def _configure_services(client, cluster):
nn_host = u.get_namenode(cluster).fqdn()
snn = u.get_secondarynamenodes(cluster)
snn_host = snn[0].fqdn() if snn else None
rm_host = u.get_resourcemanager(cluster).fqdn() if u.get_resourcemanager(
cluster) else None
hs_host = u.get_historyserver(cluster).fqdn() if u.get_historyserver(
cluster) else None
dn_hosts = [dn.fqdn() for dn in u.get_datanodes(cluster)]
nm_hosts = [tt.fqdn() for tt in u.get_nodemanagers(cluster)]
oozie_host = u.get_oozie(cluster).fqdn() if u.get_oozie(cluster) else None
hive_host = u.get_hiveserver(cluster).fqdn() if u.get_hiveserver(
cluster) else None
services = []
if u.get_namenode(cluster):
services += ['hdfs']
if u.get_resourcemanager(cluster):
services += ['yarn']
if oozie_host:
services += ['oozie']
services += ['pig']
if hive_host:
services += ['hive']
LOG.debug("Add services: %s" % ', '.join(services))
client.services.add(services)
LOG.debug("Assign roles to hosts")
client.services.hdfs.add_nodes('PrimaryNameNode', [nn_host])
client.services.hdfs.add_nodes('DataNode', dn_hosts)
if snn:
client.services.hdfs.add_nodes('SecondaryNameNode', [snn_host])
if oozie_host:
client.services.oozie.add_nodes('Oozie', [oozie_host])
if hive_host:
client.services.hive.add_nodes('HiveServer', [hive_host])
if rm_host:
client.services.yarn.add_nodes('ResourceManager', [rm_host])
client.services.yarn.add_nodes('NodeManager', nm_hosts)
if hs_host:
client.services.yarn.add_nodes('HistoryServer', [hs_host])
def _set_cluster_info(self, cluster):
nn = utils.get_namenode(cluster)
rm = utils.get_resourcemanager(cluster)
hs = utils.get_historyserver(cluster)
oo = utils.get_oozie(cluster)
info = {}
if rm:
info['YARN'] = {
'Web UI': 'http://%s:%s' % (rm.management_ip, '8088'),
'ResourceManager': 'http://%s:%s' % (rm.management_ip, '8032')
}
if nn:
info['HDFS'] = {
'Web UI': 'http://%s:%s' % (nn.management_ip, '50070'),
'NameNode': 'hdfs://%s:%s' % (nn.hostname(), '9000')
}
if oo:
info['JobFlow'] = {
'Oozie': 'http://%s:%s' % (oo.management_ip, '11000')
}
if hs:
info['MapReduce JobHistory Server'] = {
'Web UI': 'http://%s:%s' % (hs.management_ip, '19888')
}
ctx = context.ctx()
conductor.cluster_update(ctx, cluster, {'info': info})
def start_cluster(self, cluster):
nn = utils.get_namenode(cluster)
run.format_namenode(nn)
run.start_hadoop_process(nn, 'namenode')
rm = utils.get_resourcemanager(cluster)
run.start_yarn_process(rm, 'resourcemanager')
for dn in utils.get_datanodes(cluster):
run.start_hadoop_process(dn, 'datanode')
run.await_datanodes(cluster)
for nm in utils.get_nodemanagers(cluster):
run.start_yarn_process(nm, 'nodemanager')
hs = utils.get_historyserver(cluster)
if hs:
run.start_historyserver(hs)
oo = utils.get_oozie(cluster)
if oo:
run.start_oozie_process(oo)
self._set_cluster_info(cluster)
def _set_cluster_info(self, cluster):
mng = u.get_instances(cluster, 'manager')[0]
nn = u.get_namenode(cluster)
jt = u.get_resourcemanager(cluster)
oozie = u.get_oozie(cluster)
#TODO(alazarev) make port configurable (bug #1262895)
info = {
'IDH Manager': {
'Web UI': 'https://%s:9443' % mng.management_ip
}
}
if jt:
#TODO(alazarev) make port configurable (bug #1262895)
info['Yarn'] = {
'ResourceManager Web UI': 'http://%s:8088' % jt.management_ip
}
#TODO(alazarev) make port configurable (bug #1262895)
info['Yarn']['ResourceManager'] = '%s:8032' % jt.hostname()
if nn:
#TODO(alazarev) make port configurable (bug #1262895)
info['HDFS'] = {'Web UI': 'http://%s:50070' % nn.management_ip}
#TODO(alazarev) make port configurable (bug #1262895)
info['HDFS']['NameNode'] = 'hdfs://%s:8020' % nn.hostname()
if oozie:
#TODO(alazarev) make port configurable (bug #1262895)
info['JobFlow'] = {
'Oozie': 'http://%s:11000' % oozie.management_ip
}
ctx = context.ctx()
conductor.cluster_update(ctx, cluster, {'info': info})
def _set_cluster_info(self, cluster):
mng = u.get_instances(cluster, 'manager')[0]
nn = u.get_namenode(cluster)
