def __post_init__(self):
super().__post_init__()
if self.cluster_name is None or isinstance(self.cluster_name, str):
self.cluster_name = ClusterName() # todo rm cluster_name after removing config.cluster_name dependencies
self.entity_name = self.cluster_name
if self.kubeconfig_path is None:
self.kubeconfig_path = utils.get_kubeconfig_path(self.cluster_name.get())
def download_kubeconfig(self, kube_api_client: ApiClient) -> str:
log.info(f"Downloading kubeconfig for HyperShift cluster {self.name}")
kubeconfig_data = (Secret(
kube_api_client=kube_api_client,
namespace=f"clusters-{self.name}",
name="admin-kubeconfig",
).get().data["kubeconfig"])
hypershift_kubeconfig_path = utils.get_kubeconfig_path(
self.name) + "-hypershift"
with open(hypershift_kubeconfig_path, "wt") as kubeconfig_file:
kubeconfig_file.write(b64decode(kubeconfig_data).decode())
kubeconfig_file.flush()
self.kubeconfig_path = hypershift_kubeconfig_path
return self.kubeconfig_path
def nodes_flow_kube_api(cluster_name, machine_net, cluster_deployment, agent_cluster_install):
tf_folder = utils.get_tf_folder(cluster_name, args.namespace)
baremetal_template = os.path.join(tf_folder, consts.Platforms.BARE_METAL)
nodes_details = utils.get_tfvars(baremetal_template)
tf = terraform_utils.TerraformUtils(working_dir=baremetal_template)
is_ipv4 = machine_net.has_ip_v4 or not machine_net.has_ip_v6
nodes_number = args.master_count + args.number_of_workers
create_nodes_and_wait_till_registered(
inventory_client=None,
cluster=None,
nodes_details=nodes_details,
tf=tf,
is_ipv4=is_ipv4,
nodes_number=nodes_number,
cluster_deployment=cluster_deployment,
)
if args.master_count == 1:
set_single_node_ip(
client=None,
cluster_id=None,
main_cidr=args.vm_network_cidr if is_ipv4 else args.vm_network_cidr6,
is_ipv4=is_ipv4,
cluster_deployment=cluster_deployment,
tf=tf,
)
else:
log.info("VIPs already configured")
kubeapi_utils.set_agents_hostnames(
cluster_deployment=cluster_deployment,
is_ipv4=is_ipv4,
static_network_mode=args.with_static_network_config,
tf=tf,
nodes_number=nodes_number,
)
if args.install_cluster:
install_cluster.run_installation_flow_kube_api(
cluster_deployment=cluster_deployment,
agent_cluster_install=agent_cluster_install,
nodes_number=nodes_number,
kubeconfig_path=utils.get_kubeconfig_path(cluster_name)
)
def kubeconfig_path(self) -> str:
if self._kubeconfig_path == "":
log.info(
f"Downloading kubeconfig for HyperShift cluster {self.name}")
kubeconfig_data = (Secret(
kube_api_client=self.management_kube_api_client,
namespace=f"clusters-{self.name}",
name="admin-kubeconfig",
).get().data["kubeconfig"])
hypershift_kubeconfig_path = utils.get_kubeconfig_path(
self.name) + "-hypershift"
log.info(f"Kubeconfig path {hypershift_kubeconfig_path}")
with open(hypershift_kubeconfig_path, "wt") as kubeconfig_file:
kubeconfig_file.write(b64decode(kubeconfig_data).decode())
kubeconfig_file.flush()
self._kubeconfig_path = hypershift_kubeconfig_path
return self._kubeconfig_path
def start_install_and_wait_for_installed(self):
