def run(ceph_cluster, **kw):
"""
CEPH-83571453-RADOS:
Corrupt an object in ec pool followed by
list-inconsistent-* commands
1. create a jerasure ec pool with k=4,m=2
2. create an object in the pool
3. chose primary osd from the acting set and go to the backend
4. corrupt object attrib from the backend
5. run deep-scrub on the pool
6. rados list-inconsistent-pg <pool>
7. rados list-inconsistent-obj <pg>
Args:
ceph_cluster (ceph.ceph.Ceph): ceph cluster
"""
log.info("Running CEPH-83571453")
log.info(run.__doc__)
ceph_nodes = kw.get("ceph_nodes")
config = kw.get("config")
build = config.get("build", config.get("rhbuild"))
mons = []
role = "client"
for mnode in ceph_nodes:
if mnode.role == role:
mons.append(mnode)
ctrlr = mons[0]
log.info("chosing mon {cmon} as ctrlrmon".format(cmon=ctrlr.hostname))
helper = RadosHelper(ctrlr, config, log)
"""create ec pool with k=4, m=2"""
k = 4
m = 2
pname = "eccorrupt_{rand}_{k}_{m}".format(rand=random.randint(0, 10000),
k=k,
m=m)
profile = pname
if build.startswith("4"):
prof_cmd = "osd erasure-code-profile set {profile} k={k} m={m} \
crush-failure-domain=osd".format(profile=profile, k=k, m=m)
else:
prof_cmd = "osd erasure-code-profile set {profile} k={k} m={m} \
ruleset-failure-domain=osd crush-failure-domain=osd".format(
profile=profile, k=k, m=m)
try:
(outbuf, err) = helper.raw_cluster_cmd(prof_cmd)
log.info(outbuf)
log.info("created profile {ec}".format(ec=profile))
except Exception:
log.error("ec profile creation failed")
log.error(traceback.format_exc())
return 1
"""create ec pool"""
try:
helper.create_pool(pname, 1, profile)
log.info("Pool {pname} is create".format(pname=pname))
except Exception:
log.error("failed to create pool")
log.error(traceback.format_exc())
return 1
"""check whether pool exists"""
try:
helper.get_pool_num(pname)
except Exception:
log.error("Unable to find pool")
log.error(traceback.format_exc())
return 1
time.sleep(10)
oname = "OBJ_{pname}".format(pname=pname)
cmd = "osd map {pname} {obj} --format json".format(pname=pname, obj=oname)
(outbuf, err) = helper.raw_cluster_cmd(cmd)
log.info(outbuf)
cmdout = json.loads(outbuf)
targt_pg = cmdout["pgid"]
"""considering primary only as of now because of bug
1544680
"""
targt_osd_id = cmdout["up"][0]
"""write data and take snaps"""
putobj = "sudo rados -p {pool} put {obj} {path}".format(pool=pname,
obj=oname,
path="/etc/hosts")
for i in range(10):
(out, err) = ctrlr.exec_command(cmd=putobj)
snapcmd = "sudo rados mksnap -p {pool} {sname}".format(pool=pname,
sname="snap" +
str(i))
(out, err) = ctrlr.exec_command(cmd=snapcmd)
log.info("put {obj}, snap {snap}".format(obj=oname,
snap="snap" + str(i)))
"""
Goto destination osd, stop the osd
use ceph-objectstore-tool to corrupt
snap info
"""
# target_osd = ceph_cluster.get_osd_by_id(targt_osd_id)
# target_osd_node = target_osd.node
target_osd_hostname = ceph_cluster.get_osd_metadata(targt_osd_id).get(
"hostname")
log.info(target_osd_hostname)
target_osd_node = ceph_cluster.get_node_by_hostname(target_osd_hostname)
cot_environment = target_osd_node
osd_service = ceph_cluster.get_osd_service_name(targt_osd_id)
partition_path = ceph_cluster.get_osd_metadata(targt_osd_id).get(
"osd_data")
helper.kill_osd(target_osd_node, osd_service)
time.sleep(10)
osd_metadata = ceph_cluster.get_osd_metadata(targt_osd_id)
osd_data = osd_metadata.get("osd_data")
osd_journal = osd_metadata.get("osd_journal")
if ceph_cluster.containerized:
docker_image_string = "{docker_registry}/{docker_image}:{docker_tag}".format(
docker_registry=ceph_cluster.ansible_config.get(
"ceph_docker_registry"),
docker_image=ceph_cluster.ansible_config.get("ceph_docker_image"),
docker_tag=ceph_cluster.ansible_config.get(
"ceph_docker_image_tag"),
)
cot_environment = helper.get_mgr_proxy_container(
target_osd_node, docker_image_string)
device_mount_data, err = cot_environment.exec_command(
