def get_job_table(self, state_filter, limit = 50, offset = 0,
cmd_filter = None):
limit += 1
data = []
state_filter = state_filter.split('|')
if state_filter:
for s in state_filter[:]:
if s != 'all' and s not in self.all_state:
state_filter.remove(s)
state_filter = tuple(state_filter)
if not state_filter:
state_filter = tuple([ 'fail', 'pending', 'running' ])
if 'all' in state_filter:
state_filter = tuple([ x for x in self.all_state ])
with db.connection(self):
fmt_strs = ', '.join(['%s'] * len(state_filter))
q = 'SELECT * FROM job WHERE job_state IN (' + fmt_strs + ')'
if cmd_filter:
q += 'AND job_jobname = %s'
args = state_filter + (cmd_filter,) + (limit, ) + (offset,)
else:
args = state_filter + (limit, ) + (offset,)
q+= ' ORDER BY job_id DESC LIMIT %s OFFSET %s'
#print >> sys.stderr, q % (state_filter + (limit, ) + (offset,))
self.cursor.execute(q, args)
data = self.cursor.fetchall()
has_next = True if len(data) == limit else False
return data[:limit-1], has_next
def add_request(self, jobname, run_cmd, args, max_vmem,
cpu_bound = True, force = False):
job_id = 0
with db.connection(self):
job_id = self._add_request(jobname, run_cmd, args,
max_vmem, cpu_bound, force)
return job_id
def get_job_table(self, state_filter, limit = 50, offset = 0,
cmd_filter = None):
limit += 1
data = []
state_filter = state_filter.split('|')
if state_filter:
for s in state_filter[:]:
if s != 'all' and s not in self.all_state:
state_filter.remove(s)
state_filter = tuple(state_filter)
if not state_filter:
state_filter = tuple([ 'fail', 'pending', 'running' ])
if 'all' in state_filter:
state_filter = tuple([ x for x in self.all_state ])
with db.connection(self):
fmt_strs = ', '.join(['%s'] * len(state_filter))
q = 'SELECT * FROM job WHERE job_state IN (' + fmt_strs + ')'
if cmd_filter:
q += 'AND job_jobname = %s'
args = state_filter + (cmd_filter,) + (limit, ) + (offset,)
else:
args = state_filter + (limit, ) + (offset,)
q+= ' ORDER BY job_id DESC LIMIT %s OFFSET %s'
#print >> sys.stderr, q % (state_filter + (limit, ) + (offset,))
self.cursor.execute(q, args)
data = self.cursor.fetchall()
has_next = True if len(data) == limit else False
return data[:limit-1], has_next
def add_request(self, jobname, run_cmd, args, max_vmem,
cpu_bound = True, force = False):
job_id = 0
with db.connection(self):
job_id = self._add_request(jobname, run_cmd, args,
max_vmem, cpu_bound, force)
return job_id
def update_accounting(self):
jobs = {}
with db.connection(self):
q = 'SELECT job_id, sge_jobnumber, sge_hostname FROM accounting WHERE sge_hostname=""'
self.cursor.execute(q)
for data in self.cursor.fetchall():
jobs[data['sge_jobnumber']] = [ data ]
if not len(jobs):
return
self.search_accounting(jobs)
with db.connection(self):
fields = [ 'hostname', 'qsub_time', 'start_time', 'end_time',
'failed', 'exit_status', 'ru_utime', 'ru_stime',
'ru_wallclock', 'used_maxvmem' ]
set_str = []
for f in fields:
set_str.append('sge_%s=%%(%s)s' % (f, f))
set_str = ', '.join(set_str)
for sge_jobnumber in jobs:
sge_jobnumber = int(sge_jobnumber)
# Accounting not found, it'll found in the next run.
if len(jobs[sge_jobnumber]) <= 1:
