def test_get_jobtracker(self):
cl = tu.create_cluster('cl1', 't1', 'vanilla', '1.2.1',
[self.ng_manager, self.ng_jobtracker])
self.assertEqual('jt1', u.get_jobtracker(cl).instance_id)
cl = tu.create_cluster('cl1', 't1', 'vanilla', '1.2.1',
[self.ng_manager])
self.assertIsNone(u.get_jobtracker(cl))
def test_get_jobtracker(self):
cl = tu.create_cluster('cl1', 't1', 'vanilla', '1.2.1',
[self.ng_manager, self.ng_jobtracker])
self.assertEqual('jt1', u.get_jobtracker(cl).instance_id)
cl = tu.create_cluster('cl1', 't1', 'vanilla', '1.2.1',
[self.ng_manager])
self.assertIsNone(u.get_jobtracker(cl))
def _set_cluster_info(self, cluster):
nn = vu.get_namenode(cluster)
jt = vu.get_jobtracker(cluster)
oozie = vu.get_oozie(cluster)
info = {}
if jt:
ui_port = c_helper.get_port_from_config("MapReduce", "mapred.job.tracker.http.address", cluster)
jt_port = c_helper.get_port_from_config("MapReduce", "mapred.job.tracker", cluster)
info["MapReduce"] = {
"Web UI": "http://%s:%s" % (jt.management_ip, ui_port),
"JobTracker": "%s:%s" % (jt.hostname(), jt_port),
}
if nn:
ui_port = c_helper.get_port_from_config("HDFS", "dfs.http.address", cluster)
nn_port = c_helper.get_port_from_config("HDFS", "fs.default.name", cluster)
info["HDFS"] = {
"Web UI": "http://%s:%s" % (nn.management_ip, ui_port),
"NameNode": "hdfs://%s:%s" % (nn.hostname(), nn_port),
}
if oozie:
# TODO(yrunts) change from hardcode value
info["JobFlow"] = {"Oozie": "http://%s:11000" % oozie.management_ip}
ctx = context.ctx()
conductor.cluster_update(ctx, cluster, {"info": info})
def generate_sahara_configs(cluster, node_group=None):
nn_hostname = vu.get_instance_hostname(vu.get_namenode(cluster))
jt_hostname = vu.get_instance_hostname(vu.get_jobtracker(cluster))
oozie_hostname = vu.get_instance_hostname(vu.get_oozie(cluster))
hive_hostname = vu.get_instance_hostname(vu.get_hiveserver(cluster))
storage_path = node_group.storage_paths() if node_group else None
# inserting common configs depends on provisioned VMs and HDFS placement
# TODO(aignatov): should be moved to cluster context
cfg = {
'fs.default.name':
'hdfs://%s:8020' % nn_hostname,
'dfs.name.dir':
extract_hadoop_path(storage_path, '/lib/hadoop/hdfs/namenode'),
'dfs.data.dir':
extract_hadoop_path(storage_path, '/lib/hadoop/hdfs/datanode'),
'dfs.hosts':
'/etc/hadoop/dn.incl',
'dfs.hosts.exclude':
'/etc/hadoop/dn.excl',
}
if jt_hostname:
mr_cfg = {
'mapred.job.tracker':
'%s:8021' % jt_hostname,
'mapred.system.dir':
extract_hadoop_path(storage_path, '/mapred/mapredsystem'),
'mapred.local.dir':
extract_hadoop_path(storage_path, '/lib/hadoop/mapred'),
'mapred.hosts':
'/etc/hadoop/tt.incl',
'mapred.hosts.exclude':
'/etc/hadoop/tt.excl',
}
cfg.update(mr_cfg)
if oozie_hostname:
o_cfg = {
'hadoop.proxyuser.hadoop.hosts': "localhost," + oozie_hostname,
'hadoop.proxyuser.hadoop.groups': 'hadoop',
}
cfg.update(o_cfg)
LOG.debug('Applied Oozie configs for core-site.xml')
cfg.update(o_h.get_oozie_required_xml_configs())
LOG.debug('Applied Oozie configs for oozie-site.xml')
if hive_hostname:
h_cfg = {
'hive.warehouse.subdir.inherit.perms':
True,
'javax.jdo.option.ConnectionURL':
