def get_alerts_data(self, service=None):
if self._data is not None:
# return cached data
return self._data.get(service, []) if service else self._data
self._data = {}
self._cluster_services = []
try:
ambari = plugin_utils.get_instance(self.cluster,
p_common.AMBARI_SERVER)
password = self.cluster.extra.get("ambari_password")
with client.AmbariClient(ambari, password=password) as ambari:
resp = ambari.get_alerts_data(self.cluster)
for alert in resp:
alert = alert.get('Alert', {})
service = alert.get('service_name').lower()
if service not in self._data:
self._data[service] = []
self._cluster_services.append(service)
self._data[service].append(alert)
except Exception as e:
prefix = _("Can't get response from Ambari Monitor")
msg = _("%(problem)s: %(description)s") % {
'problem': prefix,
'description': six.text_type(e)
}
# don't put in exception to logs, it will be done by log.exception
LOG.exception(prefix)
self._exception_store = msg
def check_health(self):
imp_map = {'OK': 'GREEN', 'WARNING': 'YELLOW', 'CRITICAL': 'RED'}
other_map = {'OK': 'GREEN'}
color_counter = collections.Counter()
important_services = self.get_important_services()
for alert in self.provider.get_alerts_data(self.service):
alert_summary = alert.get('state', 'UNKNOWN')
if self.service in important_services:
target = imp_map.get(alert_summary, 'RED')
else:
target = other_map.get(alert_summary, 'YELLOW')
color_counter[target] += 1
if color_counter['RED'] > 0 and color_counter['YELLOW'] > 0:
raise health_check_base.RedHealthError(
_("Ambari Monitor has responded that cluster has "
"%(red)d critical and %(yellow)d warning alert(s)") % {
'red': color_counter['RED'],
'yellow': color_counter['YELLOW']
})
elif color_counter['RED'] > 0:
raise health_check_base.RedHealthError(
_("Ambari Monitor has responded that cluster has "
"%(red)d critical alert(s)") % {'red': color_counter['RED']})
elif color_counter['YELLOW'] > 0:
raise health_check_base.YellowHealthError(
_("Ambari Monitor has responded that cluster "
"has %d warning alert(s)") % color_counter['YELLOW'])
return _("No alerts found")
def _check_storm(cluster):
dr_count = utils.get_instances_count(cluster, common.DRPC_SERVER)
ni_count = utils.get_instances_count(cluster, common.NIMBUS)
su_count = utils.get_instances_count(cluster, common.STORM_UI_SERVER)
sv_count = utils.get_instances_count(cluster, common.SUPERVISOR)
if dr_count > 1:
raise ex.InvalidComponentCountException(common.DRPC_SERVER,
_("0 or 1"), dr_count)
if ni_count > 1:
raise ex.InvalidComponentCountException(common.NIMBUS, _("0 or 1"),
ni_count)
if su_count > 1:
raise ex.InvalidComponentCountException(common.STORM_UI_SERVER,
_("0 or 1"), su_count)
if dr_count == 0 and ni_count == 1:
raise ex.RequiredServiceMissingException(common.DRPC_SERVER,
required_by=common.NIMBUS)
if dr_count == 1 and ni_count == 0:
raise ex.RequiredServiceMissingException(
common.NIMBUS, required_by=common.DRPC_SERVER)
if su_count == 1 and (dr_count == 0 or ni_count == 0):
raise ex.RequiredServiceMissingException(
common.NIMBUS, required_by=common.STORM_UI_SERVER)
if dr_count == 1 and sv_count == 0:
raise ex.RequiredServiceMissingException(
common.SUPERVISOR, required_by=common.DRPC_SERVER)
if sv_count > 0 and dr_count == 0:
raise ex.RequiredServiceMissingException(common.DRPC_SERVER,
required_by=common.SUPERVISOR)
def check_health(self):
important_services = self.provider.get_important_services()
observed_data = self.provider.get_health_status(self.service)
imp_map = {'BAD': 'red', 'CONCERNING': 'yellow', 'GOOD': 'green'}
summary = observed_data['summary']
checks = observed_data.get('checks', [])
failed_checks = []
for check in checks:
if check['summary'] != 'GOOD':
failed_checks.append('%(name)s - %(summary)s state' % {
'name': check['name'],
'summary': check['summary']
})
additional_info = None
if failed_checks:
additional_info = _("The following checks did not pass: %s"
) % ",".join(failed_checks)