continue
q = "UPDATE accounting SET " + set_str
# We can't let execute() do the quoting for jobnumber, but
# sge_jobnumber is forced to int so this code is sql injection
# safe.
q += ' WHERE sge_jobnumber=%d' % sge_jobnumber
# Kludge, execute() don't accept a namedtuple nor an
# OrderedDict so convert it explicitly to a dict.
d = jobs[sge_jobnumber][1]._asdict()
d = dict(zip(d.keys(), d.values()))
self.cursor.execute(q, d)
job = jobs[sge_jobnumber][0]
new_state = 'success'
if int(d['failed']) or int(d['exit_status']):
new_state = 'fail'
q = 'UPDATE job SET job_state=%s WHERE job_id=%s'
self.cursor.execute(q, [ new_state, job['job_id'] ])
def pending_request(self, limit = 16, offset = 0):
data = []
with db.connection(self):
self.cursor.execute("SELECT * FROM job WHERE job_state='pending' LIMIT %s OFFSET %s",
[ limit, offset ])
data = self.cursor.fetchall()
return data
def pending_request(self, limit = 16, offset = 0):
data = []
with db.connection(self):
self.cursor.execute("SELECT * FROM job WHERE job_state='pending' LIMIT %s OFFSET %s",
[ limit, offset ])
data = self.cursor.fetchall()
return data
def update_accounting(self):
jobs = {}
with db.connection(self):
q = 'SELECT job_id, sge_jobnumber, sge_hostname FROM accounting WHERE sge_hostname=""'
self.cursor.execute(q)
for data in self.cursor.fetchall():
jobs[data['sge_jobnumber']] = [ data ]
if not len(jobs):
return
self.search_accounting(jobs)
with db.connection(self):
fields = [ 'hostname', 'qsub_time', 'start_time', 'end_time',
'failed', 'exit_status', 'ru_utime', 'ru_stime',
'ru_wallclock', 'used_maxvmem' ]
set_str = []
for f in fields:
set_str.append('sge_%s=%%(%s)s' % (f, f))
set_str = ', '.join(set_str)
for sge_jobnumber in jobs:
sge_jobnumber = int(sge_jobnumber)
# Accounting not found, it'll found in the next run.
if len(jobs[sge_jobnumber]) <= 1:
continue
q = "UPDATE accounting SET " + set_str
# We can't let execute() do the quoting for jobnumber, but
# sge_jobnumber is forced to int so this code is sql injection
# safe.
q += ' WHERE sge_jobnumber=%d' % sge_jobnumber
# Kludge, execute() don't accept a namedtuple nor an
# OrderedDict so convert it explicitly to a dict.
d = jobs[sge_jobnumber][1]._asdict()
d = dict(zip(d.keys(), d.values()))
self.cursor.execute(q, d)
job = jobs[sge_jobnumber][0]
new_state = 'success'
if int(d['failed']) or int(d['exit_status']):
new_state = 'fail'
q = 'UPDATE job SET job_state=%s WHERE job_id=%s'
self.cursor.execute(q, [ new_state, job['job_id'] ])
def update_db(lang, bookname):
import hocr_request
db_hocr = hocr_request.DbHocr()
with db.connection(db_hocr):
path = cache_path(bookname, lang)
if os.path.exists(path + 'sha1.sum'):
sha1 = read_sha1(path)
db_hocr.add_update_row(bookname, lang, sha1)
else:
print >> sys.stderr, "Can't locate sha1.sum", path
def check_running(self):
sge_running = qstat.running_jobs('')
if sge_running:
with db.connection(self):
q = 'SELECT job_id, sge_jobnumber, job_args FROM job WHERE job_state="running"'
self.cursor.execute(q)
for r in self.cursor.fetchall():
if not r['sge_jobnumber'] in sge_running:
self._exec_check(r)
return len(sge_running)
return None
def check_running(self):
sge_running = qstat.running_jobs('')
if sge_running:
with db.connection(self):
q = 'SELECT job_id, sge_jobnumber, job_args FROM job WHERE job_state="running"'
self.cursor.execute(q)
for r in self.cursor.fetchall():
if not r['sge_jobnumber'] in sge_running:
self._exec_check(r)
return len(sge_running)
return None
def update_db(lang, bookname):
import hocr_request
db_hocr = hocr_request.DbHocr()
with db.connection(db_hocr):
path = cache_path(bookname, lang)
if os.path.exists(path + 'sha1.sum'):
sha1 = read_sha1(path)
db_hocr.add_update_row(bookname, lang, sha1)
else:
print >> sys.stderr, "Can't locate sha1.sum", path
def exec_request(self, r):
sge_job_nr = 0
# This is a bit convoluted but we need it to avoid a race condition:
# we set the job as running before starting it so on if this script
# run twice in parallel we don't try to start the same job twice. Then
# when the job really started or fail to start we update its state
# again. As we don't know yet the sge job number, we setup it as zero.
# Note this could be done in pending_request() but I prefer to protect
# it locally.
really_pending = False
with db.connection(self):
q = 'UPDATE job SET job_state=%s, sge_jobnumber=%s WHERE job_id=%s AND job_state="pending"'
if self.cursor.execute(q, [ 'running', 0, r['job_id'] ]):
really_pending = True
if not really_pending:
print >> sys.stderr, "run request for job_id %s cancelled, as it's no longer pending" % r['job_id']
return
cmdline_arg = job_cmdline_arg(r, 'job_run_cmd')
sge_cmdline = sge_cmdline_arg(r)
ls = subprocess.Popen(sge_cmdline + cmdline_arg,
stdin=None, stdout=subprocess.PIPE,
close_fds = True)
text = ls.stdout.read()
ls.wait()
try:
sge_job_nr = int(re.search('Your job (\d+) ', text).group(1))
new_state = 'running'
except:
utils.print_traceback("sge failure to exec job: %d" % r['job_id'], text)
new_state = 'sge_fail'