cmd='mount | grep "{partition_path} "'.format(
partition_path=partition_path),
check_ec=False,
)
if not device_mount_data:
cot_environment.exec_command(
cmd="sudo mount {partition_path} {directory}".format(
partition_path=partition_path, directory=osd_data))
slist_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
--head --op list {obj}".format(osd_data=osd_data,
osd_journal=osd_journal,
obj=oname)
(outbuf, err) = cot_environment.exec_command(cmd=slist_cmd)
log.info(outbuf)
corrupt_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
{outbuf} clear-snapset \
corrupt".format(osd_data=osd_data,
osd_journal=osd_journal,
outbuf="'" + (outbuf) + "'")
(outbuf, err) = cot_environment.exec_command(cmd=corrupt_cmd)
log.info(outbuf)
helper.revive_osd(target_osd_node, osd_service)
time.sleep(10)
run_scrub = "pg deep-scrub {pgid}".format(pgid=targt_pg)
(outbuf, err) = helper.raw_cluster_cmd(run_scrub)
log.info(outbuf)
while "HEALTH_ERR" and "active+clean+inconsistent" not in outbuf:
status = "-s --format json"
(outbuf, err) = helper.raw_cluster_cmd(status)
log.info("HEALTH_ERR found as expected")
log.info("inconsistent foud as expected")
timeout = 300
found = 0
while timeout:
incon_pg = "sudo rados list-inconsistent-pg {pname}".format(
pname=pname)
(outbuf, err) = ctrlr.exec_command(cmd=incon_pg)
log.info(outbuf)
if targt_pg not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("pg not listed as inconsistent")
return 1
timeout = 300
found = 0
while timeout:
incon_obj = "sudo rados list-inconsistent-snapset \
{pg}".format(pg=targt_pg)
(outbuf, err) = ctrlr.exec_command(cmd=incon_obj)
log.info(outbuf)
if oname not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("object is not listed in inconsistent obj")
return 1
return 0
def run(ceph_cluster, **kw):
"""
CEPH-9925 - [RADOS]:
Rewrite a known omap item of a replica and list-inconsistent-obj
Steps:
1. create an object in a replica pool
2. add some omap keys and corresponding values to the object
3. chose one of the replica and using ceph-objectstore-rool corrupt omap key or
value
4. Run deep-scrub - scrub should report inconsistency
5. run rados list-inconsistent-pg <pool> - should list the pg in
which object is inconsistent
6. Run rados list-inconsistent-obj <pg> should report omap digest mismatch error
Args:
ceph_cluster (ceph.ceph.Ceph): ceph cluster
"""
log.info("Running CEPH-9924")
log.info(run.__doc__)
ceph_nodes = kw.get("ceph_nodes")
config = kw.get("config")
mons = []
role = "client"
for mnode in ceph_nodes:
if mnode.role == role:
mons.append(mnode)
ctrlr = mons[0]
log.info("chosing mon {cmon} as ctrlrmon".format(cmon=ctrlr.hostname))
helper = RadosHelper(ctrlr, config, log)
"""create an replica pool"""
pname0 = "replica_pool_{rand}".format(rand=random.randint(0, 10000))
pname = pname0
try:
helper.create_pool(pname, 128)
log.info("Pool {pname} is create".format(pname=pname))
except Exception:
log.error("failed to create pool")
log.error(traceback.format_exc())
return 1
"""check whether pool exists"""
try:
helper.get_pool_num(pname)
except Exception:
log.error("Unable to find pool")
log.error(traceback.format_exc())
return 1
time.sleep(10)
oname = "OBJ_{pname}".format(pname=pname)
putobj = "sudo rados -p {pool} put {obj} {path}".format(pool=pname,
obj=oname,
path="/etc/hosts")
(out, err) = ctrlr.exec_command(cmd=putobj)
""" creating omap key/value pairs for an object"""
for i in range(4):
omapcmd = "sudo rados -p {pool} setomapval {obj} {keey} {valu}".format(
pool=pname, obj=oname, keey="key" + str(i), valu="value" + str(i))
(out, err) = ctrlr.exec_command(cmd=omapcmd)
log.info("put {obj}, omap key {keey} value {valu}".format(
obj=oname, keey="key" + str(i), valu="value" + str(i)))
"""
Goto destination osd, stop the osd service to
use ceph-objectstore-tool to corrupt
omap keys
"""
cmd = "osd map {pname} {obj} --format json".format(pname=pname, obj=oname)
