def get_job_script(JobID):
"""Return the job payload script, or raise NotImplemented if unavailable."""
if FASRC:
#at FASRC, we have our slurmctld prolog store the script in a database
shv = ['mysql', 'slurm_jobscripts', '-BNr', '-e', 'select script from jobscripts where id_job = %d;' % int(JobID)]
return util.runsh(shv)
else:
raise NotImplementedError("job script retrieval requires implementation in config.py")
def get_job_script(JobID):
"""Return the job payload script, or raise NotImplemented if unavailable."""
if FASRC:
#at FASRC, we have our slurmctld prolog store the script in a database
shv = [
'mysql', 'slurm_jobscripts', '-BNr', '-e',
'select script from jobscripts where id_job = %d;' % int(JobID)
]
return util.runsh(shv)
else:
raise NotImplementedError(
"job script retrieval requires implementation in config.py")
def load_data(self, keys=[]):
for key in keys:
if not self.has_key(key):
#--- sacct data
#(not implemented yet)
#--- other one-off queries
if key == 'JobScript':
try:
self['JobScript'] = config.get_job_script(self['JobID'])
except NotImplementedError:
self['JobScript'] = None
elif key == 'SacctReport':
shv = ['sacct', '--format', _sacct_format_readable, '-j', self['JobID']]
self['SacctReport'] = util.runsh(shv).rstrip()
#--- derived
if key == 'ReqMem_bytes_total':
if self.has_key('ReqMem_bytes_per_core') and self['ReqMem_bytes_per_core'] is not None:
if self.has_key('ReqMem_bytes_per_node') and self['ReqMem_bytes_per_node'] is not None:
raise Exception("both memory-per-node and memory-per-core used, and this cannot handle that")
self['ReqMem_bytes_total'] = self['ReqMem_bytes_per_core'] * self['NCPUS']
elif self.has_key('ReqMem_bytes_per_node') and self['ReqMem_bytes_per_node'] is not None:
self['ReqMem_bytes_total'] = self['ReqMem_bytes_per_node'] * self['NNodes']
else:
self['ReqMem_bytes_total'] = None
elif key == 'CPU_Efficiency':
if self.has_key('TotalCPU') and \
self.has_key('CPUTime') and self['CPUTime']!=0:
self['CPU_Efficiency'] = float(self['TotalCPU'])/self['CPUTime']
else:
self['CPU_Efficiency'] = None
elif key == 'CPU_Wasted':
if self.has_key('TotalCPU') and self.has_key('CPUTime'):
self['CPU_Wasted'] = max(self['CPUTime'] - self['TotalCPU'], 0)
else:
self['CPU_Efficiency'] = None
elif key == 'JobScriptPreview':
try:
self['JobScriptPreview'] = slurmmon.job_script_preview(self['JobScript'], self)
except KeyError:
self['JobScriptPreview'] = None
def get_jobs_running_on_host(hostname):
"""Return jobs running on the given host."""
shv = ['squeue', '-o', '%A %u %C %D', '--noheader', '-w', hostname]
stdout = util.runsh(shv).strip()
for line in stdout.split('\n'):
line = line.strip()
if line!='':
try:
JobID, User, NCPUS, NNodes = line.split()
j = Job()
j['JobID'] = JobID
j['User'] = User
j['NCPUS'] = NCPUS
j['NNodes'] = NNodes
yield j
except Exception, e:
sys.stderr.write("*** ERROR *** unable to parse squeue job text [%r]: %s\n" % (line, e))
def get_jobs_running_on_host(hostname):
"""Return jobs running on the given host."""
shv = ['squeue', '-o', '%A %u %C %D', '--noheader', '-w', hostname]
stdout = util.runsh(shv).strip()
for line in stdout.split('\n'):
line = line.strip()
if line != '':
try:
JobID, User, NCPUS, NNodes = line.split()
j = Job()
j['JobID'] = JobID
j['User'] = User
j['NCPUS'] = NCPUS
j['NNodes'] = NNodes
yield j
except Exception, e:
sys.stderr.write(
"*** ERROR *** unable to parse squeue job text [%r]: %s\n"
% (line, e))
def test_shquote_runsh(self):
"""That echo is identity for a funky_string."""
self.assertEqual(
u.runsh('echo -n %s' % u.shquote(ShTestCase.funky_string)),
ShTestCase.funky_string
)
def test_runsh_list(self): """That runsh() works on an argv list.""" self.assertEqual( u.runsh(['echo','foo']), 'foo\n', )
def test_runsh_string(self):
"""That runsh() works on sh code as a string."""
self.assertEqual(
u.runsh('echo foo'),
'foo\n',
)
def load_data(self, keys=[]):
for key in keys:
if not self.has_key(key):
#--- sacct data
#(not implemented yet)
#--- other one-off queries
if key == 'JobScript':
try:
self['JobScript'] = config.get_job_script(
self['JobID'])
except NotImplementedError:
self['JobScript'] = None
elif key == 'SacctReport':
shv = [
'sacct', '--format', _sacct_format_readable, '-j',
self['JobID']
]
self['SacctReport'] = util.runsh(shv).rstrip()
#--- derived
if key == 'ReqMem_bytes_total':
if self.has_key('ReqMem_bytes_per_core') and self[
'ReqMem_bytes_per_core'] is not None:
if self.has_key('ReqMem_bytes_per_node') and self[
'ReqMem_bytes_per_node'] is not None:
raise Exception(
"both memory-per-node and memory-per-core used, and this cannot handle that"
)
self['ReqMem_bytes_total'] = self[
'ReqMem_bytes_per_core'] * self['NCPUS']
elif self.has_key('ReqMem_bytes_per_node') and self[
'ReqMem_bytes_per_node'] is not None:
self['ReqMem_bytes_total'] = self[
'ReqMem_bytes_per_node'] * self['NNodes']
else:
self['ReqMem_bytes_total'] = None
elif key == 'CPU_Efficiency':
if self.has_key('TotalCPU') and \
self.has_key('CPUTime') and self['CPUTime']!=0:
self['CPU_Efficiency'] = float(
self['TotalCPU']) / self['CPUTime']
else:
self['CPU_Efficiency'] = None
elif key == 'CPU_Wasted':
if self.has_key('TotalCPU') and self.has_key('CPUTime'):
self['CPU_Wasted'] = max(
self['CPUTime'] - self['TotalCPU'], 0)
else:
self['CPU_Efficiency'] = None
elif key == 'JobScriptPreview':
try:
self['JobScriptPreview'] = slurmmon.job_script_preview(
self['JobScript'], self)
except KeyError:
self['JobScriptPreview'] = None