jt = u.get_resourcemanager(cluster)
oozie = u.get_oozie(cluster)
#TODO(alazarev) make port configurable (bug #1262895)
info = {'IDH Manager': {
'Web UI': 'https://%s:9443' % mng.management_ip
}}
if jt:
#TODO(alazarev) make port configurable (bug #1262895)
info['Yarn'] = {
'ResourceManager Web UI': 'http://%s:8088' % jt.management_ip
}
#TODO(alazarev) make port configurable (bug #1262895)
info['Yarn']['ResourceManager'] = '%s:8032' % jt.hostname()
if nn:
#TODO(alazarev) make port configurable (bug #1262895)
info['HDFS'] = {
'Web UI': 'http://%s:50070' % nn.management_ip
}
#TODO(alazarev) make port configurable (bug #1262895)
info['HDFS']['NameNode'] = 'hdfs://%s:8020' % nn.hostname()
if oozie:
#TODO(alazarev) make port configurable (bug #1262895)
info['JobFlow'] = {
'Oozie': 'http://%s:11000' % oozie.management_ip
}
ctx = context.ctx()
conductor.cluster_update(ctx, cluster, {'info': info})
def _get_hadoop_configs(node_group):
cluster = node_group.cluster
nn_hostname = utils.get_namenode(cluster).hostname()
res_hostname = utils.get_resourcemanager(cluster).hostname()
dirs = _get_hadoop_dirs(node_group)
confs = {
'Hadoop': {
'fs.defaultFS': 'hdfs://%s:9000' % nn_hostname
},
'HDFS': {
'dfs.namenode.name.dir': ','.join(dirs['hadoop_name_dirs']),
'dfs.namenode.data.dir': ','.join(dirs['hadoop_data_dirs']),
'dfs.hosts': '%s/dn-include' % HADOOP_CONF_DIR,
'dfs.hosts.exclude': '%s/dn-exclude' % HADOOP_CONF_DIR
},
'YARN': {
'yarn.nodemanager.aux-services':
'mapreduce_shuffle',
'yarn.resourcemanager.hostname':
'%s' % res_hostname,
'yarn.resourcemanager.nodes.include-path':
'%s/nm-include' % (HADOOP_CONF_DIR),
'yarn.resourcemanager.nodes.exclude-path':
'%s/nm-exclude' % (HADOOP_CONF_DIR)
},
'MapReduce': {
'mapreduce.framework.name': 'yarn'
},
}
oozie = utils.get_oozie(cluster)
if oozie:
hadoop_cfg = {
'hadoop.proxyuser.hadoop.hosts': '*',
'hadoop.proxyuser.hadoop.groups': 'hadoop'
}
confs['Hadoop'].update(hadoop_cfg)
oozie_cfg = o_helper.get_oozie_required_xml_configs(HADOOP_CONF_DIR)
if c_helper.is_mysql_enabled(cluster):
oozie_cfg.update(o_helper.get_oozie_mysql_configs())
confs['JobFlow'] = oozie_cfg
if c_helper.get_config_value(c_helper.ENABLE_SWIFT.applicable_target,
c_helper.ENABLE_SWIFT.name, cluster):
swift_configs = {}
for config in swift.get_swift_configs():
swift_configs[config['name']] = config['value']
confs['Hadoop'].update(swift_configs)
if c_helper.is_data_locality_enabled(cluster):
confs['Hadoop'].update(th.TOPOLOGY_CONFIG)
confs['Hadoop'].update(
{"topology.script.file.name": HADOOP_CONF_DIR + "/topology.sh"})
return confs, c_helper.get_env_configs()
def _get_hadoop_configs(node_group):
cluster = node_group.cluster
nn_hostname = utils.get_namenode(cluster).hostname()
res_hostname = utils.get_resourcemanager(cluster).hostname()
dirs = _get_hadoop_dirs(node_group)
confs = {
'Hadoop': {
'fs.defaultFS': 'hdfs://%s:9000' % nn_hostname
},
'HDFS': {
'dfs.namenode.name.dir': ','.join(dirs['hadoop_name_dirs']),
'dfs.namenode.data.dir': ','.join(dirs['hadoop_data_dirs']),
'dfs.hosts': '%s/dn-include' % HADOOP_CONF_DIR,
'dfs.hosts.exclude': '%s/dn-exclude' % HADOOP_CONF_DIR
},
'YARN': {
'yarn.nodemanager.aux-services': 'mapreduce_shuffle',
'yarn.resourcemanager.hostname': '%s' % res_hostname,
'yarn.resourcemanager.nodes.include-path': '%s/nm-include' % (
HADOOP_CONF_DIR),
'yarn.resourcemanager.nodes.exclude-path': '%s/nm-exclude' % (
HADOOP_CONF_DIR)
},
'MapReduce': {
'mapreduce.framework.name': 'yarn'
},
}
oozie = utils.get_oozie(cluster)
if oozie:
hadoop_cfg = {
'hadoop.proxyuser.hadoop.hosts': '*',
'hadoop.proxyuser.hadoop.groups': 'hadoop'
}
confs['Hadoop'].update(hadoop_cfg)
oozie_cfg = o_helper.get_oozie_required_xml_configs(HADOOP_CONF_DIR)
if c_helper.is_mysql_enabled(cluster):
oozie_cfg.update(o_helper.get_oozie_mysql_configs())
confs['JobFlow'] = oozie_cfg
if c_helper.get_config_value(c_helper.ENABLE_SWIFT.applicable_target,
c_helper.ENABLE_SWIFT.name, cluster):