cluster_name = self.config.day1_cluster_name
# Running twice as a workaround for an issue with terraform not spawning a new node on first apply.
for _ in range(2):
with utils.file_lock_context():
utils.run_command(
f"make _apply_terraform CLUSTER_NAME={cluster_name} PLATFORM={consts.Platforms.BARE_METAL}"
)
time.sleep(5)
num_nodes_to_wait = self.config.day2_workers_count
installed_status = consts.NodesStatus.DAY2_INSTALLED
tfvars = utils.get_tfvars(self.config.tf_folder)
tf_network_name = tfvars["libvirt_network_name"]
config = TerraformConfig()
config.nodes_count = num_nodes_to_wait
libvirt_controller = LibvirtController(config=config,
entity_config=ClusterConfig())
libvirt_controller.wait_till_nodes_are_ready(
network_name=tf_network_name)
# Wait for day2 nodes
waiting.wait(
lambda: self.are_libvirt_nodes_in_cluster_hosts(),
timeout_seconds=consts.NODES_REGISTERED_TIMEOUT,
sleep_seconds=10,
waiting_for="Nodes to be registered in inventory service",
)
self.set_nodes_hostnames_if_needed(tf_network_name)
wait_till_all_hosts_are_in_status(
client=self.api_client,
cluster_id=self.config.cluster_id,
nodes_count=self.config.day2_workers_count,
statuses=[consts.NodesStatus.KNOWN],
interval=30,
)
# Start day2 nodes installation
log.info("Start installing all known nodes in the cluster %s",
self.config.cluster_id)
kubeconfig = utils.get_kubeconfig_path(self.config.day1_cluster_name)
ocp_ready_nodes = self.get_ocp_cluster_ready_nodes_num(kubeconfig)
hosts = self.api_client.get_cluster_hosts(self.config.cluster_id)
[
self.api_client.install_day2_host(self.config.infra_env_id,
host["id"]) for host in hosts
if host["status"] == "known"
]
log.info(
"Waiting until all nodes of cluster %s have been installed (reached added-to-existing-cluster)",
self.config.cluster_id,
)
wait_till_all_hosts_are_in_status(
client=self.api_client,
cluster_id=self.config.cluster_id,
nodes_count=num_nodes_to_wait,
statuses=[installed_status],
interval=30,
)
log.info(
"Waiting until installed nodes has actually been added to the OCP cluster"
)
waiting.wait(
lambda: self.wait_nodes_join_ocp_cluster(
ocp_ready_nodes, self.config.day2_workers_count, kubeconfig),
timeout_seconds=consts.NODES_REGISTERED_TIMEOUT,
sleep_seconds=30,
waiting_for="Day2 nodes to be added to OCP cluster",
expected_exceptions=Exception,
)
log.info("%d worker nodes were successfully added to OCP cluster",
self.config.day2_workers_count)
def day2_nodes_flow(client, terraform_cluster_dir_prefix, tf_folder, cluster,
has_ipv_6, num_worker_nodes, install_cluster_flag,
day2_type_flag, with_static_network_config,
base_cluster_name):
tf_network_name, total_num_nodes = get_network_num_nodes_from_tf(tf_folder)
# Running twice as a workaround for an issue with terraform not spawning a new node on first apply.
for _ in range(2):
with utils.file_lock_context():
utils.run_command(
f'make _apply_terraform CLUSTER_NAME={terraform_cluster_dir_prefix} PLATFORM={consts.Platforms.BARE_METAL}'
)
time.sleep(5)
if day2_type_flag == "ocp":
num_nodes_to_wait = total_num_nodes
installed_status = consts.NodesStatus.INSTALLED
else:
num_nodes_to_wait = num_worker_nodes
installed_status = consts.NodesStatus.DAY2_INSTALLED
wait_till_nodes_are_ready(nodes_count=num_nodes_to_wait,
network_name=tf_network_name)
waiting.wait(
lambda: are_libvirt_nodes_in_cluster_hosts(client, cluster.id,
num_nodes_to_wait),
timeout_seconds=consts.NODES_REGISTERED_TIMEOUT,
sleep_seconds=10,
waiting_for="Nodes to be registered in inventory service",
)
set_nodes_hostnames_if_needed(client, tf_folder,
with_static_network_config, has_ipv_6,
tf_network_name, cluster.id)
wait_till_all_hosts_are_in_status(
client=client,
cluster_id=cluster.id,
nodes_count=num_worker_nodes,
statuses=[consts.NodesStatus.KNOWN],
interval=30,
)
if install_cluster_flag:
log.info("Start installing all known nodes in the cluster %s",
cluster.id)
kubeconfig = utils.get_kubeconfig_path(base_cluster_name)
ocp_orig_ready_nodes = get_ocp_cluster_ready_nodes_num(kubeconfig)
hosts = client.get_cluster_hosts(cluster.id)
[
client.install_day2_host(cluster.id, host['id']) for host in hosts
if host["status"] == 'known'
]
log.info(
"Start waiting until all nodes of cluster %s have been installed( reached added-to-existing-clustertate)",
cluster.id)
wait_till_all_hosts_are_in_status(
