mirror of
https://github.com/ceph/ceph
synced 2025-02-24 19:47:44 +00:00
Merge PR #32158 into master
* refs/pull/32158/head: cephadm: fix exception when no time sync is running cephadm: properly extract osd fsid during adoption cephadm: do ceph-volume activate+deactivate as part of systemd unit ceph-daemon: behave with no /etc/ceph/ceph.conf Reviewed-by: Jan Fajerski <jfajerski@suse.com> Reviewed-by: Sebastian Wagner <swagner@suse.com>
This commit is contained in:
Sage Weil
commit
cd780f109c
@ -419,10 +419,10 @@ def get_legacy_config_fsid(cluster, legacy_dir=None):
|
||||
if legacy_dir is not None:
|
||||
config_file = os.path.abspath(legacy_dir + config_file)
|
||||
|
||||
config = read_config(config_file)
|
||||
|
||||
if config.has_section('global') and config.has_option('global', 'fsid'):
|
||||
return config.get('global', 'fsid')
|
||||
if os.path.exists(config_file):
|
||||
config = read_config(config_file)
|
||||
if config.has_section('global') and config.has_option('global', 'fsid'):
|
||||
return config.get('global', 'fsid')
|
||||
return None
|
||||
|
||||
def get_legacy_daemon_fsid(cluster, daemon_type, daemon_id, legacy_dir=None):
|
||||
@ -599,8 +599,9 @@ def extract_uid_gid():
|
||||
return (int(uid), int(gid))
|
||||
|
||||
def deploy_daemon(fsid, daemon_type, daemon_id, c, uid, gid,
|
||||
config, keyring):
|
||||
# type: (str, str, Union[int, str], CephContainer, int, int, Optional[str], Optional[str]) -> None
|
||||
config, keyring,
|
||||
osd_fsid=None):
|
||||
# type: (str, str, Union[int, str], CephContainer, int, int, Optional[str], Optional[str], Optional[str]) -> None
|
||||
if daemon_type == 'mon' and not os.path.exists(
|
||||
get_data_dir(fsid, 'mon', daemon_id)):
|
||||
assert config
|
||||
@ -644,32 +645,53 @@ def deploy_daemon(fsid, daemon_type, daemon_id, c, uid, gid,
|
||||
uid, gid,
|
||||
config, keyring)
|
||||
|
||||
if daemon_type == 'osd' and args.osd_fsid:
|
||||
pc = CephContainer(
|
||||
image=args.image,
|
||||
entrypoint='/usr/sbin/ceph-volume',
|
||||
args=[
|
||||
'lvm', 'activate',
|
||||
str(daemon_id), args.osd_fsid,
|
||||
'--no-systemd'
|
||||
],
|
||||
container_args=['--privileged'],
|
||||
volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
|
||||
cname='ceph-%s-activate-%s.%s' % (fsid, daemon_type, daemon_id),
|
||||
)
|
||||
pc.run()
|
||||
|
||||
deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c)
|
||||
deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
|
||||
osd_fsid=osd_fsid)
|
||||
update_firewalld(daemon_type)
|
||||
|
||||
def deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
|
||||
enable=True, start=True):
|
||||
# type: (str, int, int, str, Union[int, str], CephContainer, bool, bool) -> None
|
||||
enable=True, start=True,
|
||||
osd_fsid=None):
|
||||
# type: (str, int, int, str, Union[int, str], CephContainer, bool, bool, Optional[str]) -> None
|
||||
# cmd
|
||||
data_dir = get_data_dir(fsid, daemon_type, daemon_id)
|
||||
with open(data_dir + '/cmd', 'w') as f:
|
||||
f.write('#!/bin/sh\n' + ' '.join(c.run_cmd()) + '\n')
|
||||
os.fchmod(f.fileno(), 0o700)
|
||||
with open(data_dir + '/unit.run', 'w') as f:
|
||||
if daemon_type == 'osd':
|
||||
# osds have a pre-start step
|
||||
assert osd_fsid
|
||||
prestart = CephContainer(
|
||||
image=args.image,
|
||||
entrypoint='/usr/sbin/ceph-volume',
|
||||
args=[
|
||||
'lvm', 'activate',
|
||||
str(daemon_id), osd_fsid,
|
||||
'--no-systemd'
|
||||
],
|
||||
container_args=['--privileged'],
|
||||
volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
|
||||
cname='ceph-%s-%s.%s-activate' % (fsid, daemon_type, daemon_id),
|
||||
)
|
||||
f.write(' '.join(prestart.run_cmd()) + '\n')
