemon process "
"found : %s" % pids))
g.log.info(
"Successful in verifying self heal daemon process"
" on all nodes %s", self.servers)
# Select the bricks to bring offline
bricks_to_bring_offline = (select_volume_bricks_to_bring_offline(
self.mnode, self.volname))
g.log.info("Brick List to bring offline : %s", bricks_to_bring_offline)
# Bring down the selected bricks
ret = bring_bricks_offline(self.volname, bricks_to_bring_offline)
self.assertTrue(ret, "Failed to bring down the bricks")
g.log.info("Brought down the brick process "
"for %s", bricks_to_bring_offline)
# Write files on all mounts
all_mounts_procs, num_files_to_write = [], 100
for mount_obj in self.mounts:
cmd = ("/usr/bin/env python %s create_files "
"-f %s --base-file-name file %s" %
(self.script_upload_path, num_files_to_write,
mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
all_mounts_procs.append(proc)
# Validate IO
ret = validate_io_procs(all_mounts_procs, self.mounts)
self.assertTrue(ret, "IO failed on some of the clients")
g.log.info("IO is successful on all mounts")
# Get online bricks list
online_bricks = get_online_bricks_list(self.mnode, self.volname)
g.log.info("Online Bricks for volume %s : %s", self.volname,
online_bricks)
# Get the nodes where bricks are running
bring_offline_glustershd_nodes = []
for brick in online_bricks:
bring_offline_glustershd_nodes.append(brick.split(":")[0])
g.log.info("self heal deamon on nodes %s to be killed",
bring_offline_glustershd_nodes)
# Kill the self heal daemon process on nodes
ret = bring_self_heal_daemon_process_offline(
bring_offline_glustershd_nodes)
self.assertTrue(
ret, ("Unable to bring self heal daemon process"
" offline for nodes %s" % bring_offline_glustershd_nodes))
g.log.info(
"Sucessfully brought down self heal process for "
"nodes %s", bring_offline_glustershd_nodes)
# Check the heal info
heal_info = get_heal_info_summary(self.mnode, self.volname)
g.log.info("Successfully got heal info %s for the volume %s",
heal_info, self.volname)
# Bring bricks online
ret = bring_bricks_online(self.mnode, self.volname,
bricks_to_bring_offline, 'glusterd_restart')
self.assertTrue(
ret,
("Failed to bring bricks: %s online" % bricks_to_bring_offline))
# Issue heal
ret = trigger_heal_full(self.mnode, self.volname)
self.assertFalse(ret,
("Able to trigger heal on volume %s where "
"self heal daemon is not running" % self.volname))
g.log.info(
"Expected : Unable to trigger heal on volume %s where "
"self heal daemon is not running", self.volname)
# Wait for 130 sec to heal
ret = monitor_heal_completion(self.mnode, self.volname, 130)
self.assertFalse(ret, ("Heal Completed on volume %s" % self.volname))
g.log.info("Expected : Heal pending on volume %s", self.volname)
# Check the heal info
heal_info_after_triggering_heal = get_heal_info_summary(
self.mnode, self.volname)
g.log.info("Successfully got heal info for the volume %s",
self.volname)
# Compare with heal pending with the files wrote
for node in online_bricks:
self.assertGreaterEqual(
int(heal_info_after_triggering_heal[node]['numberOfEntries']),
num_files_to_write,
("Some of the files are healed from source bricks %s where "
"self heal daemon is not running" % node))
g.log.info("EXPECTED: No files are healed from source bricks where "
"self heal daemon is not running")
# Unmount and Mount volume again as volume options were set
# after mounting the volume
for mount_obj in self.mounts:
ret, _, _ = umount_volume(mount_obj.client_system,
mount_obj.mountpoint)
self.assertEqual(ret, 0,
"Failed to unmount %s" % mount_obj.client_system)
ret, _, _ = mount_volume(self.volname,
mtype='glusterfs',
mpoint=mount_obj.mountpoint,
mserver=self.mnode,
mclient=mount_obj.client_system)
self.assertEqual(ret, 0,
"Failed to mount %s" % mount_obj.client_system)
all_mounts_procs = []
for mount_obj in self.mounts:
cmd = ("/usr/bin/env python %s read %s" %
(self.script_upload_path, mount_obj.mountpoint))
proc = g.run_async(mount_obj.client_system,
cmd,
user=mount_obj.user)
all_mounts_procs.append(proc)
# Validate IO
ret = validate_io_procs(all_mounts_procs, self.mounts)
self.assertTrue(ret, "Reads failed on some of the clients")
g.log.info("Reads successful on all mounts")
# Wait for heal to complete
ret = monitor_heal_completion(self.mnode, self.volname)
self.assertTrue(ret, "Unable to heal the pending entries")
g.log.info("Successfully healed the pending entries for volume %s",
self.volname)
def test_manual_heal_full_should_trigger_heal(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 1x3
- Start heal full
- Make sure heal is completed
- Get arequals from all bricks and compare with arequal from mountpoint
"""
# pylint: disable=too-many-statements,too-many-locals
# Start IO on mounts
self.all_mounts_procs = []
for mount_obj in self.mounts:
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
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 for mount 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
bricks_to_add = form_bricks_list(self.mnode, self.volname, 2,
self.servers, self.all_servers_info)
g.log.info('Brick list to add: %s', bricks_to_add)
# Add bricks
ret, _, _ = add_brick(self.mnode, self.volname, bricks_to_add,
replica_count=3)
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
bricks_list = get_all_bricks(self.mnode, self.volname)
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')
# Start healing
ret = trigger_heal_full(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 on bricks and compare with mount_point_total
# It should be the same
ret, arequals = collect_bricks_arequal(bricks_list)
self.assertTrue(ret, 'Failed to get arequal on bricks')
for arequal in arequals:
brick_total = arequal.splitlines()[-1].split(':')[-1]
self.assertEqual(mount_point_total, brick_total,
'Arequals for mountpoint and %s are not equal')
g.log.info('All arequals are equal for replicated')
def test_heal_full_after_deleting_files(self):
"""
- Create IO
- Calculate arequal from mount
- Delete data from backend from the EC volume
- Trigger heal full
- Check if heal is completed
- Check for split-brain
- 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 dirs with file
g.log.info('Creating dirs with file...')
