def _provision_cluster(cluster_id):
ctx, cluster, plugin = _prepare_provisioning(cluster_id)
if CONF.use_identity_api_v3 and cluster.is_transient:
trusts.create_trust_for_cluster(cluster)
# updating cluster infra
cluster = g.change_cluster_status(cluster, "InfraUpdating")
plugin.update_infra(cluster)
# creating instances and configuring them
cluster = conductor.cluster_get(ctx, cluster_id)
INFRA.create_cluster(cluster)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
# configure cluster
cluster = g.change_cluster_status(cluster, "Configuring")
try:
plugin.configure_cluster(cluster)
except Exception as ex:
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
LOG.exception(
_LE("Can't configure cluster '%(name)s' (reason: %(reason)s)"),
{'name': cluster.name, 'reason': ex})
g.change_cluster_status(cluster, "Error")
return
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
# starting prepared and configured cluster
cluster = g.change_cluster_status(cluster, "Starting")
try:
plugin.start_cluster(cluster)
except Exception as ex:
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
LOG.exception(
_LE("Can't start services for cluster '%(name)s' (reason: "
"%(reason)s)"), {'name': cluster.name, 'reason': ex})
g.change_cluster_status(cluster, "Error")
return
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
# cluster is now up and ready
cluster = g.change_cluster_status(cluster, "Active")
# schedule execution pending job for cluster
for je in conductor.job_execution_get_all(ctx, cluster_id=cluster.id):
job_manager.run_job(je.id)
def scale_cluster(self, cluster, node_group_id_map):
ctx = context.ctx()
instance_ids = []
try:
instance_ids = self._scale_cluster_instances(cluster,
node_group_id_map)
cluster = conductor.cluster_get(ctx, cluster)
g.clean_cluster_from_empty_ng(cluster)
cluster = conductor.cluster_get(ctx, cluster)
instances = g.get_instances(cluster, instance_ids)
self._await_active(cluster, instances)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return []
self._assign_floating_ips(instances)
self._await_networks(cluster, instances)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return []
cluster = conductor.cluster_get(ctx, cluster)
volumes.attach_to_instances(
g.get_instances(cluster, instance_ids))
except Exception as ex:
with excutils.save_and_reraise_exception():
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return []
self._log_operation_exception(
"Can't scale cluster '%s' (reason: %s)", cluster, ex)
cluster = conductor.cluster_get(ctx, cluster)
self._rollback_cluster_scaling(
cluster, g.get_instances(cluster, instance_ids), ex)
instance_ids = []
cluster = conductor.cluster_get(ctx, cluster)
g.clean_cluster_from_empty_ng(cluster)
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Active"})
LOG.info(g.format_cluster_status(cluster))
# we should be here with valid cluster: if instances creation
# was not successful all extra-instances will be removed above
if instance_ids:
self._configure_instances(cluster)
return instance_ids
def create_cluster(self, cluster):
ctx = context.ctx()
try:
# create all instances
conductor.cluster_update(ctx, cluster, {"status": "Spawning"})
LOG.info(g.format_cluster_status(cluster))
self._create_instances(cluster)
# wait for all instances are up and networks ready
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Waiting"})
LOG.info(g.format_cluster_status(cluster))
instances = g.get_instances(cluster)
self._await_active(cluster, instances)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
self._assign_floating_ips(instances)
self._await_networks(cluster, instances)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
cluster = conductor.cluster_get(ctx, cluster)
# attach volumes
volumes.attach(cluster)
# prepare all instances
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Preparing"})
LOG.info(g.format_cluster_status(cluster))
self._configure_instances(cluster)
except Exception as ex:
with excutils.save_and_reraise_exception():
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
self._log_operation_exception(
"Can't start cluster '%s' (reason: %s)", cluster, ex)
cluster = conductor.cluster_update(
ctx, cluster, {"status": "Error",
"status_description": str(ex)})
LOG.info(g.format_cluster_status(cluster))
self._rollback_cluster_creation(cluster, ex)
def _provision_scaled_cluster(cluster_id, node_group_id_map):
ctx, cluster, plugin = _prepare_provisioning(cluster_id)
# Decommissioning surplus nodes with the plugin
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Decommissioning"})
LOG.info(g.format_cluster_status(cluster))
instances_to_delete = []
for node_group in cluster.node_groups:
new_count = node_group_id_map[node_group.id]
if new_count < node_group.count:
instances_to_delete += node_group.instances[new_count:
node_group.count]
if instances_to_delete:
plugin.decommission_nodes(cluster, instances_to_delete)
# Scaling infrastructure
cluster = conductor.cluster_update(ctx, cluster, {"status": "Scaling"})
LOG.info(g.format_cluster_status(cluster))
instances = INFRA.scale_cluster(cluster, node_group_id_map)
# Setting up new nodes with the plugin
if instances:
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Configuring"})
LOG.info(g.format_cluster_status(cluster))
try:
instances = g.get_instances(cluster, instances)