if self.service in important_services:
overall = imp_map.get(summary, 'red')
else:
overall = 'green'
if summary != 'GOOD':
overall = 'yellow'
msg = _("Cloudera Manager has responded that service is in "
"the %s state") % summary
if additional_info:
msg = _("%(problem)s. %(description)s") % {
'problem': msg,
'description': additional_info
}
if overall == 'red':
raise health_check_base.RedHealthError(msg)
elif overall == 'yellow':
raise health_check_base.YellowHealthError(msg)
return msg
def _create_config_obj(self, item, target='general', scope='cluster',
high_priority=False):
def _prepare_value(value):
if isinstance(value, str):
return value.strip().lower()
return value
conf_name = _prepare_value(item.get('name', None))
conf_value = _prepare_value(item.get('value', None))
if not conf_name:
raise ex.HadoopProvisionError(_("Config missing 'name'"))
if conf_value is None:
raise ex.PluginInvalidDataException(
_("Config '%s' missing 'value'") % conf_name)
if high_priority or item.get('priority', 2) == 1:
priority = 1
else:
priority = 2
return p.Config(
name=conf_name,
applicable_target=target,
scope=scope,
config_type=item.get('config_type', "string"),
config_values=item.get('config_values', None),
default_value=conf_value,
is_optional=item.get('is_optional', True),
description=item.get('description', None),
priority=priority)
def _hue_validation(cls, cluster):
hue_count = cls.get_inst_count(cluster, 'HUE_SERVER')
if hue_count > 1:
raise ex.InvalidComponentCountException('HUE_SERVER', _('0 or 1'),
hue_count)
shs_count = cls.get_inst_count(cluster, 'SPARK_YARN_HISTORY_SERVER')
hms_count = cls.get_inst_count(cluster, 'HIVE_METASTORE')
oo_count = cls.get_inst_count(cluster, 'OOZIE_SERVER')
rm_count = cls.get_inst_count(cluster, 'YARN_RESOURCEMANAGER')
if shs_count > 1:
raise ex.InvalidComponentCountException(
'SPARK_YARN_HISTORY_SERVER', _('0 or 1'), shs_count)
if shs_count and not rm_count:
raise ex.RequiredServiceMissingException(
'YARN_RESOURCEMANAGER',
required_by='SPARK_YARN_HISTORY_SERVER')
if oo_count < 1 and hue_count:
raise ex.RequiredServiceMissingException('OOZIE_SERVER',
required_by='HUE_SERVER')
if hms_count < 1 and hue_count:
raise ex.RequiredServiceMissingException('HIVE_METASTORE',
required_by='HUE_SERVER')
def _await_cldb(self, cluster_context, instances=None, timeout=600):
instances = instances or cluster_context.get_instances()
cldb_node = cluster_context.get_instance(mfs.CLDB)
start_time = timeutils.utcnow()
retry_count = 0
with cldb_node.remote() as r:
LOG.debug("Waiting {count} seconds for CLDB initialization".format(
count=timeout))
while timeutils.delta_seconds(start_time,
timeutils.utcnow()) < timeout:
ec, out = r.execute_command(NODE_LIST_CMD,
raise_when_error=False)
resp = json.loads(out)
status = resp['status']
if str(status).lower() == 'ok':
ips = [n['ip'] for n in resp['data']]
retry_count += 1
for i in instances:
if (i.internal_ip not in ips
and (retry_count > DEFAULT_RETRY_COUNT)):
msg = _("Node failed to connect to CLDB: %s"
) % i.internal_ip
raise ex.HadoopProvisionError(msg)
break
else:
context.sleep(DELAY)
else:
raise ex.HadoopProvisionError(_("CLDB failed to start"))
def _hdfs_ha_validation(cls, cluster):
jn_count = cls.get_inst_count(cluster, 'HDFS_JOURNALNODE')
zk_count = cls.get_inst_count(cluster, 'ZOOKEEPER_SERVER')
require_anti_affinity = cls.PU.c_helper.get_required_anti_affinity(
cluster)
if jn_count > 0:
if jn_count < 3:
raise ex.InvalidComponentCountException(
'HDFS_JOURNALNODE', _('not less than 3'), jn_count)
if not jn_count % 2:
raise ex.InvalidComponentCountException(
'HDFS_JOURNALNODE', _('be odd'), jn_count)
if zk_count < 1:
raise ex.RequiredServiceMissingException('ZOOKEEPER',
required_by='HDFS HA')
if require_anti_affinity:
if 'HDFS_SECONDARYNAMENODE' not in \
cls._get_anti_affinity(cluster):
raise ex.NameNodeHAConfigurationError(
_('HDFS_SECONDARYNAMENODE should be enabled '
'in anti_affinity.'))