'jdbc:derby:;databaseName=/opt/hive/metastore_db;create=true'
}
cfg.update(h_cfg)
LOG.debug('Applied Hive config for hive metastore server')
return cfg
def scale_cluster(self, cluster, instances):
self._setup_instances(cluster, instances)
run.refresh_nodes(remote.get_remote(vu.get_namenode(cluster)), "dfsadmin")
jt = vu.get_jobtracker(cluster)
if jt:
run.refresh_nodes(remote.get_remote(jt), "mradmin")
self._start_tt_dn_processes(instances)
def scale_cluster(self, cluster, instances):
self._setup_instances(cluster, instances)
run.refresh_nodes(remote.get_remote(vu.get_namenode(cluster)),
"dfsadmin")
jt = vu.get_jobtracker(cluster)
if jt:
run.refresh_nodes(remote.get_remote(jt), "mradmin")
self._start_tt_dn_processes(instances)
def generate_sahara_configs(cluster, node_group=None):
nn_hostname = vu.get_instance_hostname(vu.get_namenode(cluster))
jt_hostname = vu.get_instance_hostname(vu.get_jobtracker(cluster))
oozie_hostname = vu.get_instance_hostname(vu.get_oozie(cluster))
hive_hostname = vu.get_instance_hostname(vu.get_hiveserver(cluster))
storage_path = node_group.storage_paths() if node_group else None
# inserting common configs depends on provisioned VMs and HDFS placement
# TODO(aignatov): should be moved to cluster context
cfg = {
'fs.default.name': 'hdfs://%s:8020' % nn_hostname,
'dfs.name.dir': extract_hadoop_path(storage_path,
'/lib/hadoop/hdfs/namenode'),
'dfs.data.dir': extract_hadoop_path(storage_path,
'/lib/hadoop/hdfs/datanode'),
'dfs.hosts': '/etc/hadoop/dn.incl',
'dfs.hosts.exclude': '/etc/hadoop/dn.excl',
}
if jt_hostname:
mr_cfg = {
'mapred.job.tracker': '%s:8021' % jt_hostname,
'mapred.system.dir': extract_hadoop_path(storage_path,
'/mapred/mapredsystem'),
'mapred.local.dir': extract_hadoop_path(storage_path,
'/lib/hadoop/mapred'),
'mapred.hosts': '/etc/hadoop/tt.incl',
'mapred.hosts.exclude': '/etc/hadoop/tt.excl',
}
cfg.update(mr_cfg)
if oozie_hostname:
o_cfg = {
'hadoop.proxyuser.hadoop.hosts': "localhost," + oozie_hostname,
'hadoop.proxyuser.hadoop.groups': 'hadoop',
}
cfg.update(o_cfg)
LOG.debug('Applied Oozie configs for core-site.xml')
cfg.update(o_h.get_oozie_required_xml_configs())
LOG.debug('Applied Oozie configs for oozie-site.xml')
if hive_hostname:
h_cfg = {
'hive.warehouse.subdir.inherit.perms': True,
'javax.jdo.option.ConnectionURL':
'jdbc:derby:;databaseName=/opt/hive/metastore_db;create=true'
}
cfg.update(h_cfg)
LOG.debug('Applied Hive config for hive metastore server')
return cfg
def start_cluster(self, cluster):
nn_instance = vu.get_namenode(cluster)
with remote.get_remote(nn_instance) as r:
run.format_namenode(r)
run.start_processes(r, "namenode")
for snn in vu.get_secondarynamenodes(cluster):
run.start_processes(remote.get_remote(snn), "secondarynamenode")
jt_instance = vu.get_jobtracker(cluster)
if jt_instance:
run.start_processes(remote.get_remote(jt_instance), "jobtracker")
self._start_tt_dn_processes(utils.get_instances(cluster))
self._await_datanodes(cluster)
LOG.info(_LI("Hadoop services in cluster %s have been started"),
cluster.name)
oozie = vu.get_oozie(cluster)
if oozie:
with remote.get_remote(oozie) as r:
