def get_location(workflow: str):
"""Extract host and port from a workflow's contact file.
NB: if it fails to load the workflow contact file, it will exit.
Args:
workflow (str): workflow name
Returns:
Tuple[str, int, int]: tuple with the host name and port numbers.
Raises:
ClientError: if the workflow is not running.
CylcVersionError: if target is a Cylc 7 (or earlier) workflow.
"""
try:
contact = load_contact_file(workflow)
except ServiceFileError:
raise WorkflowStopped(workflow)
host = contact[ContactFileFields.HOST]
host = get_fqdn_by_host(host)
port = int(contact[ContactFileFields.PORT])
if ContactFileFields.PUBLISH_PORT in contact:
pub_port = int(contact[ContactFileFields.PUBLISH_PORT])
else:
version = (contact['CYLC_VERSION']
if 'CYLC_VERSION' in contact else None)
raise CylcVersionError(version=version)
return host, port, pub_port
async def test_load_contact_file_async(myflow):
cont = await load_contact_file_async(myflow.workflow)
assert cont[CFF.HOST] == myflow.host
# compare the async interface to the sync interface
cont2 = load_contact_file(myflow.workflow)
assert cont == cont2
def _check_contact_file(scheduler):
try:
contact_data = workflow_files.load_contact_file(scheduler.workflow)
if contact_data != scheduler.contact_data:
raise CylcError('contact file modified')
except (AssertionError, IOError, ValueError, ServiceFileError):
raise CylcError('%s: contact file corrupted/modified and may be left' %
workflow_files.get_contact_file(scheduler.workflow))
def main(parser, options, reg):
"""CLI for "cylc get-workflow-contact"."""
try:
data = load_contact_file(reg)
except ServiceFileError:
raise CylcError(
f"{reg}: cannot get contact info, workflow not running?")
else:
for key, value in sorted(data.items()):
print("%s=%s" % (key, value))
async def test_scan_cleans_stuck_contact_files(
run,
scheduler,
flow,
one_conf,
run_dir,
test_dir,
):
"""Ensure scan tidies up contact files from crashed flows."""
# create a flow
reg = flow(one_conf, name='-crashed-')
schd = scheduler(reg)
srv_dir = Path(run_dir, reg, WorkflowFiles.Service.DIRNAME)
tmp_dir = test_dir / 'srv'
cont = srv_dir / WorkflowFiles.Service.CONTACT
# run the flow, copy the contact, stop the flow, copy back the contact
async with run(schd):
copytree(srv_dir, tmp_dir)
rmtree(srv_dir)
copytree(tmp_dir, srv_dir)
rmtree(tmp_dir)
# the old contact file check uses the CLI command that the flow was run
# with to check that whether the flow is running. Because this is an
# integration test the process is the pytest process and it is still
# running so we need to change the command so that Cylc sees the flow as
# having crashed
contact_info = load_contact_file(reg)
contact_info[ContactFileFields.COMMAND] += 'xyz'
dump_contact_file(reg, contact_info)
# make sure this flow shows for a regular filesystem-only scan
opts = ScanOptions(states='running,paused', format='name')
flows = []
await main(opts, write=flows.append, scan_dir=test_dir)
assert len(flows) == 1
assert '-crashed-' in flows[0]
# the contact file should still be there
assert cont.exists()
# make sure this flow shows for a regular filesystem-only scan
opts = ScanOptions(states='running,paused', format='name', ping=True)
flows = []
await main(opts, write=flows.append, scan_dir=test_dir)
assert len(flows) == 0
# the contact file should have been removed by the scan
assert not cont.exists()
def main(parser: COP, options: 'Values', workflow_id: str) -> None:
"""CLI for "cylc get-workflow-contact"."""
workflow_id, *_ = parse_id(
workflow_id,
constraint='workflows',
)
try:
data = load_contact_file(workflow_id)
except ServiceFileError:
raise CylcError(
f"{workflow_id}: cannot get contact info, workflow not running?")
else:
for key, value in sorted(data.items()):
print("%s=%s" % (key, value))
def send_request(self, command, args=None, timeout=None):
"""Send a request, using ssh.
Determines ssh_cmd, cylc_path and login_shell settings from the contact
file.
Converts message to JSON and sends this to stdin. Executes the Cylc
command, then deserialises the output.