command = ("/usr/bin/env python %s create_deep_dirs_with_files "
"-d 2 -l 2 -n 2 -f 20 %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
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 areequal before deleting the files from brick
g.log.info('Getting areequal before getting bricks offline...')
ret, result_before_killing_procs = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting areequal before getting bricks offline '
'is successful')
subvols = get_subvols(self.mnode, self.volname)['volume_subvols']
# Delete data from backend from the erasure node
for subvol in subvols:
erasure = subvol[-1]
g.log.info('Clearing ec brick %s', erasure)
node, brick_path = erasure.split(':')
ret, _, err = g.run(node, 'cd %s/ ; rm -rf *' % brick_path)
g.log.error('Clearing ec brick %s is unsuccessful', erasure)
self.assertFalse(ret, err)
g.log.info('Clearing data from brick is successful')
# Trigger heal full
ret = trigger_heal_full(self.mnode, self.volname)
# 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 areequal after healing
g.log.info('Getting areequal after getting bricks online...')
ret, result_after_healing = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting areequal after getting bricks online '
'is successful')
# Comparing areequals
self.assertEqual(
result_before_killing_procs, result_after_healing,
'Areequals areequals before before killing arbiter '
'processes and after healing are not equal')
g.log.info('Areequals areequals before before killing arbiter '
'processes and after healing equal')
def test_replace_brick_self_heal_io_in_progress(self):
"""
- Create directory on mount point and write files/dirs
- Create another set of files (1K files)
- While creation of files/dirs are in progress Kill one brick
- Remove the contents of the killed brick(simulating disk replacement)
- When the IO's are still in progress, restart glusterd on the nodes
where we simulated disk replacement to bring back bricks online
- Start volume heal
- Wait for IO's to complete
- Verify whether the files are self-healed
- Calculate arequals of the mount point and all the bricks
"""
# pylint: disable=too-many-locals,too-many-statements,too-many-branches
# Create dirs with files
g.log.info('Creating dirs with file...')
command = ("/usr/bin/env python %s create_deep_dirs_with_files "
"-d 2 -l 2 -n 2 -f 10 %s" %
(self.script_upload_path, self.mounts[0].mountpoint))
ret, _, err = g.run(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
self.assertFalse(ret, err)
g.log.info("IO is successful")
# Creating another set of files (1K files)
self.all_mounts_procs = []
# Create dirs with files
g.log.info('Creating 1K files...')
command = ("/usr/bin/env python %s create_files "
"-f 1500 --fixed-file-size 10k %s" %
(self.script_upload_path, self.mounts[0].mountpoint))
proc = g.run_async(self.mounts[0].client_system,
command,
user=self.mounts[0].user)
self.all_mounts_procs.append(proc)
self.io_validation_complete = False
# Validate IO
ret = validate_io_procs(self.all_mounts_procs, self.mounts[0])
self.assertTrue(ret, "IO failed on some of the clients")
self.io_validation_complete = True
g.log.info("IO is successful on all mounts")
# 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
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)
# Remove the content of the killed bricks
for brick in bricks_to_bring_offline:
brick_node, brick_path = brick.split(':')
# Removing files
command = ('cd %s ; rm -rf *' % brick_path)
ret, _, err = g.run(brick_node, command)
self.assertFalse(ret, err)
g.log.info('Files are deleted on brick %s', brick)
# Bring brick online
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
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
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
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_full(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')
# Check arequals for "replicated"
all_bricks = get_all_bricks(self.mnode, self.volname)
if self.volume_type == "replicated":
# Get arequal after bricks are online
ret, arequals = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal after successfully bringing'
'bricks online.')
