def ssh_copy_id(cls, # pylint: disable=too-many-arguments
instance: object,
name: str,
user: str,
key_file: str,
comment: str = "test@local") -> None:
container = cls.get_container(instance, name)
pub_key = pub_key_from_private_key_file(key_file)
shell_command = f"umask 077 && echo 'ssh-rsa {pub_key} {comment}' >> ~/.ssh/authorized_keys"
res = container.exec_run(["sh", "-c", shell_command], user=user)
if res.exit_code:
raise DockerException(f"{container}: {res.output.decode('utf-8')}")
def _create_instance(self, node_index, dc_idx, spot=False):
# if size of disk is larget than 80G, then
# change the timeout of job completion to default * 3.
gce_job_default_timeout = None
if self._gce_image_size and int(self._gce_image_size) > 80:
gce_job_default_timeout = self._gce_services[
dc_idx].connection.timeout
self._gce_services[
dc_idx].connection.timeout = gce_job_default_timeout * 3
self.log.info("Job complete timeout is set to %ss" %
self._gce_services[dc_idx].connection.timeout)
name = f"{self.node_prefix}-{dc_idx}-{node_index}".lower()
gce_disk_struct = self._get_disks_struct(name=name, dc_idx=dc_idx)
# Name must start with a lowercase letter followed by up to 63
# lowercase letters, numbers, or hyphens, and cannot end with a hyphen
assert len(name) <= 63, "Max length of instance name is 63"
startup_script = self.test_config.get_startup_script()
if self.params.get("scylla_linux_distro") in (
"ubuntu-bionic",
"ubuntu-xenial",
"ubuntu-focal",
):
# we need to disable sshguard to prevent blocking connections from the builder
startup_script += dedent("""
sudo systemctl disable sshguard
sudo systemctl stop sshguard
""")
username = self.params.get("gce_image_username")
public_key = pub_key_from_private_key_file(
self.params.get("user_credentials_path"))
create_node_params = dict(
name=name,
size=self._gce_instance_type,
image=self._gce_image,
ex_network=self._gce_network,
ex_disks_gce_struct=gce_disk_struct,
ex_metadata={
**self.tags,
"Name":
name,
"NodeIndex":
node_index,
"startup-script":
startup_script,
"user-data":
json.dumps(
dict(scylla_yaml=dict(cluster_name=self.name),
start_scylla_on_first_boot=False)),
"block-project-ssh-keys":
"true",
"ssh-keys":
f"{username}:ssh-rsa {public_key}",
},
ex_service_accounts=self._service_accounts,
ex_preemptible=spot)
instance = self._create_node_with_retries(
name=name,
dc_idx=dc_idx,
create_node_params=create_node_params,
spot=spot)
self.log.info('Created %s instance %s',
'spot' if spot else 'on-demand', instance)
if gce_job_default_timeout:
self.log.info('Restore default job timeout %s' %
gce_job_default_timeout)
self._gce_services[
dc_idx].connection.timeout = gce_job_default_timeout
return instance
def _create_instance(self, node_index, dc_idx, spot=False):
# if size of disk is larget than 80G, then
# change the timeout of job completion to default * 3.
gce_job_default_timeout = None
if self._gce_image_size and int(self._gce_image_size) > 80:
gce_job_default_timeout = self._gce_services[
dc_idx].connection.timeout
self._gce_services[
dc_idx].connection.timeout = gce_job_default_timeout * 3
self.log.info("Job complete timeout is set to %ss" %
self._gce_services[dc_idx].connection.timeout)
name = f"{self.node_prefix}-{dc_idx}-{node_index}".lower()
gce_disk_struct = list()
gce_disk_struct.append(
self._get_root_disk_struct(name=name,
disk_type=self._gce_image_type,
dc_idx=dc_idx))
for i in range(self._gce_n_local_ssd):
gce_disk_struct.append(
self._get_local_ssd_disk_struct(name=name,
index=i,
dc_idx=dc_idx))
if self._add_disks:
for disk_type, disk_size in self._add_disks.items():
disk_size = int(disk_size)
if disk_size:
gce_disk_struct.append(
self._get_persistent_disk_struct(name=name,
disk_size=disk_size,
disk_type=disk_type,
dc_idx=dc_idx))
self.log.info(gce_disk_struct)
# Name must start with a lowercase letter followed by up to 63
# lowercase letters, numbers, or hyphens, and cannot end with a hyphen
assert len(name) <= 63, "Max length of instance name is 63"
startup_script = cluster.Setup.get_startup_script()
if self.params.get("scylla_linux_distro") in (
"ubuntu-bionic",
"ubuntu-xenial",
"ubuntu-focal",
):
# we need to disable sshguard to prevent blocking connections from the builder
startup_script += dedent("""
sudo systemctl disable sshguard
sudo systemctl stop sshguard
""")
username = self.params.get("gce_image_username")
public_key = pub_key_from_private_key_file(
self.params.get("user_credentials_path"))
create_node_params = dict(
name=name,
size=self._gce_instance_type,
image=self._gce_image,
ex_network=self._gce_network,
ex_disks_gce_struct=gce_disk_struct,
ex_metadata={
**self.tags,
"Name":
name,
"NodeIndex":
node_index,
"startup-script":
startup_script,
"user-data":
json.dumps(
dict(scylla_yaml=dict(cluster_name=self.name),
start_scylla_on_first_boot=False)),
"block-project-ssh-keys":
"true",
"ssh-keys":
f"{username}:ssh-rsa {public_key}",
},
ex_service_accounts=self._service_accounts,
ex_preemptible=spot)
try:
instance = self._gce_services[dc_idx].create_node(
**create_node_params)
except GoogleBaseError as details:
if not spot:
raise
self.log.warning(
'Unable to create a spot instance, try to create an on-demand one: %s',
details)
# May happen that spot instance was preempted during instance creating. This instance won't be
# destroyed but stopped and stays with the name in the instances pool. In this case we should destroy this
# instance before creating new one
if "Instance failed to start due to preemption" in str(details) and \
(failed_instance := self._get_instances_by_name(dc_idx=dc_idx, name=name)):
self._gce_services[dc_idx].destroy_node(node=failed_instance,
destroy_boot_disk=True)
create_node_params['ex_preemptible'] = spot = False
instance = self._gce_services[dc_idx].create_node(
**create_node_params)