def _stop_server(self, context, body=None):
"""Noout and stop all osd service, then stop the server.
body = {u'servers': [{u'cluster_id': 1, u'id': u'1'},
{u'cluster_id': 1, u'id': u'2'}]}
"""
LOG.info("DEBUG in stop server in scheduler manager.")
server_list = body['servers']
need_change_mds = False
for ser in server_list:
ser_ref = db.init_node_get(context, ser['id'])
ser['host'] = ser_ref['host']
if ser_ref['mds'] == 'yes':
need_change_mds = True
active_monitor = self._get_active_monitor(context)
LOG.info("stop_server of scheduer manager %s" % server_list)
for item in server_list:
res = db.init_node_get(context, item['id'])
self._start_stop(context, item['id'])
res = self._agent_rpcapi.stop_server(context,
item['id'],
res['host'])
self._agent_rpcapi.update_osd_state(context,
active_monitor['host'])
LOG.info('need_change_mds = %s' % need_change_mds)
if need_change_mds:
self.add_mds(context, server_list)
return True
def get_server(self, context, id):
LOG.info('get_server_list in conductor manager')
server = db.init_node_get(context, id)
ret = self._set_error(context)
if ret:
server['status'] = 'unavailable'
LOG.info("CEPH_LOG log server %s" % server)
return server
def get_server(self, context, id):
LOG.info('get_server_list in conductor manager')
server = db.init_node_get(context, id)
ret = self._set_error(context)
if ret:
server['status'] = 'unavailable'
LOG.info("CEPH_LOG log server %s" % server)
return server
def _add_servers(self, context, body=None):
"""Add the servers into ceph cluster.
It's notable that, the type of body['servers']
looks as below:
[
{u'is_storage': True,
u'is_monitor': True,
u'id': u'1',
u'zone_id': u'1'},
{u'is_storage': True,
u'is_monitor': False,
u'id': u'2',
u'zone_id': u'2'}
]
Here we also need to fetch info from DB.
"""
def _update_ssh_key():
server_list = db.init_node_get_all(context)
for ser in server_list:
if ser['status'] == 'Active' or ser['status'] == 'available':
self._agent_rpcapi.update_ssh_keys(context, ser['host'])
server_list = body['servers']
for ser in server_list:
ser_ref = db.init_node_get(context, ser['id'])
ser['host'] = ser_ref['host']
ser['cluster_id'] = self._agent_rpcapi.cluster_id(context,
ser['host'])
# It need to change the role defined in
# server.manifest
if ser['is_monitor'] == False:
if ser_ref['type'].find('monitor') != -1:
values = {'type': 'storage'}
db.init_node_update(context, ser_ref['id'], values)
self._update_server_list_status(context, server_list, 'running')
_update_ssh_key()
self.add_monitor(context, server_list)
# Begin to add osds.
LOG.info("start to add storage")
self.add_osd(context, server_list)
self._judge_drive_ext_threshold(context)
return True
def _remove_servers(self, context, body=None):
"""
[
{u'remove_storage': True,
u'remove_monitor': True,
u'id': u'1',
u'zone_id': u'1'},
{u'remove_storage': True,
u'remove_monitor': False,
u'id': u'2',
u'zone_id': u'2'}
]
"""
server_list = body['servers']
LOG.info('remove_servers = %s ' % server_list)
if len(server_list) <= 0:
return True
need_change_mds = False
for ser in server_list:
ser_ref = db.init_node_get(context, ser['id'])
ser['host'] = ser_ref['host']
ser['type'] = ser_ref['type']
ser['remove_monitor'] = ser['type'].find('monitor') != -1
ser['status'] = ser_ref['status']
ser['mds'] = ser_ref['mds']
if ser['mds'] == 'yes':
need_change_mds = True
LOG.info("start to remove monitor")
self.remove_monitors(context, server_list)
LOG.info("start to remove storage")
self.remove_osd(context, server_list)
if need_change_mds:
LOG.info("start to remove mds")
self.remove_mds(context, server_list)
self.add_mds(context, server_list)
return True
def _start_server(self, context, body=None):
"""Start all osd service, then start the server.
body = {u'servers': [{u'cluster_id': 1, u'id': u'1'},
{u'cluster_id': 1, u'id': u'2'}]}
"""
LOG.info("DEBUG in start server in scheduler manager.")
server_list = body['servers']
active_monitor = self._get_active_monitor(context)
for item in server_list:
#host = self._get_monitor_by_cluster_id(context, item['cluster_id'])
res = db.init_node_get(context, item['id'])
if not res:
LOG.error("No available node for node id %s" % \
item['id'])
try:
raise StorageServerStartFailed
except Exception, e:
LOG.error("%s: %s " %(e.code, e.message))
raise
self._start_start(context, item['id'])
res = self._agent_rpcapi.start_server(context, item['id'], res['host'])
self._agent_rpcapi.update_osd_state(context, active_monitor['host'])
def create_cluster(self, context, server_list):
"""Add the servers into ceph cluster.
