def check_creds(user, passwd):
# Sanity check KDC credentials
if not '@' in user:
raise ProvisionatorException(
"Supplied KDC account username contains '@")
elif len(passwd) == 0:
raise ProvisionatorException("Supplied KDC account password is empty")
def enable_rmha(conf):
ha_hosts_map = {}
standby_rm_key = "StandbyResourceManager"
try:
# Open a connection to CM and get a CM object
api = util.get_api_handle(conf)
cm = api.get_cloudera_manager()
cl = None
if 'cluster' in conf and 'name' in conf['cluster']:
cl = api.get_cluster(conf['cluster']['name'])
else:
raise ProvisionatorException("No cluster specified")
host_id_map = util.host_id_map(config, api)
#todo: check if enabled already
yarn_svc = config.get_cluster_service_by_type(conf, 'YARN')
yarn = cl.get_service(yarn_svc['name'])
standby_rm = yarn_svc['haconfig']['standby_rm']
ha_hosts_map[standby_rm_key] = standby_rm
LOG.info("host_id_map: " + str(host_id_map))
LOG.info("ha_host_map: " + str(ha_hosts_map))
for ha_host in ha_hosts_map:
LOG.info ("ha_hosts_map[ha_host]: " + ha_hosts_map[ha_host])
try:
LOG.info("Host ID for " + ha_host + ": "+ host_id_map[ha_hosts_map[ha_host]])
except KeyError:
LOG.error("Cannot find host " + ha_hosts_map[ha_host] + " for role \""
+ ha_host + "\" in list of hosts (" + str(list(host_id_map)) + ")")
return
zk_svc = config.get_cluster_service_by_type(conf, 'ZOOKEEPER')
zk_svc_name = zk_svc['name']
LOG.info("Running YARN HA configuration now...")
# new_rm_host_id - ID of host where Standby Resource Manager will be created.
# zk_service_name - Name of the ZooKeeper service to use for auto-failover.
cmd = yarn.enable_rm_ha(
host_id_map[standby_rm],
zk_service_name=zk_svc_name
)
util.wait_for_command(cmd, True)
#TODO: in some cases time.sleep(x) is needed to avoid race conditions
try:
mgmt_svc = cm.get_service()
LOG.info("Restarting management services")
mgmt.restart(cm)
except ApiException:
pass
LOG.info("Restarting cluster")
cluster.restart(cl)
except ApiException, e:
raise ProvisionatorException(e)
def enable_kerberos(conf):
try:
# Open a connection to CM and get a CM object
api = util.get_api_handle(conf)
cm = api.get_cloudera_manager()
cl = None
if 'cluster' in conf and 'name' in conf['cluster']:
cl = api.get_cluster(conf['cluster']['name'])
else:
raise ProvisionatorException("No cluster specified")
# Check the HDFS service to see if Kerberos is already enabled
hdfs_svc = config.get_cluster_service_by_type(conf, 'HDFS')
hdfs = cl.get_service(hdfs_svc['name'])
hdfs_cfg, hdfs_roletype_cfg = hdfs.get_config(view='full')
if not hdfs_cfg['hadoop_security_authentication'].value == 'kerberos':
# Kerberos has not been enabled - add the KDC creds
if 'kdc_user' in conf['cm'] and 'kdc_pass' in conf['cm']:
check_creds(conf['cm']['kdc_user'], conf['cm']['kdc_pass'])
LOG.info("Importing credentials for administrative account")
cmd = cm.import_admin_credentials(conf['cm']['kdc_user'],
conf['cm']['kdc_pass'])
util.wait_for_command(cmd, True)
# OK let's do this
datanode_transceiver_port = DATANODE_TRANSCEIVER_PORT
datanode_web_port = DATANODE_WEB_PORT
cm_cfg = cm.get_config()
if 'SINGLE_USER_ENABLED' in cm_cfg:
# TODO: don't hardcode this
datanode_transceiver_port = 4004
datanode_web_port = 4006
LOG.info("Running Kerberos configuration now...")
cmd = cl.configure_for_kerberos(
datanode_transceiver_port=datanode_transceiver_port,
datanode_web_port=datanode_web_port)
util.wait_for_command(cmd, True)
#TODO: in some cases time.sleep(x) is needed to avoid race conditions
try:
mgmt_svc = cm.get_service()
LOG.info("Restarting management services")
mgmt.restart(cm)
except ApiException:
pass
LOG.info("Restarting cluster")
cluster.restart(cl)
else:
LOG.info("Kerberos already enabled")
except ApiException, e:
raise ProvisionatorException(e)
def get_parcel_by_short_version(cluster, version):
parcels = [
p for p in cluster.get_all_parcels() if p.version.startswith(version)
]
matching_parcels = sorted(parcels, key=lambda p: p.version, reverse=True)
if len(matching_parcels) < 1:
LOG.error("No matching parcels found for %s" % version)
raise ProvisionatorException("No matching parcels found for %s" %
version)
return matching_parcels[0]
def wait_for_command(cmd, exception_on_fail=False, timeout=300):
status = cmd.wait(timeout)
if status.active == 'true':
if exception_on_fail:
raise ProvisionatorException(
"Command %s (%s) timed out after %d seconds" %
(status.name, status.id, timeout))
else:
LOG.warn("Command %s (%s) timed out after %d seconds" %
(status.name, status.id, timeout))
if status.active == 'false' and status.success == 'false':
if exception_on_fail:
raise ProvisionatorException(
"Command %s (%s) failed with: %s" %
(status.name, status.id, status.resultMessage))
else:
LOG.warn("Command %s (%s) failed with: %s" %
(status.name, status.id, status.resultMessage))
return status.active == 'false' and status.success == 'true'