|
||||
f.write(' '.join(c.run_cmd()) + '\n')
|
||||
os.fchmod(f.fileno(), 0o600)
|
||||
with open(data_dir + '/unit.poststop', 'w') as f:
|
||||
if daemon_type == 'osd':
|
||||
assert osd_fsid
|
||||
poststop = CephContainer(
|
||||
image=args.image,
|
||||
entrypoint='/usr/sbin/ceph-volume',
|
||||
args=[
|
||||
'lvm', 'deactivate',
|
||||
str(daemon_id), osd_fsid,
|
||||
'--no-systemd'
|
||||
],
|
||||
container_args=['--privileged'],
|
||||
volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
|
||||
cname='ceph-%s-%s.%s-deactivate' % (fsid, daemon_type,
|
||||
daemon_id),
|
||||
)
|
||||
f.write(' '.join(poststop.run_cmd()) + '\n')
|
||||
os.fchmod(f.fileno(), 0o600)
|
||||
|
||||
# systemd
|
||||
install_base_units(fsid)
|
||||
@ -881,8 +903,9 @@ LimitNPROC=1048576
|
||||
EnvironmentFile=-/etc/environment
|
||||
ExecStartPre=-{container_path} rm ceph-{fsid}-%i
|
||||
ExecStartPre=-{install_path} -d -m0770 -o {uid} -g {gid} /var/run/ceph/{fsid}
|
||||
ExecStart=/bin/bash {data_dir}/{fsid}/%i/cmd
|
||||
ExecStart=/bin/bash {data_dir}/{fsid}/%i/unit.run
|
||||
ExecStop=-{container_path} rm -f ceph-{fsid}-%i
|
||||
ExecStopPost=-/bin/bash {data_dir}/{fsid}/%i/unit.poststop
|
||||
Restart=on-failure
|
||||
RestartSec=10s
|
||||
TimeoutStartSec=120
|
||||
@ -1349,7 +1372,8 @@ def command_deploy():
|
||||
(uid, gid) = extract_uid_gid()
|
||||
c = get_container(args.fsid, daemon_type, daemon_id)
|
||||
deploy_daemon(args.fsid, daemon_type, daemon_id, c, uid, gid,
|
||||
config, keyring)
|
||||
config, keyring,
|
||||
osd_fsid=args.osd_fsid)
|
||||
if crash_keyring:
|
||||
deploy_crash(args.fsid, uid, gid, config, crash_keyring)
|
||||
|
||||
@ -1611,6 +1635,19 @@ def command_adopt():
|
||||
if not fsid:
|
||||
raise Error('could not detect legacy fsid; set fsid in ceph.conf')
|
||||
|
||||
data_dir_src = ('/var/lib/ceph/%s/%s-%s' %
|
||||
(daemon_type, args.cluster, daemon_id))
|
||||
data_dir_src = os.path.abspath(args.legacy_dir + data_dir_src)
|
||||
|
||||
osd_fsid = None
|
||||
if daemon_type == 'osd':
|
||||
path = os.path.join(data_dir_src, 'fsid')
|
||||
try:
|
||||
with open(path, 'r') as f:
|
||||
osd_fsid = f.read().strip()
|
||||
except IOError:
|
||||
raise Error('unable to read OSD fsid from %s' % path)
|
||||
|
||||
# NOTE: implicit assumption here that the units correspond to the
|
||||
# cluster we are adopting based on the /etc/{defaults,sysconfig}/ceph
|
||||
# CLUSTER field.
|
||||
@ -1626,9 +1663,6 @@ def command_adopt():
|
||||
|
||||
# data
|
||||
logger.info('Moving data...')
|
||||
data_dir_src = ('/var/lib/ceph/%s/%s-%s' %
|
||||
(daemon_type, args.cluster, daemon_id))
|
||||
data_dir_src = os.path.abspath(args.legacy_dir + data_dir_src)
|
||||
data_dir_dst = make_data_dir(fsid, daemon_type, daemon_id,
|
||||
uid=uid, gid=gid)
|
||||
move_files(glob(os.path.join(data_dir_src, '*')),
|
||||
@ -1659,7 +1693,8 @@ def command_adopt():
|
||||
c = get_container(fsid, daemon_type, daemon_id)
|
||||
deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
|
||||
enable=True, # unconditionally enable the new unit
|
||||
start=(state == 'running'))
|
||||
start=(state == 'running'),
|
||||
osd_fsid=osd_fsid)
|
||||
update_firewalld(daemon_type)
|
||||
|
||||
else:
|
||||
@ -1765,8 +1800,7 @@ def command_check_host():
|
||||
|
||||
# check for configured+running chronyd or ntp
|
||||
if not check_time_sync():
|
||||
raise RuntimeError('No time synchronization is active (checked all of %s)' %
|
||||
units)
|
||||
raise RuntimeError('No time synchronization is active')
|
||||
|
||||
logger.info('Host looks OK')
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user