if c_helper.is_mysql_enable(cluster):
run.mysql_start(r, oozie)
run.oozie_create_db(r)
run.oozie_share_lib(r, nn_instance.hostname())
run.start_oozie(r)
LOG.info(_LI("Oozie service at '%s' has been started"),
nn_instance.hostname())
hive_server = vu.get_hiveserver(cluster)
if hive_server:
with remote.get_remote(hive_server) as r:
run.hive_create_warehouse_dir(r)
run.hive_copy_shared_conf(
r, edp.get_hive_shared_conf_path('hadoop'))
if c_helper.is_mysql_enable(cluster):
if not oozie or hive_server.hostname() != oozie.hostname():
run.mysql_start(r, hive_server)
run.hive_create_db(r)
run.hive_metastore_start(r)
LOG.info(_LI("Hive Metastore server at %s has been "
"started"),
hive_server.hostname())
LOG.info(_LI('Cluster %s has been started successfully'), cluster.name)
self._set_cluster_info(cluster)
def start_cluster(self, cluster):
nn_instance = vu.get_namenode(cluster)
with remote.get_remote(nn_instance) as r:
run.format_namenode(r)
run.start_processes(r, "namenode")
for snn in vu.get_secondarynamenodes(cluster):
run.start_processes(remote.get_remote(snn), "secondarynamenode")
jt_instance = vu.get_jobtracker(cluster)
if jt_instance:
run.start_processes(remote.get_remote(jt_instance), "jobtracker")
self._start_tt_dn_processes(utils.get_instances(cluster))
self._await_datanodes(cluster)
LOG.info(_LI("Hadoop services in cluster %s have been started"),
cluster.name)
oozie = vu.get_oozie(cluster)
if oozie:
with remote.get_remote(oozie) as r:
if c_helper.is_mysql_enable(cluster):
run.mysql_start(r, oozie)
run.oozie_create_db(r)
run.oozie_share_lib(r, nn_instance.hostname())
run.start_oozie(r)
LOG.info(_LI("Oozie service at '%s' has been started"),
nn_instance.hostname())
hive_server = vu.get_hiveserver(cluster)
if hive_server:
with remote.get_remote(hive_server) as r:
run.hive_create_warehouse_dir(r)
run.hive_copy_shared_conf(
r, edp.get_hive_shared_conf_path('hadoop'))
if c_helper.is_mysql_enable(cluster):
if not oozie or hive_server.hostname() != oozie.hostname():
run.mysql_start(r, hive_server)
run.hive_create_db(r, cluster.extra['hive_mysql_passwd'])
run.hive_metastore_start(r)
LOG.info(
_LI("Hive Metastore server at %s has been "
"started"), hive_server.hostname())
LOG.info(_LI('Cluster %s has been started successfully'), cluster.name)
self._set_cluster_info(cluster)
def _validate_additional_ng_scaling(self, cluster, additional):
jt = vu.get_jobtracker(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, "Vanilla plugin cannot scale nodegroup"
" with processes: " + ' '.join(ng.node_processes))
if not jt and 'tasktracker' in ng.node_processes:
raise ex.NodeGroupCannotBeScaled(
ng.name, "Vanilla plugin cannot scale node group with "
"processes which have no master-processes run "
"in cluster")
def _validate_additional_ng_scaling(self, cluster, additional):
jt = vu.get_jobtracker(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, _("Vanilla plugin cannot scale nodegroup"
" with processes: %s") %
' '.join(ng.node_processes))
if not jt and 'tasktracker' in ng.node_processes:
raise ex.NodeGroupCannotBeScaled(
ng.name, _("Vanilla plugin cannot scale node group with "
