def fix_split(name):
assert os.path.exists(name) and os.path.isdir(name)
queue = QueueDir(name)
assert queue.qsize() == 1
item_list = queue.get()
queue.extend(item_list)
logger.info(queue)
def _queue_jds(name):
assert os.path.exists(name) and os.path.isdir(name)
queue = QueueDir(name)
jds = {}
for i in range(queue.qsize()):
jd = queue.peek(i)
jds[i] = jd
return jds
def _queue_ids(name):
assert os.path.exists(name) and os.path.isdir(name)
queue = QueueDir(name)
ids = []
for i in range(queue.qsize()):
jd = queue.peek(i)
job_id = jd['job_id']
ids.append(job_id)
return ids
def fill(dst, template, count):
queue = QueueDir(dst)
with open(template) as fh:
jd_template = json.load(fh)
jd_min = None
if 'job_id' in jd_template:
jd_min = jd_template['job_id']
assert count > 0
for i in range(count):
jd = copy(jd_template)
if jd_min is not None:
jd['job_id'] = jd_min + i
queue.put(jd)
def reset_fail(name):
assert os.path.exists(name) and os.path.isdir(name)
name = name.rstrip('/')
assert name.endswith('.fail')
origname = name.replace('.fail', '')
groupname = os.path.basename(origname)
qfail = QueueDir(name)
qorig = QueueDir(origname)
for jd in qfail:
outdir = 'output-%s/%d' % (groupname, jd['job_id'])
if os.path.exists(outdir):
shutil.rmtree(outdir)
qorig.put(jd)
def check_dupes(name, do_remove=False):
assert os.path.exists(name) and os.path.isdir(name)
queue = QueueDir(name)
queue_files = queue.list_files()
jds = {}
for i in range(queue.qsize()):
jd = queue.peek(i)
key = jd['job_id']
jd_rec = {'file': queue_files[i], 'jd': jd, 'id': i}
if key in (jds):
jds[key].append(jd_rec)
else:
jds[key] = [jd_rec]
for key, dupes in jds.iteritems():
if len(dupes) > 1:
print "Dupes: %s" % dupes
if do_remove:
for jd_rec in dupes[0:-1]:
print "remove: %s" % jd_rec['file']
os.remove(jd_rec['file']) # hack
def stat_host(basedir, name, exptotal=None):
stat = {WORK: 0}
total_count = 0
for key, suff in queue_exts.iteritems():
queue_dir = "%s/%s%s" % (basedir, name, suff)
if os.path.exists(queue_dir):
q = QueueDir(queue_dir)
stat[key] = q.qsize()
total_count += stat[key]
else:
stat[key] = 0
if os.path.exists("%s/%s.locker" % (basedir, name)):
with open("%s/%s.locker" % (basedir, name)) as fh:
queue_work = cPickle.load(fh)
stat[WORK] = len(queue_work)
total_count += len(queue_work)
stat[TOTAL] = total_count
update_calc_stat(stat, exptotal)
stat['name'] = name
return stat
def mv(src, dst, count=None, put_back=False):
assert src is not None and dst is not None
assert os.path.exists(src) and os.path.isdir(src)
q_src = QueueDir(src)
q_dst = QueueDir(dst, default_mask=q_src.mask)
if count is None:
count = q_src.qsize()
if count > 0:
jds = q_src.get_n(count)
q_dst.extend(jds)
if put_back:
q_src.extend(jds)
else:
logger.warn("WARN: empty source queue")
logger.info("SRC: %s" % q_src)
logger.info("DST: %s" % q_dst)
def unlock(name):
lockfile = "%s.locker" % name
assert os.path.exists(name) and os.path.isdir(name)
jds = _lock_ids(lockfile)
assert len(jds) > 0
queue_orig = QueueDir(name)
queue_succ = QueueDir(name + ".success")
for job_id, jd in jds.iteritems():
if _has_output(name, jd):
queue_succ.put(jd)
logger.info("%d -> success" % job_id)
else:
queue_orig.put(jd)
logger.info("%d -> orig" % job_id)
os.remove(lockfile)
def check_success(start_id, stop_id):
pool = multiprocessing.Pool(POOL_SIZE)
group_names = ["mc%02d" % i for i in range(1,21)]
print group_names
unsuccessful = pool.map(_find_no_output, group_names)
unsuccessful = [x for x in unsuccessful if len(x) > 0]
print unsuccessful
with open("no_output.dump", 'w') as fh:
cPickle.dump(unsuccessful, fh)
for unx in unsuccessful:
queue_succ_name = unx.keys()[0].split(':')[0]
queue_succ = QueueDir(queue_succ_name)
queue_fail = QueueDir(queue_succ_name.replace('success', 'fail'))
for key in sorted(unx.keys(), key=lambda x: int(x.split(':')[1]), reverse=True):
id = int(key.split(':')[1])
jd = unx[key]
print "%s -> fail (%d)" % (key, jd['job_id'])
queue_fail.put(jd)
queue_succ.remove(id)
def fix_interrupts(name):
assert os.path.exists(name) and os.path.isdir(name)
assert name.endswith('fail')
queue_fail = QueueDir(name)
queue_success = QueueDir(name.replace('fail', 'success'))
restore_count = 0
queue_fail_size = queue_fail.qsize()
fail_files = queue_fail.list_files()
success_cache = {}
for i in range(queue_success.qsize()):
jd = queue_success.peek(i)
key = jd['job_id']
jd_rec = {'jd': jd, 'id': i}
success_cache[key] = jd_rec
for i in range(queue_fail.qsize() - 1, -1, -1):
jd = queue_fail.peek(i)
if _has_output(name, jd):
if jd['job_id'] in success_cache:
print "WARN: already in success (%s)" % fail_files[i]
continue
print "seemsOK: %d" % jd['job_id']
restore_count += 1
queue_fail.remove(i)
jd['ex_status'] = jd['status']
jd['status'] = 'SUCCESS'
queue_success.put(jd)
print "restored %d JDs of %d" % (restore_count, queue_fail_size)
