def read_jobs(self):
self.jobs = {}
execute = excute_local if not self._ssh else execute_remote
handle = execute('%s/squeue -a -h -t all -o \'JobId=%%i TimeUsed=%%M Partition=%%P JobState=%%T '
'ReqNodes=%%D ReqCPUs=%%C TimeLimit=%%l Name=%%j NodeList=%%N\'' % (self._path))
if handle.returncode:
raise ArcError('squeue error: %s' % '\n'.join(handle.stderr), 'SLURMInfo')
for line in handle.stdout:
try:
job = dict(item.split('=', 1) for item in LRMSInfo.split(line.strip()))
if job.has_key('TimeUsed'):
job['TimeUsed'] = SLURMInfo.as_period(job['TimeUsed'])
if job.has_key('TimeLimit'):
job['TimeLimit'] = SLURMInfo.as_period(job['TimeLimit'])
self.jobs[job['JobId']] = job
except ValueError: # Couldn't split: blank line, header etc ..
continue
def read_nodes(self):
self.nodes = {}
execute = excute_local if not self._ssh else execute_remote
handle = execute('%s/scontrol show node --oneliner' % (self._path))
if handle.returncode:
raise ArcError('scontrol error: %s' % '\n'.join(handle.stderr), 'SLURMInfo')
for line in handle.stdout:
try:
_ = dict(item.split('=', 1) for item in LRMSInfo.split(line.strip()))
record = dict((k, _[k]) for k in ('NodeName', 'CPUTot', 'RealMemory', 'State', 'Sockets', 'OS', 'Arch'))
# Node status can be followed by different symbols
# according to it being unresponsive, powersaving, etc.
# Get rid of them
record['State'] = record['State'].rstrip('*~#+')
self.nodes[record['NodeName']] = record
except KeyError: # Node is probably down if attributes are missing, just skip it
continue
except ValueError: # Couldn't split: blank line, header etc ..
continue
def read_partitions(self):
self.partitions = {}
execute = excute_local if not self._ssh else execute_remote
handle = execute('%s/sinfo -a -h -o \'PartitionName=%%P TotalCPUs=%%C '
'TotalNodes=%%D MaxTime=%%l\'' % (self._path))
if handle.returncode:
raise ArcError('sinfo error: %s' % '\n'.join(handle.stderr), 'SLURMInfo')
for line in handle.stdout:
try:
part = dict(item.split('=', 1) for item in LRMSInfo.split(line.strip()))
part['PartitionName'] = part['PartitionName'].rstrip('*')
part['MaxTime'] = SLURMInfo.as_period(part['MaxTime'])
# Format of '%C' is: Number of CPUs by state in the format 'allocated/idle/other/total'
part['AllocatedCPUs'], part['IdleCPUs'], part['OtherCPUs'], part['TotalCPUs'] = \
map(SLURMInfo.parse_number, part['TotalCPUs'].split('/'))
part['TotalNodes'] = SLURMInfo.parse_number(part['TotalNodes'])
self.partitions[part['PartitionName']] = part;
except ValueError: # Couldn't split: blank line, header etc ..
continue
def get_lrms_options_schema():
return LRMSInfo.get_lrms_options_schema(slurm_bin_path = '*')
def get_lrms_options_schema():
return LRMSInfo.get_lrms_options_schema()