"processes which have no master-processes run "
"in cluster"))
def decommission_nodes(self, cluster, instances):
tts = vu.get_tasktrackers(cluster)
dns = vu.get_datanodes(cluster)
decommission_dns = False
decommission_tts = False
for i in instances:
if 'datanode' in i.node_group.node_processes:
dns.remove(i)
decommission_dns = True
if 'tasktracker' in i.node_group.node_processes:
tts.remove(i)
decommission_tts = True
nn = vu.get_namenode(cluster)
jt = vu.get_jobtracker(cluster)
if decommission_tts:
sc.decommission_tt(jt, instances, tts)
if decommission_dns:
sc.decommission_dn(nn, instances, dns)
def decommission_nodes(self, cluster, instances):
tts = vu.get_tasktrackers(cluster)
dns = vu.get_datanodes(cluster)
decommission_dns = False
decommission_tts = False
for i in instances:
if "datanode" in i.node_group.node_processes:
dns.remove(i)
decommission_dns = True
if "tasktracker" in i.node_group.node_processes:
tts.remove(i)
decommission_tts = True
nn = vu.get_namenode(cluster)
jt = vu.get_jobtracker(cluster)
if decommission_tts:
sc.decommission_tt(jt, instances, tts)
if decommission_dns:
sc.decommission_dn(nn, instances, dns)
def _set_cluster_info(self, cluster):
nn = vu.get_namenode(cluster)
jt = vu.get_jobtracker(cluster)
oozie = vu.get_oozie(cluster)
info = {}
if jt:
ui_port = c_helper.get_port_from_config(
'MapReduce', 'mapred.job.tracker.http.address', cluster)
jt_port = c_helper.get_port_from_config('MapReduce',
'mapred.job.tracker',
cluster)
info['MapReduce'] = {
'Web UI': 'http://%s:%s' % (jt.management_ip, ui_port),
'JobTracker': '%s:%s' % (jt.hostname(), jt_port)
}
if nn:
ui_port = c_helper.get_port_from_config('HDFS', 'dfs.http.address',
cluster)
nn_port = c_helper.get_port_from_config('HDFS', 'fs.default.name',
cluster)
info['HDFS'] = {
'Web UI': 'http://%s:%s' % (nn.management_ip, ui_port),
'NameNode': 'hdfs://%s:%s' % (nn.hostname(), nn_port)
}
if oozie:
# TODO(yrunts) change from hardcode value
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):
nn = vu.get_namenode(cluster)
jt = vu.get_jobtracker(cluster)
oozie = vu.get_oozie(cluster)
info = {}
if jt:
ui_port = c_helper.get_port_from_config(
'MapReduce', 'mapred.job.tracker.http.address', cluster)
jt_port = c_helper.get_port_from_config(
'MapReduce', 'mapred.job.tracker', cluster)
info['MapReduce'] = {
'Web UI': 'http://%s:%s' % (jt.management_ip, ui_port),
'JobTracker': '%s:%s' % (jt.hostname(), jt_port)
}
if nn:
ui_port = c_helper.get_port_from_config('HDFS', 'dfs.http.address',
cluster)
nn_port = c_helper.get_port_from_config('HDFS', 'fs.default.name',
cluster)
info['HDFS'] = {
'Web UI': 'http://%s:%s' % (nn.management_ip, ui_port),
'NameNode': 'hdfs://%s:%s' % (nn.hostname(), nn_port)
}
if oozie:
# TODO(yrunts) change from hardcode value
info['JobFlow'] = {
'Oozie': 'http://%s:11000' % oozie.management_ip
}
ctx = context.ctx()
conductor.cluster_update(ctx, cluster, {'info': info})
def start_jobtracker(self, cluster):
jt = vu.get_jobtracker(cluster)
if jt:
self._start_jobtracker(jt)
def start_jobtracker(self, cluster):
jt = vu.get_jobtracker(cluster)
if jt:
self._start_jobtracker(jt)