if 'HDFS_NAMENODE' not in cls._get_anti_affinity(cluster):
raise ex.NameNodeHAConfigurationError(
_('HDFS_NAMENODE should be enabled in anti_affinity.'))
def _impala_validation(cls, cluster):
ics_count = cls.get_inst_count(cluster, 'IMPALA_CATALOGSERVER')
iss_count = cls.get_inst_count(cluster, 'IMPALA_STATESTORE')
id_count = cls.get_inst_count(cluster, 'IMPALAD')
dn_count = cls.get_inst_count(cluster, 'HDFS_DATANODE')
hms_count = cls.get_inst_count(cluster, 'HIVE_METASTORE')
if ics_count > 1:
raise ex.InvalidComponentCountException('IMPALA_CATALOGSERVER',
_('0 or 1'), ics_count)
if iss_count > 1:
raise ex.InvalidComponentCountException('IMPALA_STATESTORE',
_('0 or 1'), iss_count)
if ics_count == 1:
datanode_ng = u.get_node_groups(cluster, "HDFS_DATANODE")
impalad_ng = u.get_node_groups(cluster, "IMPALAD")
datanodes = set(ng.id for ng in datanode_ng)
impalads = set(ng.id for ng in impalad_ng)
if datanodes != impalads:
raise ex.InvalidClusterTopology(
_("IMPALAD must be installed on every HDFS_DATANODE"))
if iss_count != 1:
raise ex.RequiredServiceMissingException('IMPALA_STATESTORE',
required_by='IMPALA')
if id_count < 1:
raise ex.RequiredServiceMissingException('IMPALAD',
required_by='IMPALA')
if dn_count < 1:
raise ex.RequiredServiceMissingException('HDFS_DATANODE',
required_by='IMPALA')
if hms_count < 1:
raise ex.RequiredServiceMissingException('HIVE_METASTORE',
required_by='IMPALA')
def _yarn_ha_validation(cls, cluster):
rm_count = cls.get_inst_count(cluster, 'YARN_RESOURCEMANAGER')
zk_count = cls.get_inst_count(cluster, 'ZOOKEEPER_SERVER')
stdb_rm_count = cls.get_inst_count(cluster, 'YARN_STANDBYRM')
require_anti_affinity = cls.PU.c_helper.get_required_anti_affinity(
cluster)
if stdb_rm_count > 1:
raise ex.InvalidComponentCountException('YARN_STANDBYRM',
_('0 or 1'), stdb_rm_count)
if stdb_rm_count > 0:
if rm_count < 1:
raise ex.RequiredServiceMissingException(
'YARN_RESOURCEMANAGER', required_by='RM HA')
if zk_count < 1:
raise ex.RequiredServiceMissingException('ZOOKEEPER',
required_by='RM HA')
if require_anti_affinity:
if 'YARN_RESOURCEMANAGER' not in \
cls._get_anti_affinity(cluster):
raise ex.ResourceManagerHAConfigurationError(
_('YARN_RESOURCEMANAGER should be enabled in '
'anti_affinity.'))
if 'YARN_STANDBYRM' not in cls._get_anti_affinity(cluster):
raise ex.ResourceManagerHAConfigurationError(
_('YARN_STANDBYRM should be'
' enabled in anti_affinity.'))
