|
1
|
|
|
|
|
2
|
|
|
import os |
|
3
|
|
|
import re |
|
4
|
|
|
import math |
|
5
|
|
|
from string import Template |
|
|
|
|
|
|
6
|
|
|
from datetime import timedelta |
|
7
|
|
|
from subprocess import Popen, PIPE, check_output |
|
|
|
|
|
|
8
|
|
|
from mycluster.mycluster import get_data |
|
9
|
|
|
from mycluster.mycluster import load_template |
|
10
|
|
|
|
|
11
|
|
|
""" |
|
12
|
|
|
|
|
13
|
|
|
bjobs -u all -q emerald |
|
14
|
|
|
bqueues -l emerald |
|
15
|
|
|
|
|
16
|
|
|
""" |
|
|
|
|
|
|
17
|
|
|
|
|
18
|
|
|
|
|
19
|
|
|
def scheduler_type(): |
|
20
|
|
|
return 'lsf' |
|
21
|
|
|
|
|
22
|
|
|
def name(): |
|
23
|
|
|
lsid_output = check_output(['lsid']).splitlines() |
|
24
|
|
|
for line in lsid_output: |
|
25
|
|
|
if line.startswith('My cluster name is'): |
|
26
|
|
|
return line.rsplit(' ',1)[1].strip() |
|
|
|
|
|
|
27
|
|
|
|
|
28
|
|
|
return 'undefined' |
|
29
|
|
|
|
|
30
|
|
|
def queues(): |
|
31
|
|
|
queue_list = [] |
|
32
|
|
|
|
|
33
|
|
|
with os.popen('bqueues -w -u `whoami`') as f: |
|
34
|
|
|
f.readline() # read header |
|
35
|
|
|
for line in f: |
|
36
|
|
|
q = line.split(' ')[0].strip() |
|
37
|
|
|
queue_list.append(q) |
|
38
|
|
|
|
|
39
|
|
|
return queue_list |
|
40
|
|
|
|
|
41
|
|
|
|
|
42
|
|
|
def accounts(): |
|
43
|
|
|
return [] |
|
44
|
|
|
|
|
45
|
|
|
|
|
46
|
|
|
def available_tasks(queue_id): |
|
47
|
|
|
|
|
48
|
|
|
# split queue id into queue and parallel env |
|
49
|
|
|
# list free slots |
|
50
|
|
|
free_tasks = 0 |
|
51
|
|
|
max_tasks = 0 |
|
52
|
|
|
run_tasks = 0 |
|
53
|
|
|
queue_name = queue_id |
|
|
|
|
|
|
54
|
|
|
q_output = check_output(['bqueues',queue_name]).splitlines() |
|
|
|
|
|
|
55
|
|
|
for line in q_output: |
|
56
|
|
|
if line.startswith(queue_name): |
|
57
|
|
|
new_line = re.sub(' +',' ',line).strip() |
|
|
|
|
|
|
58
|
|
|
try: |
|
59
|
|
|
max_tasks = int(new_line.split(' ')[4]) |
|
60
|
|
|
except: |
|
|
|
|
|
|
61
|
|
|
pass |
|
62
|
|
|
pen_tasks = int(new_line.split(' ')[8]) |
|
|
|
|
|
|
63
|
|
|
run_tasks = int(new_line.split(' ')[9]) |
|
|
|
|
|
|
64
|
|
|
sus_tasks = int(new_line.split(' ')[10]) |
|
|
|
|
|
|
65
|
|
|
|
|
66
|
|
|
|
|
67
|
|
|
return {'available' : max_tasks-run_tasks, 'max tasks' : max_tasks} |
|
68
|
|
|
|
|
69
|
|
|
def tasks_per_node(queue_id): |
|
70
|
|
|
host_list = None |
|
71
|
|
|
q_output = check_output(['bqueues','-l',queue_id]).splitlines() |
|
|
|
|
|
|
72
|
|
|
for line in q_output: |
|
73
|
|
|
if line.startswith('HOSTS:'): |
|
74
|
|
|
host_list = line.strip().rsplit(' ',1)[1].replace('/','') |
|
|
|
|
|
|
75
|
|
|
if host_list == 'none': |
|
76
|
|
|
return 0 |
|
77
|
|
|
bhosts_output = check_output(['bhosts','-l',host_list]).splitlines() |
|
|
|
|
|
|
78
|
|
|
line = re.sub(' +',' ',bhosts_output[2]).strip() |
|
|
|
|
|
|
79
|
|
|
tasks = int(line.split(' ')[3]) |
|
80
|
|
|
|
|
81
|
|
|
return tasks |
|
82
|
|
|
|
|
83
|
|
|
def node_config(queue_id): |
|
84
|
|
|
# Find first node with queue and record node config |
|