# Now we can really update the job state, see comment above.
with db.connection(self):
q = 'UPDATE job SET job_state=%s, sge_jobnumber=%s WHERE job_id=%s'
self.cursor.execute(q, [ new_state, sge_job_nr, r['job_id'] ])
def exec_request(self, r):
sge_job_nr = 0
# This is a bit convoluted but we need it to avoid a race condition:
# we set the job as running before starting it so on if this script
# run twice in parallel we don't try to start the same job twice. Then
# when the job really started or fail to start we update its state
# again. As we don't know yet the sge job number, we setup it as zero.
# Note this could be done in pending_request() but I prefer to protect
# it locally.
really_pending = False
with db.connection(self):
q = 'UPDATE job SET job_state=%s, sge_jobnumber=%s WHERE job_id=%s AND job_state="pending"'
if self.cursor.execute(q, [ 'running', 0, r['job_id'] ]):
really_pending = True
if not really_pending:
print >> sys.stderr, "run request for job_id %s cancelled, as it's no longer pending" % r['job_id']
return
cmdline_arg = job_cmdline_arg(r, 'job_run_cmd')
sge_cmdline = sge_cmdline_arg(r)
ls = subprocess.Popen(sge_cmdline + cmdline_arg,
stdin=None, stdout=subprocess.PIPE,
close_fds = True)
text = ls.stdout.read()
ls.wait()
try:
sge_job_nr = int(re.search('Your job (\d+) ', text).group(1))
new_state = 'running'
except:
utils.print_traceback("sge failure to exec job: %d" % r['job_id'], text)
new_state = 'sge_fail'
# Now we can really update the job state, see comment above.
with db.connection(self):
q = 'UPDATE job SET job_state=%s, sge_jobnumber=%s WHERE job_id=%s'
self.cursor.execute(q, [ new_state, sge_job_nr, r['job_id'] ])
def delete(bookname, lang):
if type(bookname) == type(u''):
bookname = bookname.encode('utf-8')
bookname = bookname.replace(' ', '_')
path = hocr.cache_path(bookname, lang)
sha1 = hocr.read_sha1(path)
db_hocr = hocr_request.DbHocr()
with db.connection(db_hocr):
q = 'delete from hocr where sha1=%s and lang=%s and title=%s'
db_hocr.cursor.execute(q, [sha1, lang, bookname])
print db_hocr.cursor.fetchall()
if os.path.exists(path + 'sha1.sum'):
os.remove(path + 'sha1.sum')
def delete(bookname, lang):
if type(bookname) == type(u''):
bookname = bookname.encode('utf-8')
bookname = bookname.replace(' ', '_')
path = hocr.cache_path(bookname, lang)
sha1 = hocr.read_sha1(path)
db_hocr = hocr_request.DbHocr()
with db.connection(db_hocr):
q = 'delete from hocr where sha1=%s and lang=%s and title=%s'
db_hocr.cursor.execute(q, [sha1, lang, bookname])
print db_hocr.cursor.fetchall()
if os.path.exists(path + 'sha1.sum'):
os.remove(path + 'sha1.sum')
def get_accounting_table(self, limit = 50, offset = 0, job_ids = None):
limit += 1
data = []
if not job_ids:
job_ids = []
if type(job_ids) != type([]):
job_ids = [ job_ids ]
with db.connection(self):
q = 'SELECT * from accounting '
if job_ids:
fmt_strs = ', '.join(['%s'] * len(job_ids))
q += 'WHERE job_id in (' + fmt_strs + ') '
q += 'ORDER BY job_id DESC, sge_jobnumber DESC, sge_hostname LIMIT %s OFFSET %s'
self.cursor.execute(q, tuple(job_ids) + (limit, ) + (offset,))
data = self.cursor.fetchall()
has_next = True if len(data) == limit else False
return data[:limit-1], has_next
def get_accounting_table(self, limit = 50, offset = 0, job_ids = None):
limit += 1
data = []
if not job_ids:
job_ids = []
if type(job_ids) != type([]):
job_ids = [ job_ids ]
with db.connection(self):
q = 'SELECT * from accounting '
if job_ids:
fmt_strs = ', '.join(['%s'] * len(job_ids))
q += 'WHERE job_id in (' + fmt_strs + ') '
q += 'ORDER BY job_id DESC, sge_jobnumber DESC, sge_hostname LIMIT %s OFFSET %s'
self.cursor.execute(q, tuple(job_ids) + (limit, ) + (offset,))
data = self.cursor.fetchall()
has_next = True if len(data) == limit else False
return data[:limit-1], has_next