mount_point_total = arequals[0].splitlines()[-1].split(':')[-1]
# Get arequal on bricks and compare with mount_point_total
ret, arequals = collect_bricks_arequal(all_bricks)
self.assertTrue(ret, 'Failed to get arequal on bricks')
for arequal in arequals:
brick_total = arequal.splitlines()[-1].split(':')[-1]
self.assertEqual(
mount_point_total, brick_total,
'Arequals for mountpoint and brick '
'are not equal')
g.log.info('Arequals for mountpoint and brick are equal')
# Check arequals for "distributed-replicated"
if self.volume_type == "distributed-replicated":
# Get the subvolumes
subvols_dict = get_subvols(self.mnode, self.volname)
num_subvols = len(subvols_dict['volume_subvols'])
g.log.info("Number of subvolumes in volume %s:", num_subvols)
# Get arequals and compare
for i in range(0, num_subvols):
# Get arequal for first brick
subvol_brick_list = subvols_dict['volume_subvols'][i]
ret, arequal = collect_bricks_arequal(subvol_brick_list[0])
self.assertTrue(ret, 'Failed to get arequal on first brick')
first_brick_total = arequal[0].splitlines()[-1].split(':')[-1]
# Get arequal for every brick and compare with first brick
ret, arequals = collect_bricks_arequal(subvol_brick_list)
self.assertTrue(ret, 'Failed to get arequal on bricks')
for arequal in arequals:
brick_total = arequal.splitlines()[-1].split(':')[-1]
self.assertEqual(
first_brick_total, brick_total,
'Arequals for subvol and brick are '
'not equal')
g.log.info('Arequals for subvol and brick are equal')
def test_heal_full_node_reboot(self):
"""
- Create IO from mountpoint.
- Calculate arequal from mount.
- Delete data from backend from the EC volume.
- Trigger heal full.
- Disable Heal.
- Again Enable and do Heal full.
- Reboot a Node.
- 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 dirs with file
g.log.info('Creating dirs with file...')
command = ("/usr/bin/env python %s create_deep_dirs_with_files "
"-d 2 -l 2 -n 2 -f 20 %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
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 deleting the files from brick
g.log.info('Getting arequal before getting bricks offline...')
ret, result_before_killing_procs = collect_mounts_arequal(self.mounts)
self.assertTrue(ret, 'Failed to get arequal')
g.log.info('Getting arequal before getting bricks offline '
'is successful')
subvols = get_subvols(self.mnode, self.volname)['volume_subvols']
# Delete data from backend from the erasure node
for subvol in subvols:
erasure = subvol[-1]
g.log.info('Clearing ec brick %s', erasure)
node, brick_path = erasure.split(':')
ret, _, err = g.run(node, 'cd %s/ ; rm -rf *' % brick_path)
self.assertFalse(ret, err)
g.log.info('Clearing ec brick %s is successful', erasure)
g.log.info('Clearing data from brick is unsuccessful')
# Trigger heal full
ret = trigger_heal_full(self.mnode, self.volname)
self.assertTrue(ret, 'Unable to trigger full heal.')
# Disable Heal and Enable Heal Full Again
g.log.info("Disabling Healon the Servers")
ret = disable_heal(self.mnode, self.volname)
self.assertTrue(ret, "Disabling Failed")
g.log.info("Healing is Now Disabled")
g.log.info("Enbaling Heal Now")
ret = enable_heal(self.mnode, self.volname)
self.assertTrue(ret, "Enabling Heal failed")
g.log.info("Healing is now enabled")
ret = trigger_heal_full(self.mnode, self.volname)
self.assertTrue(ret, 'Unable to trigger full heal.')
# Reboot A Node
g.log.info("Rebooting Node from the Cluster")
subvols_dict = get_subvols(self.mnode, self.volname)
nodes_to_reboot = []
for subvol in subvols_dict['volume_subvols']:
# Define nodes to reboot
brick_list = subvol[1:2]
for brick in brick_list:
node, brick_path = brick.split(':')
if node not in nodes_to_reboot:
nodes_to_reboot.append(node)
# Reboot nodes on subvol and wait while rebooting
g.log.info("Rebooting the nodes %s", nodes_to_reboot)
ret = reboot_nodes(nodes_to_reboot)
self.assertTrue(ret, 'Failed to reboot nodes %s '
% nodes_to_reboot)
# Check if nodes are online
counter = 0
timeout = 700
_rc = False
while counter < timeout:
ret, reboot_results = are_nodes_online(nodes_to_reboot)
if not ret:
g.log.info("Nodes are offline, Retry after 5 seconds ... ")
sleep(5)
counter = counter + 5
_rc = True
break
if not _rc:
for node in reboot_results:
if not reboot_results[node]:
g.log.error("Node %s is offline even after "
"%d minutes", node, timeout / 60.0)
g.log.info("All nodes %s are up and running", nodes_to_reboot)
# Trigger Heal Full
ret = trigger_heal_full(self.mnode, self.volname)
if not ret:
sleep(10)
ret = trigger_heal_full(self.mnode, self.volname)
self.assertTrue(ret, 'Unable to trigger full heal.')
# 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 healing
g.log.info('Getting arequal after getting bricks online...')
ret, result_after_healing = 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
self.assertEqual(result_before_killing_procs, result_after_healing,
'Arequals before killing arbiter '
'processes and after healing are not equal')
g.log.info('Arequals before killing arbiter '
'processes and after healing are equal')