client=client,
cluster_id=cluster.id,
nodes_count=num_nodes_to_wait,
statuses=[installed_status],
interval=30,
)
log.info(
"Start waiting until installed nodes has actually been added to the OCP cluster"
)
waiting.wait(lambda: wait_nodes_join_ocp_cluster(
ocp_orig_ready_nodes, num_worker_nodes, day2_type_flag, kubeconfig
),
timeout_seconds=consts.NODES_REGISTERED_TIMEOUT,
sleep_seconds=30,
waiting_for="Day2 nodes to be added to OCP cluster",
expected_exceptions=Exception)
log.info("%d worker nodes were successfully added to OCP cluster",
num_worker_nodes)
def nodes_flow(
client,
cluster_name,
cluster,
machine_net,
cluster_deployment=None,
agent_cluster_install=None,
):
tf_folder = utils.get_tf_folder(cluster_name, args.namespace)
nodes_details = utils.get_tfvars(tf_folder)
if cluster:
nodes_details["cluster_inventory_id"] = cluster.id
utils.set_tfvars(tf_folder, nodes_details)
tf = terraform_utils.TerraformUtils(working_dir=tf_folder)
is_ipv4 = machine_net.has_ip_v4 or not machine_net.has_ip_v6
nodes_number = args.master_count + args.number_of_workers
create_nodes_and_wait_till_registered(
inventory_client=client,
cluster=cluster,
nodes_details=nodes_details,
tf=tf,
is_ipv4=is_ipv4,
nodes_number=nodes_number,
cluster_deployment=cluster_deployment,
)
main_cidr = args.vm_network_cidr if is_ipv4 else args.vm_network_cidr6
secondary_cidr = machine_net.provisioning_cidr_v4 if is_ipv4 else machine_net.provisioning_cidr_v6
if client:
cluster_info = client.cluster_get(cluster.id)
macs = get_libvirt_nodes_macs(nodes_details["libvirt_network_name"])
if is_none_platform_mode():
macs += get_libvirt_nodes_macs(nodes_details["libvirt_secondary_network_name"])
if not (cluster_info.api_vip and cluster_info.ingress_vip):
if not args.kube_api:
wait_till_hosts_with_macs_are_in_status(
client=client,
cluster_id=cluster.id,
macs=macs,
statuses=[
consts.NodesStatus.INSUFFICIENT,
consts.NodesStatus.PENDING_FOR_INPUT,
consts.NodesStatus.KNOWN
],
)
if args.master_count == 1:
set_single_node_ip(
client=client,
cluster_id=cluster.id,
main_cidr=main_cidr,
is_ipv4=is_ipv4,
cluster_deployment=cluster_deployment,
tf=tf,
)
if not args.kube_api:
set_cluster_machine_cidr(
client=client,
cluster_id=cluster.id,
machine_net=machine_net,
set_vip_dhcp_allocation=False,
)
elif is_none_platform_mode():
pass
elif args.vip_dhcp_allocation and not args.kube_api:
set_cluster_machine_cidr(client, cluster.id, machine_net)
else:
set_cluster_vips(client, cluster.id, machine_net)
else:
log.info("VIPs already configured")
if args.kube_api:
kubeapi_utils.set_agents_hostnames(
cluster_deployment=cluster_deployment,
is_ipv4=is_ipv4,
static_network_mode=args.with_static_network_config,
tf=tf,
nodes_number=nodes_number,
)
else:
set_hosts_roles(
client=client,
cluster=cluster,
nodes_details=nodes_details,
machine_net=machine_net,
tf=tf,
master_count=args.master_count,
static_network_mode=args.with_static_network_config,
)
if is_none_platform_mode() and args.master_count > 1:
master_ips = helper_cluster.Cluster.get_master_ips(client, cluster.id,
main_cidr) + helper_cluster.Cluster.get_master_ips(
client, cluster.id, secondary_cidr)
worker_ips = helper_cluster.Cluster.get_worker_ips(client, cluster.id,
main_cidr) + helper_cluster.Cluster.get_worker_ips(
client, cluster.id, secondary_cidr)
if not worker_ips:
worker_ips = master_ips
load_balancer_ip = _get_host_ip_from_cidr(
machine_net.cidr_v6 if machine_net.has_ip_v6 and not machine_net.has_ip_v4 else machine_net.cidr_v4)
lb_controller = LoadBalancerController(tf)
lb_controller.set_load_balancing_config(load_balancer_ip, master_ips, worker_ips)
if not args.kube_api:
wait_till_hosts_with_macs_are_in_status(
client=client,
cluster_id=cluster.id,
macs=macs,
statuses=[consts.NodesStatus.KNOWN],
)
if args.vip_dhcp_allocation:
vips_info = helper_cluster.Cluster.get_vips_from_cluster(client, cluster.id)
tf.set_new_vips(api_vip=vips_info["api_vip"], ingress_vip=vips_info["ingress_vip"])
if args.install_cluster:
install_cluster.run_install_flow(
client=client,
cluster_id=cluster.id,
kubeconfig_path=utils.get_kubeconfig_path(cluster_name),
pull_secret=args.pull_secret,
tf=tf,
cluster_deployment=cluster_deployment,
agent_cluster_install=agent_cluster_install,
nodes_number=nodes_number,
)
# Validate DNS domains resolvability
validate_dns(client, cluster.id)