plugin.scale_cluster(cluster, instances)
except Exception as ex:
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
LOG.exception("Can't scale cluster '%s' (reason: %s)",
cluster.name, ex)
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Error"})
LOG.info(g.format_cluster_status(cluster))
return
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
cluster = conductor.cluster_update(ctx, cluster, {"status": "Active"})
LOG.info(g.format_cluster_status(cluster))
def launch_instances(self, ctx, cluster, target_count):
# create all instances
cluster = conductor.cluster_update(ctx, cluster, {"status": self.STAGES[0]})
LOG.info(g.format_cluster_status(cluster))
tmpl = heat.ClusterTemplate(cluster)
self._configure_template(ctx, tmpl, cluster, target_count)
stack = tmpl.instantiate(update_existing=self.UPDATE_STACK)
stack.wait_till_active()
self.inst_ids = self._populate_cluster(ctx, cluster, stack)
# wait for all instances are up and networks ready
cluster = conductor.cluster_update(ctx, cluster, {"status": self.STAGES[1]})
LOG.info(g.format_cluster_status(cluster))
instances = g.get_instances(cluster, self.inst_ids)
self._await_networks(cluster, instances)
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
# prepare all instances
cluster = conductor.cluster_update(ctx, cluster, {"status": self.STAGES[2]})
LOG.info(g.format_cluster_status(cluster))
instances = g.get_instances(cluster, self.inst_ids)
volumes.mount_to_instances(instances)
self._configure_instances(cluster)
def create_cluster(self, cluster):
ctx = context.ctx()
launcher = _CreateLauncher()
try:
target_count = self._get_ng_counts(cluster)
self._nullify_ng_counts(cluster)
cluster = conductor.cluster_get(ctx, cluster)
launcher.launch_instances(ctx, cluster, target_count)
except Exception as ex:
with excutils.save_and_reraise_exception():
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
self._log_operation_exception(
"Can't start cluster '%s' (reason: %s)", cluster, ex)
cluster = conductor.cluster_update(
ctx, cluster, {"status": "Error",
"status_description": str(ex)})
LOG.info(g.format_cluster_status(cluster))
self._rollback_cluster_creation(cluster)
def create_cluster(self, cluster):
ctx = context.ctx()
launcher = _CreateLauncher()
try:
target_count = self._get_ng_counts(cluster)
self._nullify_ng_counts(cluster)
cluster = conductor.cluster_get(ctx, cluster)
launcher.launch_instances(ctx, cluster, target_count)
cluster = conductor.cluster_get(ctx, cluster)
self._add_volumes(ctx, cluster)
except Exception as ex:
with excutils.save_and_reraise_exception():
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
self._log_operation_exception(
_LW("Can't start cluster '%(cluster)s' "
"(reason: %(reason)s)"), cluster, ex)
cluster = g.change_cluster_status(
cluster, "Error", status_description=six.text_type(ex))
self._rollback_cluster_creation(cluster)
def _provision_scaled_cluster(cluster_id, node_group_id_map):
ctx, cluster, plugin = _prepare_provisioning(cluster_id)
# Decommissioning surplus nodes with the plugin
cluster = g.change_cluster_status(cluster, "Decommissioning")
instances_to_delete = []
for node_group in cluster.node_groups:
new_count = node_group_id_map[node_group.id]
if new_count < node_group.count:
instances_to_delete += node_group.instances[new_count:
node_group.count]
if instances_to_delete:
plugin.decommission_nodes(cluster, instances_to_delete)
# Scaling infrastructure
cluster = g.change_cluster_status(cluster, "Scaling")
instances = INFRA.scale_cluster(cluster, node_group_id_map)
# Setting up new nodes with the plugin
if instances:
cluster = g.change_cluster_status(cluster, "Configuring")
try:
instances = g.get_instances(cluster, instances)
plugin.scale_cluster(cluster, instances)
except Exception as ex:
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
LOG.exception(
_LE("Can't scale cluster '%(name)s' (reason: %(reason)s)"),
{'name': cluster.name, 'reason': ex})
g.change_cluster_status(cluster, "Error")
return
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
g.change_cluster_status(cluster, "Active")
def scale_cluster(self, cluster, target_count):
ctx = context.ctx()
rollback_count = self._get_ng_counts(cluster)
launcher = _ScaleLauncher()
try:
launcher.launch_instances(ctx, cluster, target_count)
except Exception as ex:
with excutils.save_and_reraise_exception():
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
self._log_operation_exception(
"Can't scale cluster '%s' (reason: %s)", cluster, ex)
cluster = conductor.cluster_get(ctx, cluster)
try:
self._rollback_cluster_scaling(
ctx, cluster, rollback_count, target_count)
except Exception:
if not g.check_cluster_exists(cluster):
LOG.info(g.format_cluster_deleted_message(cluster))
return
# if something fails during the rollback, we stop
# doing anything further
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Error"})
LOG.info(g.format_cluster_status(cluster))
LOG.error("Unable to complete rollback, aborting")
raise
cluster = conductor.cluster_update(ctx, cluster,
{"status": "Active"})
LOG.info(g.format_cluster_status(cluster))
LOG.warn(
"Rollback successful. Throwing off an initial exception.")
finally:
cluster = conductor.cluster_get(ctx, cluster)
g.clean_cluster_from_empty_ng(cluster)
return launcher.inst_ids