(out, err) = helper.raw_cluster_cmd(cmd)
outbuf = out.read().decode()
log.info(outbuf)
cmdout = json.loads(outbuf)
targt_pg = cmdout["pgid"]
"""Considering non primary osd"""
targt_osd_id = cmdout["up"][1]
# target_osd = ceph_cluster.get_osd_by_id(targt_osd_id)
# target_osd_node = target_osd.node
target_osd_hostname = ceph_cluster.get_osd_metadata(targt_osd_id).get(
"hostname")
log.info(target_osd_hostname)
target_osd_node = ceph_cluster.get_node_by_hostname(target_osd_hostname)
cot_environment = target_osd_node
osd_service = ceph_cluster.get_osd_service_name(targt_osd_id)
partition_path = ceph_cluster.get_osd_metadata(targt_osd_id).get(
"osd_data")
helper.kill_osd(target_osd_node, osd_service)
time.sleep(10)
osd_metadata = ceph_cluster.get_osd_metadata(targt_osd_id)
osd_data = osd_metadata.get("osd_data")
osd_journal = osd_metadata.get("osd_journal")
if ceph_cluster.containerized:
# target_osd_node.exec_command(cmd='sudo yum install -y ceph-osd', check_ec=False)
docker_image_string = "{docker_registry}/{docker_image}:{docker_tag}".format(
docker_registry=ceph_cluster.ansible_config.get(
"ceph_docker_registry"),
docker_image=ceph_cluster.ansible_config.get("ceph_docker_image"),
docker_tag=ceph_cluster.ansible_config.get(
"ceph_docker_image_tag"),
)
cot_environment = helper.get_mgr_proxy_container(
target_osd_node, docker_image_string)
out, err = cot_environment.exec_command(
cmd='mount | grep "{partition_path} "'.format(
partition_path=partition_path),
check_ec=False,
)
device_mount_data = out.read().decode() # type: str
if not device_mount_data:
cot_environment.exec_command(
cmd="sudo mount {partition_path} {directory}".format(
partition_path=partition_path, directory=osd_data))
# docker_image_string = '{docker_registry}/{docker_image}:{docker_tag}'.format(
# docker_registry=ceph_cluster.ansible_config.get('ceph_docker_registry'),
# docker_image=ceph_cluster.ansible_config.get('ceph_docker_image'),
# docker_tag=ceph_cluster.ansible_config.get('ceph_docker_image_tag'))
# mgr_proxy = helper.get_mgr_container_proxy(target_osd_node, docker_image_string)
slist_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
--pgid {pgid} {obj} list-omap".format(osd_data=osd_data,
osd_journal=osd_journal,
obj=oname,
pgid=targt_pg)
(out, err) = cot_environment.exec_command(cmd=slist_cmd)
outbuf = out.read().decode()
keylist = outbuf.split()
log.info(outbuf)
"""corrupting an omap key by rewriting the omap key with different value"""
corrupt_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
--pgid {pgid} {obj} set-omap \
{outbuf} {path}".format(
osd_data=osd_data,
osd_journal=osd_journal,
obj=oname,
pgid=targt_pg,
outbuf=keylist[0],
path="/etc/hosts",
)
(out, err) = cot_environment.exec_command(cmd=corrupt_cmd)
outbuf = out.read().decode()
log.info(outbuf)
helper.revive_osd(target_osd_node, osd_service)
time.sleep(10)
run_scrub = "pg deep-scrub {pgid}".format(pgid=targt_pg)
(out, err) = helper.raw_cluster_cmd(run_scrub)
outbuf = out.read().decode()
log.info(outbuf)
while "HEALTH_ERR" and "active+clean+inconsistent" not in outbuf:
status = "-s --format json"
(out, err) = helper.raw_cluster_cmd(status)
outbuf = out.read().decode()
log.info("HEALTH_ERR found as expected")
log.info("inconsistent found as expected")
timeout = 100
found = 0
while timeout:
incon_pg = "sudo rados list-inconsistent-pg {pname}".format(
pname=pname)
(out, err) = ctrlr.exec_command(cmd=incon_pg)
outbuf = out.read().decode()
log.info(outbuf)
if targt_pg not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("pg not listed as inconsistent")
return 1
timeout = 100
found = 0
while timeout:
incon_obj = "sudo rados list-inconsistent-obj {pg}".format(pg=targt_pg)
(out, err) = ctrlr.exec_command(cmd=incon_obj)
outbuf = out.read().decode()
log.info(outbuf)
if oname not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("object is not listed in inconsistent obj")