def _get_ha_params():
enable_namenode_ha = provisioning.Config(
name=common.NAMENODE_HA,
applicable_target="general",
scope="cluster",
config_type="bool",
default_value=False,
is_optional=True,
description=_("Enable NameNode HA"),
priority=1)
enable_resourcemanager_ha = provisioning.Config(
name=common.RESOURCEMANAGER_HA,
applicable_target="general",
scope="cluster",
config_type="bool",
default_value=False,
is_optional=True,
description=_("Enable ResourceManager HA"),
priority=1)
enable_regionserver_ha = provisioning.Config(
name=common.HBASE_REGIONSERVER_HA,
applicable_target="general",
scope="cluster",
config_type="bool",
default_value=False,
is_optional=True,
description=_("Enable HBase RegionServer HA"),
priority=1)
return [
enable_namenode_ha, enable_resourcemanager_ha, enable_regionserver_ha
]
def _validate_existing_ng_scaling(self, cluster, existing):
scalable_processes = self._get_scalable_processes()
dn_to_delete = 0
for ng in cluster.node_groups:
if ng.id in existing:
if ng.count > existing[ng.id] and ("datanode" in
ng.node_processes):
dn_to_delete += ng.count - existing[ng.id]
if not set(ng.node_processes).issubset(scalable_processes):
raise ex.NodeGroupCannotBeScaled(
ng.name, _("Spark plugin cannot scale nodegroup"
" with processes: %s") %
' '.join(ng.node_processes))
dn_amount = len(utils.get_instances(cluster, "datanode"))
rep_factor = utils.get_config_value_or_default('HDFS',
"dfs.replication",
cluster)
if dn_to_delete > 0 and dn_amount - dn_to_delete < rep_factor:
raise ex.ClusterCannotBeScaled(
cluster.name, _("Spark plugin cannot shrink cluster because "
"there would be not enough nodes for HDFS "
"replicas (replication factor is %s)") %
rep_factor)
def wait_ambari_requests(self, requests, cluster_name):
requests = set(requests)
failed = []
context.sleep(20)
while len(requests) > 0:
completed, not_completed = set(), set()
for req_id in requests:
request = self.get_request_info(cluster_name, req_id)
status = request.get("request_status")
if status == 'COMPLETED':
completed.add(req_id)
elif status in ['IN_PROGRESS', 'PENDING']:
not_completed.add(req_id)
else:
failed.append(request)
if failed:
msg = _("Some Ambari request(s) "
"not in COMPLETED state: %(description)s.")
descrs = []
for req in failed:
descr = _(
"request %(id)d: %(name)s - in status %(status)s")
descrs.append(descr %
{'id': req.get("id"),
'name': req.get("request_context"),
'status': req.get("request_status")})
raise p_exc.HadoopProvisionError(msg % {'description': descrs})
requests = not_completed
context.sleep(5)
LOG.debug("Waiting for %d ambari request(s) to be completed",
len(not_completed))
LOG.debug("All ambari requests have been completed")
def get_service(self, node_process):
ui_name = self.get_service_name_by_node_process(node_process)
if ui_name is None:
raise e.PluginInvalidDataException(
_('Service not found in services list'))
version = self.get_chosen_service_version(ui_name)
service = self._find_service_instance(ui_name, version)
if service is None:
raise e.PluginInvalidDataException(_('Can not map service'))
return service
def _check_jn_ha(cluster):
jn_count = utils.get_instances_count(cluster, common.JOURNAL_NODE)
if jn_count < 3:
raise ex.InvalidComponentCountException(
common.JOURNAL_NODE, _("3 or more. Odd number"), jn_count,
_("At least 3 JournalNodes are required for HA"))
if jn_count % 2 != 1:
raise ex.InvalidComponentCountException(
common.JOURNAL_NODE, _("Odd number"), jn_count,
_("Odd number of JournalNodes are required for HA"))
def _check_zk_ha(cluster):
zk_count = utils.get_instances_count(cluster, common.ZOOKEEPER_SERVER)
if zk_count < 3:
raise ex.InvalidComponentCountException(