It's notable that, the type of body['servers']
looks as below:
[
{u'is_storage': True,
u'is_monitor': True,
u'id': u'1',
u'zone_id': u'1'},
{u'is_storage': True,
u'is_monitor': False,
u'id': u'2',
u'zone_id': u'2'}
]
Here we also need to fetch info from DB.
"""
# Add hostname here.
for ser in server_list:
ser_ref = db.init_node_get(context, ser['id'])
ser['host'] = ser_ref['host']
# Use mkcephfs to set up ceph system.
monitor_node = self._select_monitor(context, server_list)
LOG.info('Choose monitor node = %s' % monitor_node)
def _update(status):
LOG.debug('status = %s' % status)
self._update_server_list_status(context,
server_list,
status)
if status.lower().find('error') != -1:
raise
# Set at least 3 mons when creating cluster
nums = len(server_list)
if nums >= 3:
count = 0
rest_mon_num = 0
for ser in server_list:
if ser['is_monitor'] == True:
count += 1
if count < 3:
rest_mon_num = 3 - count
if rest_mon_num > 0:
for ser in server_list:
if ser['is_monitor'] == False:
ser['is_monitor'] = True
rest_mon_num -= 1
if rest_mon_num <= 0:
break
# Clean ceph data.
def __clean_data(host):
self._agent_rpcapi.update_ssh_keys(context, host)
self._agent_rpcapi.clean_ceph_data(context, host)
def __create_crushmap(context, server_list, host):
self._agent_rpcapi.create_crushmap(context,
server_list=server_list,
host=host)
try:
_update("Cleaning")
thd_list = []
for ser in server_list:
thd = utils.MultiThread(__clean_data, host=ser['host'])
thd_list.append(thd)
utils.start_threads(thd_list)
_update("Clean success")
except:
_update("ERROR: Cleaning")
# When clean data, we also begin to create ceph.conf
# and init osd in db.
# Do not run with the same time as clean_data.
# It maybe cleaned by clean_data.
try:
_update("Create ceph.conf")
self._agent_rpcapi.inital_ceph_osd_db_conf(context,
server_list=server_list,
host=monitor_node['host'])
_update("Create ceph.conf success")
except:
_update("ERROR: ceph.conf")
try:
_update("create crushmap")
# Then begin to create crush map file.
create_crushmap = utils.MultiThread(__create_crushmap,
context=context,
server_list=server_list,
host=monitor_node['host'])
create_crushmap.start()
except:
_update("ERROR: crushmap")
try:
# Begin to mount disks on the mount_point.
_update("Mount disks")
def __mount_disk(host):
self._agent_rpcapi.mount_disks(context, host)
thd_list = []
for ser in server_list:
thd = utils.MultiThread(__mount_disk, host=ser['host'])
thd_list.append(thd)
utils.start_threads(thd_list)
_update("Mount disks success")
except:
_update("ERROR: mount disk")
# Generate monitor keyring file.
_update("start montior")
monitor_keyring = utils.gen_mon_keyring()
def __write_monitor_keyring(host):
self._agent_rpcapi.write_monitor_keyring(context,
monitor_keyring,
host)
thd_list = []
for ser in server_list:
thd = utils.MultiThread(__write_monitor_keyring, host=ser['host'])
thd_list.append(thd)
utils.start_threads(thd_list)
_update("start monitor success")
self._track_monitors(context, server_list)
self._agent_rpcapi.prepare_osds(context,
server_list,
host=monitor_node['host'])
self._agent_rpcapi.upload_keyring_admin_into_db(context,
host=monitor_node['host'])
# Begin to start osd service.
_update('Start osds')
def __start_osd(host):
self._agent_rpcapi.start_osd(context, host)
thd_list = []
for ser in server_list:
thd = utils.MultiThread(__start_osd, host=ser['host'])
thd_list.append(thd)
utils.start_threads(thd_list)
_update('OSD success')
# add mds service
try:
_update("Start mds")
LOG.info('start mds services, host = %s' % monitor_node['host'])
self._agent_rpcapi.add_mds(context, host=monitor_node['host'])
except:
_update("ERROR: mds")
# Created begin to get ceph status
try:
_update('Ceph status')
stat = self._agent_rpcapi.get_ceph_health(context,
monitor_node['host'])
except:
_update('ERROR: ceph -s')
if stat == False:
self._update_server_list_status(context,
server_list,
"Ceph Start Error")
LOG.error('Ceph starting failed!')
raise
try:
_update('Set crushmap')
# Wait until it's created over.
while create_crushmap.is_alive():
time.sleep(1)
def __set_crushmap(context, host):
self._agent_rpcapi.set_crushmap(context,
host)
_update('Active')
set_crushmap = utils.MultiThread(__set_crushmap,
context=context,
host=monitor_node['host'])
set_crushmap.start()
except:
_update('ERROR: set crushmap')
self._update_init_node(context, server_list)
while set_crushmap.is_alive():
time.sleep(1)
self._agent_rpcapi.update_all_status(context,
host=monitor_node['host'])
self._judge_drive_ext_threshold(context)
self._update_drive_ext_threshold(context)
return {'message':'res'}