def submit_federated(clusters, jobs, group, pool):
"""
Attempts to submit the provided jobs to each cluster in clusters, until a cluster
returns a "created" status code. If no cluster returns "created" status, throws.
"""
messages = ""
for cluster in clusters:
cluster_name = cluster['name']
cluster_url = cluster['url']
try:
print_info('Attempting to submit on %s cluster...' % terminal.bold(cluster_name))
json_body = {'jobs': jobs}
if group:
json_body['groups'] = [group]
if pool:
json_body['pool'] = pool
resp = http.post(cluster, 'jobs', json_body)
print_submit_result(cluster, resp)
if resp.status_code == 201:
metrics.inc('command.submit.jobs', len(jobs))
return 0
except requests.exceptions.ReadTimeout as rt:
logging.exception(rt)
print_info(terminal.failed(
f'Encountered read timeout with {cluster_name} ({cluster_url}). Your submission may have completed.'))
return 1
except IOError as ioe:
logging.exception(ioe)
reason = f'Cannot connect to {cluster_name} ({cluster_url})'
message = submit_failed_message(cluster_name, reason)
messages += message
print_error(messages)
raise Exception(terminal.failed('Job submission failed on all of your configured clusters.'))
def run(args):
"""
Main entrypoint to the cook scheduler CLI. Loads configuration files,
processes global command line arguments, and calls other command line
sub-commands (actions) if necessary.
"""
args = vars(parser.parse_args(args))
print_version = args.pop('version')
if print_version:
print(f'cs version {version.VERSION}')
return 0
util.silent = args.pop('silent')
verbose = args.pop('verbose') and not util.silent
log_format = '%(asctime)s [%(levelname)s] [%(name)s] %(message)s'
if verbose:
logging.getLogger('').handlers = []
logging.basicConfig(format=log_format, level=logging.DEBUG)
else:
logging.disable(logging.FATAL)
logging.debug('args: %s' % args)
action = args.pop('action')
config_path = args.pop('config')
cluster = args.pop('cluster')
url = args.pop('url')
if action is None:
parser.print_help()
else:
config_map = configuration.load_config_with_defaults(config_path)
try:
metrics.initialize(config_map)
metrics.inc('command.%s.runs' % action)
clusters = load_target_clusters(config_map, url, cluster)
http.configure(config_map)
args = {k: v for k, v in args.items() if v is not None}
defaults = config_map.get('defaults')
action_defaults = (defaults.get(action)
if defaults else None) or {}
result = actions[action](clusters,
deep_merge(action_defaults,
args), config_path)
logging.debug('result: %s' % result)
return result
finally:
metrics.close()
return None
def submit_federated(clusters, jobs):
"""
Attempts to submit the provided jobs to each cluster in clusters, until a cluster
returns a "created" status code. If no cluster returns "created" status, throws.
"""
for cluster in clusters:
cluster_name = cluster['name']
try:
print_info('Attempting to submit on %s cluster...' %
colors.bold(cluster_name))
resp = http.post(cluster, 'rawscheduler', {'jobs': jobs})
print_submit_result(cluster, resp)
if resp.status_code == 201:
metrics.inc('command.submit.jobs', len(jobs))
return 0
except IOError as ioe:
logging.info(ioe)
reason = 'Cannot connect to %s (%s)' % (cluster_name,
cluster['url'])
print_info('%s\n' % submit_failed_message(cluster_name, reason))
raise Exception(
colors.failed(
'Job submission failed on all of your configured clusters.'))
def run(args, plugins):
"""
Main entrypoint to the cook scheduler CLI. Loads configuration files,
processes global command line arguments, and calls other command line
sub-commands (actions) if necessary.
plugins is a map from plugin-name -> function or Class.SubCommandPlugin
"""
# This has to happen before we parse the args, otherwise we might
# get subcommand not found.
for name, instance in plugins.items():
if isinstance(instance, SubCommandPlugin):
logging.debug('Adding SubCommandPlugin %s' % name)
try:
instance.register(subparsers.add_parser,
configuration.add_defaults)
logging.debug('Done adding SubCommandPlugin %s' % name)
name = instance.name()
if name in actions:
raise Exception(
'SubCommandPlugin %s clashes with an existing subcommand.'
% name)
actions[name] = instance.run
except Exception as e:
print('Failed to load SubCommandPlugin %s: %s' % (name, e),
file=sys.stderr)
args = vars(parser.parse_args(args))
util.silent = args.pop('silent')
verbose = args.pop('verbose') and not util.silent
log_format = '%(asctime)s [%(levelname)s] [%(name)s] %(message)s'
if verbose:
logging.getLogger('').handlers = []
logging.basicConfig(format=log_format, level=logging.DEBUG)
else:
logging.disable(logging.FATAL)
logging.debug('args: %s', args)
action = args.pop('action')
config_path = args.pop('config')
cluster = args.pop('cluster')
url = args.pop('url')
if action is None:
parser.print_help()
else:
_, config_map = configuration.load_config_with_defaults(config_path)
try:
metrics.initialize(config_map)
metrics.inc('command.%s.runs' % action)
clusters = load_target_clusters(config_map, url, cluster)
http.configure(config_map, plugins)
cook.plugins.configure(plugins)
args = {k: v for k, v in args.items() if v is not None}
defaults = config_map.get('defaults')
action_defaults = (defaults.get(action)
if defaults else None) or {}
logging.debug('going to execute % action' % action)
result = actions[action](clusters,
deep_merge(action_defaults,
args), config_path)
logging.debug('result: %s' % result)
return result
finally:
metrics.close()
return None