common.ZOOKEEPER_SERVER, _("3 or more. Odd number"), zk_count,
_("At least 3 ZooKeepers are required for HA"))
if zk_count % 2 != 1:
raise ex.InvalidComponentCountException(
common.ZOOKEEPER_SERVER, _("Odd number"), zk_count,
_("Odd number of ZooKeepers are required for HA"))
def stop(self, cluster_context, instances=None):
instances = instances or cluster_context.get_instances()
zookeepers = cluster_context.filter_instances(instances, mng.ZOOKEEPER)
utils.add_provisioning_step(cluster_context.cluster.id,
_("Stop ZooKeepers nodes"),
len(zookeepers))
self._stop_zk_nodes(zookeepers)
utils.add_provisioning_step(cluster_context.cluster.id,
_("Stop Warden nodes"), len(instances))
self._stop_warden_on_nodes(instances)
def _basic_validation(cls, cluster):
mng_count = cls.get_inst_count(cluster, 'CLOUDERA_MANAGER')
if mng_count != 1:
raise ex.InvalidComponentCountException('CLOUDERA_MANAGER', 1,
mng_count)
nn_count = cls.get_inst_count(cluster, 'HDFS_NAMENODE')
if nn_count != 1:
raise ex.InvalidComponentCountException('HDFS_NAMENODE', 1,
nn_count)
snn_count = cls.get_inst_count(cluster, 'HDFS_SECONDARYNAMENODE')
if snn_count != 1:
raise ex.InvalidComponentCountException('HDFS_SECONDARYNAMENODE',
1, snn_count)
dn_count = cls.get_inst_count(cluster, 'HDFS_DATANODE')
replicas = cls.PU.get_config_value('HDFS', 'dfs_replication', cluster)
if dn_count < replicas:
raise ex.InvalidComponentCountException(
'HDFS_DATANODE', replicas, dn_count,
_('Number of datanodes must be not'
' less than dfs_replication.'))
du_reserved = cls.PU.get_config_value('DATANODE',
'dfs_datanode_du_reserved',
cluster)
du_reserved = du_reserved / 1073741824.
for node_group in cluster.node_groups:
volume_size = node_group.volumes_size
if volume_size and volume_size < du_reserved:
raise ex.InvalidVolumeSizeException(volume_size, du_reserved)
rm_count = cls.get_inst_count(cluster, 'YARN_RESOURCEMANAGER')
if rm_count > 1:
raise ex.InvalidComponentCountException('YARN_RESOURCEMANAGER',
_('0 or 1'), rm_count)
hs_count = cls.get_inst_count(cluster, 'YARN_JOBHISTORY')
if hs_count > 1:
raise ex.InvalidComponentCountException('YARN_JOBHISTORY',
_('0 or 1'), hs_count)
if rm_count > 0 and hs_count < 1:
raise ex.RequiredServiceMissingException(
'YARN_JOBHISTORY', required_by='YARN_RESOURCEMANAGER')
nm_count = cls.get_inst_count(cluster, 'YARN_NODEMANAGER')
if rm_count == 0:
if nm_count > 0:
raise ex.RequiredServiceMissingException(
'YARN_RESOURCEMANAGER', required_by='YARN_NODEMANAGER')
class NodeRequiredServiceMissingException(e.RequiredServiceMissingException):
MISSING_MSG = _('Node "%(ng_name)s" is missing component %(component)s')
REQUIRED_MSG = _('%(message)s, required by %(required_by)s')
def __init__(self, service_name, ng_name, required_by=None):
super(NodeRequiredServiceMissingException,
self).__init__(service_name, required_by)
args = {'ng_name': ng_name, 'component': service_name}
self.message = (NodeRequiredServiceMissingException.MISSING_MSG % args)
if required_by:
args = {'message': self.message, 'required_by': required_by}
self.message = (NodeRequiredServiceMissingException.REQUIRED_MSG %
args)
def get_service_by_role(self, role, cluster=None, instance=None):
if cluster:
cm_cluster = self.get_cloudera_cluster(cluster)
elif instance:
cm_cluster = self.get_cloudera_cluster(instance.cluster)
else:
raise ValueError(_("'cluster' or 'instance' argument missed"))
if role in ['NAMENODE', 'DATANODE', 'SECONDARYNAMENODE',
'HDFS_GATEWAY']:
return cm_cluster.get_service(self.HDFS_SERVICE_NAME)
elif role in ['RESOURCEMANAGER', 'NODEMANAGER', 'JOBHISTORY',
'YARN_GATEWAY']:
return cm_cluster.get_service(self.YARN_SERVICE_NAME)
elif role in ['OOZIE_SERVER']:
return cm_cluster.get_service(self.OOZIE_SERVICE_NAME)