return 1
return 0
def run(**kw):
"""
1. Create a LRC profile and then create a ec pool
#ceph osd erasure-code-profile set $profile \
plugin=lrc \
k= m= l= \
ruleset-failure-domain=osd
try different values for k, m and l
# ceph osd pool create $poolname 1 1 erasure $profile
2. perform I/O
#rados put -p poolname obj /path/
"""
log.info("Running CEPH-9322")
log.info(run.__doc__)
ceph_nodes = kw.get('ceph_nodes')
config = kw.get('config')
build = config.get('build', config.get('rhbuild'))
mons = []
osds = []
role = 'client'
for mnode in ceph_nodes:
if mnode.role == role:
mons.append(mnode)
role = 'osd'
for osd in ceph_nodes:
if osd.role == role:
osds.append(osd)
ctrlr = mons[0]
log.info("chosing mon {cmon} as ctrlrmon".format(cmon=ctrlr.hostname))
helper = RadosHelper(ctrlr, config, log)
'''beacause of limited machines resorting to following config'''
lrc_config = [(4, 2, 3), (2, 1, 3), (2, 2, 2)]
for conf in lrc_config:
(k, m, l) = conf
suffix = "{k}_{m}_{l}".format(k=k, m=m, l=l)
prof_name = "LRCprofile{suf}".format(suf=suffix)
if build.startswith('4'):
profile = "osd erasure-code-profile set {LRC} plugin=lrc k={k} m={m} l={l} \
crush-failure-domain=osd".format(LRC=prof_name, k=k, m=m, l=l)
else:
profile = "osd erasure-code-profile set {LRC} plugin=lrc k={k} m={m} l={l} \
ruleset-failure-domain=osd crush-failure-domain=osd".format(
LRC=prof_name, k=k, m=m, l=l)
try:
(out, err) = helper.raw_cluster_cmd(profile)
outbuf = out.read().decode()
log.info(outbuf)
log.info("created profile {LRC}".format(LRC=prof_name))
except Exception:
log.error("LRC profile creation failed")
log.error(traceback.format_exc())
return 1
'''create LRC ec pool'''
pname = "lrcpool{rand}{suf}".format(rand=random.randint(0, 10000),
suf=suffix)
try:
helper.create_pool(pname, 1, prof_name)
log.info("Pool {pname} created".format(pname=pname))
except Exception:
log.error("failed to create lrcpool")
log.error(traceback.format_exc())
return 1
'''check whether pool exists'''
try:
helper.get_pool_num(pname)
except Exception:
log.error("Unable to find pool")
log.error(traceback.format_exc())
return 1
return 0
def run(ceph_cluster, **kw):
"""
CEPH-9939:
Delete snapset objects in ec pool followed by
list-inconsistent-* commands
1. create a jerasure ec pool with k=4,m=2
2. create an object in the pool
3. chose any of the osd from the acting set and go to the backend
4. delete snap object from the backend
5. run deep-scrub on the pool
6. rados list-inconsistent-pg <pool>
7. rados list-inconsistent-obj <pg>
Args:
ceph_cluster (ceph.ceph.Ceph): ceph cluster
"""
log.info("Running CEPH-9939")
log.info(run.__doc__)
ceph_nodes = kw.get('ceph_nodes')
config = kw.get('config')
mons = []
osds = []
role = 'client'
for mnode in ceph_nodes:
if mnode.role == role:
mons.append(mnode)
role = 'osd'
for osd in ceph_nodes:
if osd.role == role:
osds.append(osd)
ctrlr = mons[0]
log.info("chosing mon {cmon} as ctrlrmon".format(cmon=ctrlr.hostname))
helper = RadosHelper(ctrlr, config, log)
"""create ec pool with k=4,m=2"""
k = 4
m = 2
pname = "ecsnapdelete_{rand}_{k}_{m}".format(rand=random.randint(0, 1000),
k=k,
m=m)
profile = pname
prof_cmd = "osd erasure-code-profile set {profile}\
k={k}\
m={m}\
rulset-failure-domain=osd\
crush-failure-domain=osd".format(profile=profile, k=k, m=m)
try:
(out, err) = helper.raw_cluster_cmd(prof_cmd)
outbuf = out.read().decode()
log.info(outbuf)
log.info("created profile {ec}".format(ec=profile))
except Exception:
log.error("ec profile creation failed")
log.error(traceback.format_exc())
return 1
'''create ec pool'''
try:
helper.create_pool(pname, 1, profile)
log.info("Pool {pname} is create".format(pname=pname))
except Exception:
log.error("failed to create pool")
log.error(traceback.format_exc())
return 1
'''check whether pool exists'''
try:
helper.get_pool_num(pname)
except Exception:
log.error("Unable to find pool")