swift_configs = {}
for config in swift.get_swift_configs():
swift_configs[config['name']] = config['value']
confs['Hadoop'].update(swift_configs)
if c_helper.is_data_locality_enabled(cluster):
confs['Hadoop'].update(th.TOPOLOGY_CONFIG)
confs['Hadoop'].update({"topology.script.file.name":
HADOOP_CONF_DIR + "/topology.sh"})
return confs, c_helper.get_env_configs()
def get_nodemanagers_status(cluster):
statuses = {}
resourcemanager = u.get_resourcemanager(cluster)
status_regexp = r'^(\S+):\d+\s+(\w+)'
matcher = re.compile(status_regexp, re.MULTILINE)
yarn_report = resourcemanager.remote().execute_command(
'sudo su - -c "yarn node -all -list" hadoop')[1]
for host, status in matcher.findall(yarn_report):
statuses[host] = status.lower()
return statuses
def validate_additional_ng_scaling(cluster, additional):
rm = u.get_resourcemanager(cluster)
scalable_processes = _get_scalable_processes()
for ng_id in additional:
ng = gu.get_by_id(cluster.node_groups, ng_id)
if not set(ng.node_processes).issubset(scalable_processes):
msg = "Vanilla plugin cannot scale nodegroup with processes: %s"
raise ex.NodeGroupCannotBeScaled(ng.name,
msg % ' '.join(ng.node_processes))
if not rm and 'nodemanager' in ng.node_processes:
msg = ("Vanilla plugin cannot scale node group with processes "
"which have no master-processes run in cluster")
raise ex.NodeGroupCannotBeScaled(ng.name, msg)
def validate_additional_ng_scaling(cluster, additional):
rm = u.get_resourcemanager(cluster)
scalable_processes = _get_scalable_processes()
for ng_id in additional:
ng = gu.get_by_id(cluster.node_groups, ng_id)
if not set(ng.node_processes).issubset(scalable_processes):
msg = "Vanilla plugin cannot scale nodegroup with processes: %s"
raise ex.NodeGroupCannotBeScaled(ng.name,
msg % ' '.join(ng.node_processes))
if not rm and 'nodemanager' in ng.node_processes:
msg = ("Vanilla plugin cannot scale node group with processes "
"which have no master-processes run in cluster")
raise ex.NodeGroupCannotBeScaled(ng.name, msg)
def _validate_additional_ng_scaling(self, cluster, additional):
rm = u.get_resourcemanager(cluster)
scalable_processes = self._get_scalable_processes()
for ng_id in additional:
ng = self._get_by_id(cluster.node_groups, ng_id)
if not set(ng.node_processes).issubset(scalable_processes):
raise ex.NodeGroupCannotBeScaled(
ng.name, "Intel plugin cannot scale nodegroup"
" with processes: " + ' '.join(ng.node_processes))
if not rm and 'nodemanager' in ng.node_processes:
raise ex.NodeGroupCannotBeScaled(
ng.name, "Intel plugin cannot scale node group with "
"processes which have no master-processes run "
"in cluster")
def _validate_additional_ng_scaling(self, cluster, additional):
rm = u.get_resourcemanager(cluster)
scalable_processes = self._get_scalable_processes()
for ng_id in additional:
ng = self._get_by_id(cluster.node_groups, ng_id)
if not set(ng.node_processes).issubset(scalable_processes):
raise ex.NodeGroupCannotBeScaled(
ng.name, "Intel plugin cannot scale nodegroup"
" with processes: " +
' '.join(ng.node_processes))
if not rm and 'nodemanager' in ng.node_processes:
raise ex.NodeGroupCannotBeScaled(
ng.name, "Intel plugin cannot scale node group with "
"processes which have no master-processes run "
"in cluster")
def start_cluster(cluster):
client = c.IntelClient(u.get_instance(cluster, 'manager'), cluster.name)
LOG.debug("Starting hadoop services")
client.services.hdfs.start()
if u.get_resourcemanager(cluster):
client.services.yarn.start()
if u.get_hiveserver(cluster):
client.services.hive.start()
if u.get_oozie(cluster):
LOG.info("Setup oozie")
_setup_oozie(cluster)
client.services.oozie.start()
def refresh_yarn_nodes(cluster):
rm = u.get_resourcemanager(cluster)
rm.remote().execute_command(
'sudo su - -c "yarn rmadmin -refreshNodes" hadoop')
def refresh_yarn_nodes(cluster):
rm = u.get_resourcemanager(cluster)
rm.remote().execute_command(
'sudo su - -c "yarn rmadmin -refreshNodes" hadoop')