elif role in ['HIVESERVER2', 'HIVEMETASTORE', 'WEBHCAT']:
return cm_cluster.get_service(self.HIVE_SERVICE_NAME)
elif role in ['HUE_SERVER']:
return cm_cluster.get_service(self.HUE_SERVICE_NAME)
elif role in ['SPARK_YARN_HISTORY_SERVER']:
return cm_cluster.get_service(self.SPARK_SERVICE_NAME)
elif role in ['SERVER']:
return cm_cluster.get_service(self.ZOOKEEPER_SERVICE_NAME)
elif role in ['MASTER', 'REGIONSERVER']:
return cm_cluster.get_service(self.HBASE_SERVICE_NAME)
elif role in ['AGENT']:
return cm_cluster.get_service(self.FLUME_SERVICE_NAME)
elif role in ['SENTRY_SERVER']:
return cm_cluster.get_service(self.SENTRY_SERVICE_NAME)
elif role in ['SQOOP_SERVER']:
return cm_cluster.get_service(self.SQOOP_SERVICE_NAME)
elif role in ['SOLR_SERVER']:
return cm_cluster.get_service(self.SOLR_SERVICE_NAME)
elif role in ['HBASE_INDEXER']:
return cm_cluster.get_service(self.KS_INDEXER_SERVICE_NAME)
elif role in ['CATALOGSERVER', 'STATESTORE', 'IMPALAD', 'LLAMA']:
return cm_cluster.get_service(self.IMPALA_SERVICE_NAME)
elif role in ['KMS']:
return cm_cluster.get_service(self.KMS_SERVICE_NAME)
elif role in ['JOURNALNODE']:
return cm_cluster.get_service(self.HDFS_SERVICE_NAME)
elif role in ['YARN_STANDBYRM']:
return cm_cluster.get_service(self.YARN_SERVICE_NAME)
elif role in ['KAFKA_BROKER']:
return cm_cluster.get_service(self.KAFKA_SERVICE_NAME)
else:
raise ValueError(
_("Process %(process)s is not supported by CDH plugin") %
{'process': role})
def _check_hive(cluster):
hs_count = utils.get_instances_count(cluster, common.HIVE_SERVER)
hm_count = utils.get_instances_count(cluster, common.HIVE_METASTORE)
if hs_count > 1:
raise ex.InvalidComponentCountException(common.HIVE_SERVER,
_("0 or 1"), hs_count)
if hm_count > 1:
raise ex.InvalidComponentCountException(common.HIVE_METASTORE,
_("0 or 1"), hm_count)
if hs_count == 0 and hm_count == 1:
raise ex.RequiredServiceMissingException(
common.HIVE_SERVER, required_by=common.HIVE_METASTORE)
if hs_count == 1 and hm_count == 0:
raise ex.RequiredServiceMissingException(
common.HIVE_METASTORE, required_by=common.HIVE_SERVER)
def _check_ranger(cluster):
ra_count = utils.get_instances_count(cluster, common.RANGER_ADMIN)
ru_count = utils.get_instances_count(cluster, common.RANGER_USERSYNC)
if ra_count > 1:
raise ex.InvalidComponentCountException(common.RANGER_ADMIN,
_("0 or 1"), ra_count)
if ru_count > 1:
raise ex.InvalidComponentCountException(common.RANGER_USERSYNC,
_("0 or 1"), ru_count)
if ra_count == 1 and ru_count == 0:
raise ex.RequiredServiceMissingException(
common.RANGER_USERSYNC, required_by=common.RANGER_ADMIN)
if ra_count == 0 and ru_count == 1:
raise ex.RequiredServiceMissingException(
common.RANGER_ADMIN, required_by=common.RANGER_USERSYNC)
def validate_additional_ng_scaling(cluster, additional):
rm = vu.get_resourcemanager(cluster)
scalable_processes = _get_scalable_processes()
for ng_id in additional:
ng = u.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 _check_decommission(cluster, instances, check_func, option):
utils.plugin_option_poll(cluster, is_decommissioned, option,
_("Wait for decommissioning"), 5, {
'cluster': cluster,
'check_func': check_func,
'instances': instances
})
def validate_job_execution(self, cluster, job, data):
if not self.edp_supported(cluster.hadoop_version):
raise ex.InvalidDataException(
_('Storm {base} required to run {type} jobs').format(
base=EdpPyleusEngine.edp_base_version, type=job.type))
super(EdpPyleusEngine, self).validate_job_execution(cluster, job, data)
def invoke(self, method, relpath=None, params=None, data=None,
headers=None):
"""Invoke an API method
:return: Raw body or JSON dictionary (if response content type is
JSON).