def distribute_parcel(cluster, parcel):
"""
Distribute given parcel version
cluster -- ApiCluster object
parcel -- ApiParcel object
"""
parcel.start_distribution()
while True:
parcel = cluster.get_parcel('CDH', parcel.version)
if parcel.stage == 'DISTRIBUTED':
break
if parcel.state.errors:
raise ProvisionatorException(str(parcel.state.errors))
LOG.info("Parcel distribution progress: %s/%s", parcel.state.progress,
parcel.state.totalProgress)
time.sleep(5)
LOG.info("Distributed CDH parcel version '%s' on cluster '%s'",
parcel.version, cluster.name)
def download_parcel(cluster, parcel):
"""
Downloads given parcel version
cluster -- ApiCluster object
parcel -- ApiParcel object
"""
parcel.start_download()
while True:
parcel = cluster.get_parcel('CDH', parcel.version)
if parcel.stage == 'DOWNLOADED':
break
if parcel.state.errors:
raise ProvisionatorException(str(parcel.state.errors))
LOG.info("Parcel download progress: %s/%s", parcel.state.progress,
parcel.state.totalProgress)
time.sleep(5)
LOG.info("Downloaded CDH parcel version '%s' on cluster '%s'", parcel,
cluster.name)
def enable_hdfsha(conf):
ha_hosts_map = {}
active_namenode_key = "ActiveNamenode"
standby_namenode_key = "StandbyNamenode"
journalnode1_key = "JournalNode1"
journalnode2_key = "JournalNode2"
journalnode3_key = "JournalNode3"
try:
# Open a connection to CM and get a CM object
api = util.get_api_handle(conf)
cm = api.get_cloudera_manager()
cl = None
if 'cluster' in conf and 'name' in conf['cluster']:
cl = api.get_cluster(conf['cluster']['name'])
else:
raise ProvisionatorException("No cluster specified")
#todo: check if enabled already
hdfs_svc = config.get_cluster_service_by_type(conf, 'HDFS')
hdfs = cl.get_service(hdfs_svc['name'])
hdfs_cfg, hdfs_roletype_cfg = hdfs.get_config(view='full')
host_id_map = util.host_id_map(config, api)
active_namenode = hdfs_svc['haconfig']['active_namenode']
standby_namenode = hdfs_svc['haconfig']['standby_namenode']
journalnode1 = hdfs_svc['haconfig']['journalnode1']
journalnode2 = hdfs_svc['haconfig']['journalnode2']
journalnode3 = hdfs_svc['haconfig']['journalnode3']
ns = hdfs_svc['haconfig']['ns']
jnEditsDir = hdfs_svc['haconfig']['jnEditsDir']
ha_hosts_map[active_namenode_key] = active_namenode
ha_hosts_map[standby_namenode_key] = standby_namenode
ha_hosts_map[journalnode1_key] = journalnode1
ha_hosts_map[journalnode2_key] = journalnode2
ha_hosts_map[journalnode3_key] = journalnode3
LOG.info("host_id_map: " + str(host_id_map))
LOG.info("ha_host_map: " + str(ha_hosts_map))
for ha_host in ha_hosts_map:
LOG.info ("ha_hosts_map[ha_host]: " + ha_hosts_map[ha_host])
try:
LOG.info("Host ID for " + ha_host + ": "+ host_id_map[ha_hosts_map[ha_host]])
except KeyError:
LOG.error("Cannot find host " + ha_hosts_map[ha_host] + " for role \""
+ ha_host + "\" in list of hosts (" + str(list(host_id_map)) + ")")
return
jns = []
jns.append({'jnHostId': host_id_map[journalnode1], 'jnName': journalnode1_key, 'jnEditsDir': jnEditsDir})
jns.append({'jnHostId': host_id_map[journalnode2], 'jnName': journalnode2_key, 'jnEditsDir': jnEditsDir})
jns.append({'jnHostId': host_id_map[journalnode3], 'jnName': journalnode3_key, 'jnEditsDir': jnEditsDir})
zk_svc = config.get_cluster_service_by_type(conf, 'ZOOKEEPER')
zk_svc_name = zk_svc['name']
nn = hdfs.get_roles_by_type("NAMENODE")
if not nn:
raise ProvisionatorException("Could not find the active Namenode for current configuration")
if len([instance.name for instance in nn]) > 1:
LOG.warn("Found more than one Namnode in current configuration")
nn_name = ([instance.name for instance in nn])[0]
LOG.info("Active Namenode name in current config: " + nn_name)
LOG.info("Running HDFS HA configuration now...")
# standby_host_id - ID of host where Standby NameNode will be created.
# nameservice - Nameservice to be used while enabling HA
# jns - List of Journal Nodes. Each element must be a dict containing the following keys:
# jnHostId: ID of the host where the new JournalNode will be created.
# jnName: Name of the JournalNode role (optional)
# jnEditsDir: Edits dir of the JournalNode. Can be omitted if the config is already set at RCG level.
# zk_service_name - Name of the ZooKeeper service to use for auto-failover.
cmd = hdfs.enable_nn_ha(
nn_name,
host_id_map[standby_namenode],
ns,
jns,
standby_name_dir_list=None,
qj_name=None,
standby_name=None,
active_fc_name=None,
standby_fc_name=None,
zk_service_name=zk_svc_name,
force_init_znode=True,
clear_existing_standby_name_dirs=True,
clear_existing_jn_edits_dir=True
)
util.wait_for_command(cmd, True)
#TODO: in some cases time.sleep(x) is needed to avoid race conditions
try:
mgmt_svc = cm.get_service()
LOG.info("Restarting management services")
mgmt.restart(cm)
except ApiException:
pass
LOG.info("Restarting cluster")
cluster.restart(cl)
except ApiException, e:
raise ProvisionatorException(e)