log.error(traceback.format_exc())
return 1
time.sleep(10)
oname = "OBJ_{pname}".format(pname=pname)
cmd = "osd map {pname} {obj} --format json".format(pname=pname, obj=oname)
(out, err) = helper.raw_cluster_cmd(cmd)
outbuf = out.read().decode()
log.info(outbuf)
cmdout = json.loads(outbuf)
targt_pg = cmdout['pgid']
'''considering primary only as of now because of bug
1544680
'''
targt_osd_id = cmdout['up'][0]
'''write data and take snaps'''
putobj = "sudo rados -p {pool} put {obj} {path}".format(pool=pname,
obj=oname,
path="/etc/hosts")
for i in range(10):
(out, err) = ctrlr.exec_command(cmd=putobj)
snapcmd = "sudo rados mksnap -p {pool} {sname}".format(pool=pname,
sname="snap" +
str(i))
(out, err) = ctrlr.exec_command(cmd=snapcmd)
log.info("put {obj}, snap {snap}".format(obj=oname,
snap="snap" + str(i)))
"""Goto destination osd , stop osd use ceph-objectstore-tool
to delete snap """
target_osd = ceph_cluster.get_osd_by_id(targt_osd_id)
target_osd_node = target_osd.node
cot_environment = target_osd_node
osd_service = ceph_cluster.get_osd_service_name(targt_osd_id)
partition_path = ceph_cluster.get_osd_data_partition_path(targt_osd_id)
helper.kill_osd(target_osd_node, osd_service)
time.sleep(10)
osd_metadata = ceph_cluster.get_osd_metadata(targt_osd_id)
osd_data = osd_metadata.get('osd_data')
osd_journal = osd_metadata.get('osd_journal')
if ceph_cluster.containerized:
docker_image_string = '{docker_registry}/{docker_image}:{docker_tag}'.format(
docker_registry=ceph_cluster.ansible_config.get(
'ceph_docker_registry'),
docker_image=ceph_cluster.ansible_config.get('ceph_docker_image'),
docker_tag=ceph_cluster.ansible_config.get(
'ceph_docker_image_tag'))
cot_environment = helper.get_mgr_proxy_container(
target_osd_node, docker_image_string)
out, err = cot_environment.exec_command(
cmd='mount | grep "{partition_path} "'.format(
partition_path=partition_path),
check_ec=False)
device_mount_data = out.read().decode() # type: str
if not device_mount_data:
cot_environment.exec_command(
cmd='sudo mount {partition_path} {directory}'.format(
partition_path=partition_path, directory=osd_data))
slist_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
--op list \
{obj}|grep \\\"snapid\\\":1".format(osd_data=osd_data,
osd_journal=osd_journal,
obj=oname)
(out, err) = cot_environment.exec_command(cmd=slist_cmd)
outbuf = out.read().decode()
log.info(outbuf)
corrupt_cmd = "sudo ceph-objectstore-tool --data-path \
{osd_data} --journal-path \
{osd_journal} \
{outbuf} remove".format(osd_data=osd_data,
osd_journal=osd_journal,
outbuf="'" + (outbuf) + "'")
(out, err) = cot_environment.exec_command(cmd=corrupt_cmd)
outbuf = out.read().decode()
log.info(outbuf)
helper.revive_osd(target_osd_node, osd_service)
time.sleep(10)
run_scrub = "pg deep-scrub {pgid}".format(pgid=targt_pg)
(out, err) = helper.raw_cluster_cmd(run_scrub)
outbuf = out.read().decode()
log.info(outbuf)
while 'HEALTH_ERR' and 'active+clean+inconsistent' not in outbuf:
status = "-s --format json"
(out, err) = helper.raw_cluster_cmd(status)
outbuf = out.read().decode()
log.info("HEALTH_ERR found as expected")
log.info("inconsistent foud as expected")
timeout = 300
found = 0
while timeout:
incon_pg = "sudo rados list-inconsistent-pg \
{pname}".format(pname=pname)
(out, err) = ctrlr.exec_command(cmd=incon_pg)
outbuf = out.read().decode()
log.info(outbuf)
if targt_pg not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("pg not listed as inconsistent")
return 1
timeout = 300
found = 0
while timeout:
incon_obj = "sudo rados list-inconsistent-obj {pg}".format(pg=targt_pg)
(out, err) = ctrlr.exec_command(cmd=incon_obj)
outbuf = out.read().decode()
log.info(outbuf)
if oname not in outbuf:
time.sleep(1)
timeout = timeout - 1
else:
found = 1
break
if timeout == 0 and found == 0:
log.error("object is not listed in inconsistent obj")
return 1
return 0