"""
path = self._join_uri(relpath)
resp = self._client.execute(method,
path,
params=params,
data=data,
headers=headers)
try:
body = resp.read()
except Exception as ex:
raise ex.CMApiException(
_("Command %(method)s %(path)s failed: %(msg)s")
% {'method': method, 'path': path, 'msg': six.text_type(ex)})
LOG.debug("{method} got response: {body}".format(method=method,
body=body[:32]))
# Is the response application/json?
if (len(body) != 0 and resp.info().getmaintype() == "application"
and resp.info().getsubtype() == "json"):
try:
json_dict = json.loads(body)
return json_dict
except Exception:
LOG.error('JSON decode error: {body}'.format(body=body))
raise
else:
return body
def get(self, relpath=None, params=None):
"""Invoke the GET method on a resource
:param relpath: Optional. A relative path to this resource's path.
:param params: Key-value data.
:return: A dictionary of the JSON result.
"""
for retry in six.moves.xrange(self.retries + 1):
if retry:
context.sleep(self.retry_sleep)
try:
return self.invoke("GET", relpath, params)
except (socket.error, urllib.error.URLError) as e:
if "timed out" in six.text_type(e).lower():
if retry < self.retries:
LOG.warning("Timeout issuing GET request for "
"{path}. Will retry".format(
path=self._join_uri(relpath)))
else:
LOG.warning("Timeout issuing GET request for "
"{path}. No retries left".format(
path=self._join_uri(relpath)))
else:
raise
else:
raise ex.CMApiException(_("Get retry max time reached."))
def validate_additional_ng_scaling(cls, cluster, additional):
rm = cls.PU.get_resourcemanager(cluster)
scalable_processes = cls._get_scalable_processes()
for ng_id in additional:
ng = u.get_by_id(cluster.node_groups, ng_id)
if not set(ng.node_processes).issubset(scalable_processes):
msg = _("CDH plugin cannot scale nodegroup with processes: "
"%(processes)s")
raise ex.NodeGroupCannotBeScaled(
ng.name, msg % {'processes': ' '.join(ng.node_processes)})
if not rm and 'YARN_NODEMANAGER' in ng.node_processes:
msg = _("CDH plugin cannot scale node group with processes "
"which have no master-processes run in cluster")
raise ex.NodeGroupCannotBeScaled(ng.name, msg)
def _check_yarn(cluster):
rm_count = utils.get_instances_count(cluster, common.RESOURCEMANAGER)
nm_count = utils.get_instances_count(cluster, common.NODEMANAGER)
hs_count = utils.get_instances_count(cluster, common.HISTORYSERVER)
at_count = utils.get_instances_count(cluster, common.APP_TIMELINE_SERVER)
if cluster.cluster_configs.get("general",
{}).get(common.RESOURCEMANAGER_HA):
_check_zk_ha(cluster)
if rm_count != 2:
raise ex.InvalidComponentCountException(common.RESOURCEMANAGER, 2,
rm_count)
else:
if rm_count != 1:
raise ex.InvalidComponentCountException(common.RESOURCEMANAGER, 1,
rm_count)
if hs_count != 1:
raise ex.InvalidComponentCountException(common.HISTORYSERVER, 1,
hs_count)
if at_count != 1:
raise ex.InvalidComponentCountException(common.APP_TIMELINE_SERVER, 1,
at_count)
if nm_count == 0:
raise ex.InvalidComponentCountException(common.NODEMANAGER,
_("1 or more"), nm_count)
def check_health(self):
instances = self.cluster_context.get_instances(
node_process=management.ZOOKEEPER)
active_count = 0
for instance in instances:
if self._is_zookeeper_running(instance):
active_count += 1
if active_count == 0:
raise health_check_base.RedHealthError(_(
"ZooKeeper is not in running state"))
if active_count < len(instances):
raise health_check_base.YellowHealthError(_(
"Some ZooKeeper processes are not in running state"))
return _("ZooKeeper is in running state")