Use ``__call__`` to call this method.
Args:
command (str): The name of the endpoint to call.
args (dict): Arguments to pass to the endpoint function.
timeout (float): Override the default timeout (seconds).
Raises:
ClientError: Coverall, on error from function call
Returns:
object: Deserialized output from function called.
"""
# Set environment variable to determine the communication for use on
# the scheduler
os.environ["CLIENT_COMMS_METH"] = CommsMeth.SSH.value
cmd = ["client"]
if timeout:
cmd += [f'comms_timeout={timeout}']
cmd += [self.workflow, command]
contact = load_contact_file(self.workflow)
ssh_cmd = contact[ContactFileFields.SCHEDULER_SSH_COMMAND]
login_shell = contact[ContactFileFields.SCHEDULER_USE_LOGIN_SHELL]
cylc_path = contact[ContactFileFields.SCHEDULER_CYLC_PATH]
cylc_path = None if cylc_path == 'None' else cylc_path
if not args:
args = {}
message = json.dumps(args)
proc = _remote_cylc_cmd(
cmd,
host=self.host,
stdin_str=message,
ssh_cmd=ssh_cmd,
remote_cylc_path=cylc_path,
ssh_login_shell=login_shell,
capture_process=True)
out, err = (f.decode() for f in proc.communicate())
return_code = proc.wait()
if return_code:
raise ClientError(err, f"return-code={return_code}")
return json.loads(out)
def prepare_command(self, command: str, args: Dict, timeout: Union[float,
str]):
"""Prepare command for submission.
"""
# Set environment variable to determine the communication for use on
# the scheduler
os.environ["CLIENT_COMMS_METH"] = CommsMeth.SSH.value
cmd = ["client"]
if timeout:
cmd += [f'--comms-timeout={timeout}']
cmd += [self.workflow, command]
contact = load_contact_file(self.workflow)
ssh_cmd = contact[ContactFileFields.SCHEDULER_SSH_COMMAND]
login_shell = contact[ContactFileFields.SCHEDULER_USE_LOGIN_SHELL]
cylc_path = contact[ContactFileFields.SCHEDULER_CYLC_PATH]
cylc_path = None if cylc_path == 'None' else cylc_path
if not args:
args = {}
message = json.dumps(args)
return cmd, ssh_cmd, login_shell, cylc_path, message
def _timeout_handler(workflow: str, host: str, port: Union[int, str]):
"""Handle the eventuality of a communication timeout with the workflow.
Args:
workflow (str): workflow name
host (str): host name
port (Union[int, str]): port number
Raises:
ClientError: if the workflow has already stopped.
"""
if workflow is None:
return
try:
contact_data: Dict[str, str] = load_contact_file(workflow)
except (IOError, ValueError, ServiceFileError):
# Contact file does not exist or corrupted, workflow should be dead
return
contact_host: str = contact_data.get(ContactFileFields.HOST, '?')
contact_port: str = contact_data.get(ContactFileFields.PORT, '?')
if (
contact_host != host
or contact_port != str(port)
):
raise CylcError(
f'The workflow is no longer running at {host}:{port}\n'
f'It has moved to {contact_host}:{contact_port}'
)
# Cannot connect, perhaps workflow is no longer running and is leaving
# behind a contact file?
try:
detect_old_contact_file(workflow, contact_data)
except (AssertionError, ServiceFileError):
# old contact file exists and the workflow process still alive
return
else:
# the workflow has stopped
raise WorkflowStopped(workflow)
def get_location(workflow: str):
"""Extract host and port from a workflow's contact file.
NB: if it fails to load the workflow contact file, it will exit.
Args:
workflow (str): workflow name
Returns:
Tuple[str, int, int]: tuple with the host name and port numbers.
Raises:
ClientError: if the workflow is not running.
"""
try:
contact = load_contact_file(workflow)
except ServiceFileError:
raise WorkflowStopped(workflow)
host = contact[ContactFileFields.HOST]
host = get_fqdn_by_host(host)
port = int(contact[ContactFileFields.PORT])
pub_port = int(contact[ContactFileFields.PUBLISH_PORT])
return host, port, pub_port
def test_load_contact_file(myflow):
cont = load_contact_file(myflow.workflow)
assert cont[CFF.HOST] == myflow.host