def job(args):
# Verify command syntax
if len(args) > 2:
raise errors.BadArgsError('job', 'job[ <job_id> | <auto>[ <app>]]')
# Basic checks
_establish_default_terminal()
_ensure_scheduler_exists('job')
# If job name is not provided then probably only one job is active
if len(args) == 0 or args[0] == 'auto':
try:
job_id = csssi.detect_active_job(default_terminal)
# No active jobs
except csssi.NoActiveJobError:
raise errors.CommandFailedError(
'job', 'No active user jobs found. Cannot proceed.')
# More than one job is active
except csssi.MoreThanOneActiveJobError:
raise errors.CommandFailedError(
'job', 'Found more than one job. Cannot proceed.')
# Job Id provided
else:
job_id = args[0]
# In case app name is provided
app = args[1] if len(args) == 2 else None
# Hosts and PIDs will be stored here
pid_dict = None
# List job nodes and processes
try:
pid_dict = csssi.ls_job_pids(default_terminal, job_id, app)
# Job does not exist
except csssi.InvalidJobError:
raise errors.CommandFailedError(
'job', '{0} does not seem to be a valid job id'.format(job_id))
# Cannot determine app
except csssi.NoRunningAppFoundError:
raise errors.CommandFailedError(
'job',
'No application name lookup pattern provided and automatic MPI application detection did not succeed. Ensure job {0} is actually running a distributed application and provide the application name'
.format(job_id))
# Launch GDB and attach to PIDs
session_manager, msgs = _attach_pids(pid_dict)
# Initialize the debugging session
return debug_session.init_debugging_mode(session_manager, msgs)
def _ensure_sessions_selected(cmd):
'''Verifies that there are sessions selected at this moment.'''
if not selected_sessions:
raise errors.CommandFailedError(
cmd,
'No session(s) selected. Debugging mode failed to start. (Maybe init_debugging_mode() was not called?)'
)
def attach(args):
def _find_dead_pids_host(host, pids):
dead_pids = []
_establish_default_terminal()
_ensure_scheduler_exists('jobs')
for pid in pids:
if not default_terminal.is_pid_alive(pid):
host_path = '.'.join(hops.list_hops())
if host:
host_path += '.' + host
dead_pids.append('{host}:{pid}'.format(host=host_path
or 'localhost',
pid=pid))
return dead_pids
def _find_dead_pids(pid_dict):
# Check the status of all provided PIDs
dead_pids = []
for host, pids in pid_dict.iteriterms():
# Establish a connection per each process
dead_pids.extend(_find_dead_pids_host(host, pids))
return dead_pids
def _parse_group_pids(expr):
pid_dict = {}
for app_instance in re.finditer(r'((?:(\w+):)?(\d+))', expr):
host = app_instance.group(2)
pid = int(app_instance.group(3))
if pid_dict.has_key(host):
pid_dict[host] += [pid]
else:
pid_dict[host] = [pid]
args_string = ' '.join(args)
# Verify command syntax
if len(args) < 1 or not re.match(r'(?:(?:\w+:)?\d+|\s)+', args_string):
raise errors.BadArgsError(
'attach', 'attach [<host>:]<pid>[ [<host>:]<pid> [...]]')
# Group by host
pid_dict = _parse_group_pids(args_string)
# Check the status of all provided PIDs
dead_pids = _find_dead_pids(pid_dict)
# Stop if all processes are alive
if len(dead_pids) != 0:
raise errors.CommandFailedError(
'attach',
'Invalid PIDs provided: {0}'.format(' ,'.join(dead_pids)))
# Launch GDB and attach to PIDs
session_manager, msgs = _attach_pids(pid_dict)
# Initialize the debugging session
return debug_session.init_debugging_mode(session_manager, msgs)
def _ensure_scheduler_exists(cmd):
if not default_terminal_scheduler:
try:
global default_terminal_scheduler
default_terminal_scheduler = csssi.detect_scheduler(
default_terminal)
except csssi.NoSchedulerFoundError:
if not cmd:
return False
raise errors.CommandFailedError(
cmd,
'Unable to detect the scheduling system type on this machine.')
return True
def run(self):
try:
self.term = console.Terminal(
hops.list_hops(),
target_host=self.host,
meta=self.pid,
tag=self.tag,
prompt_re=r'\(gdb\)\ \r\n',
exit_re=r'&"quit\n"|\^exit',
)
self.term.connect()
gdb_response = self.term.query(self.cmd)
try:
self.mi_response = mi_interface.parse(gdb_response)
except pexpect.ExceptionPexpect as e:
raise errors.CommandFailedError('attach', 'attach', e)
except Exception as e:
self.error = e
def unhop(args):
# Verify command syntax
if len(args) > 1:
raise errors.BadArgsError('unhop',
'unhop[ <number of hops to remove>].')
n_hops_to_remove = 1
if len(args) == 1:
try:
n_hops_to_remove = int(args[0])
except ValueError:
raise errors.BadArgsError('unhop',
'unhop[ <number of hops to remove>].')
try:
for _ in range(n_hops_to_remove):
hops.remove_last()
except console.NoHopsError:
raise errors.CommandFailedError(
'unhop', 'No more hops currently exist. Nothing can be removed')
_establish_default_terminal(reestablish=True)
return modes.offline, None