85
|
|
|
#bqueues -l queue_id |
|
86
|
|
|
host_list = None |
|
87
|
|
|
config = {} |
|
88
|
|
|
q_output = check_output(['bqueues','-l',queue_id]).splitlines() |
|
|
|
|
|
|
89
|
|
|
for line in q_output: |
|
90
|
|
|
if line.startswith('HOSTS:'): |
|
91
|
|
|
host_list = line.strip().rsplit(' ',1)[1].replace('/','') |
|
|
|
|
|
|
92
|
|
|
if host_list == 'none': |
|
93
|
|
|
config['max task'] = 0 |
|
|
|
|
|
|
94
|
|
|
config['max thread'] = 0 |
|
95
|
|
|
config['max memory'] = 0 |
|
96
|
|
|
return config |
|
97
|
|
|
bhosts_output = check_output(['bhosts','-l',host_list]).splitlines() |
|
|
|
|
|
|
98
|
|
|
line = re.sub(' +',' ',bhosts_output[2]).strip() |
|
|
|
|
|
|
99
|
|
|
tasks = int(line.split(' ')[3]) |
|
100
|
|
|
line = re.sub(' +',' ',bhosts_output[6]).strip() |
|
|
|
|
|
|
101
|
|
|
memory = float(line.split(' ')[11].replace('G','')) |
|
|
|
|
|
|
102
|
|
|
config['max task'] = tasks |
|
|
|
|
|
|
103
|
|
|
config['max thread'] = tasks |
|
104
|
|
|
config['max memory'] = memory |
|
105
|
|
|
|
|
106
|
|
|
return config |
|
107
|
|
|
|
|
108
|
|
|
|
|
109
|
|
|
def create_submit(queue_id,**kwargs): |
|
|
|
|
|
|
110
|
|
|
queue_name = queue_id |
|
|
|
|
|
|
111
|
|
|
num_tasks = 1 |
|
112
|
|
|
if 'num_tasks' in kwargs: |
|
113
|
|
|
num_tasks = kwargs['num_tasks'] |
|
114
|
|
|
|
|
115
|
|
|
tpn = tasks_per_node(queue_id) |
|
116
|
|
|
queue_tpn = tpn |
|
117
|
|
|
if 'tasks_per_node' in kwargs: |
|
118
|
|
|
tpn = min(tpn,kwargs['tasks_per_node']) |
|
|
|
|
|
|
119
|
|
|
|
|
120
|
|
|
nc = node_config(queue_id) |
|
121
|
|
|
qc = available_tasks(queue_id) |
|
122
|
|
|
|
|
123
|
|
|
if qc['max tasks'] > 0: |
|
124
|
|
|
num_tasks = min(num_tasks,qc['max tasks']) |
|
|
|
|
|
|
125
|
|
|
|
|
126
|
|
|
num_threads_per_task = nc['max thread'] |
|
127
|
|
|
if 'num_threads_per_task' in kwargs: |
|
128
|
|
|
num_threads_per_task = kwargs['num_threads_per_task'] |
|
129
|
|
|
num_threads_per_task = min(num_threads_per_task,int(math.ceil(float(nc['max thread'])/float(tpn)))) |
|
|
|
|
|
|
130
|
|
|
|
|
131
|
|
|
my_name = kwargs.get('my_name', "myclusterjob") |
|
132
|
|
|
my_output = kwargs.get('my_output', "myclusterjob.out") |
|
133
|
|
|
my_script = kwargs.get('my_script', None) |
|
134
|
|
|
if 'mycluster-' in my_script: |
|
135
|
|
|
my_script = get_data(my_script) |
|
136
|
|
|
|
|
137
|
|
|
user_email = kwargs.get('user_email', None) |
|
138
|
|
|
project_name = kwargs.get('project_name', 'default') |
|
139
|
|
|
|
|
140
|
|
|
wall_clock = kwargs.get('wall_clock', '12:00') |
|
141
|
|
|
if ':' not in wall_clock: |
|
142
|
|
|
wall_clock = wall_clock + ':00' |
|
143
|
|
|
|
|
144
|
|
|
num_nodes = int(math.ceil(float(num_tasks)/float(tpn))) |
|
145
|
|
|
|
|
146
|
|
|
num_queue_slots = num_nodes*queue_tpn |
|
147
|
|
|
|
|
148
|
|
|
no_syscribe = kwargs.get('no_syscribe', False) |
|
149
|
|
|
|
|
150
|
|
|
record_job = not no_syscribe |
|
151
|
|
|
|
|
152
|
|
|
openmpi_args = kwargs.get('openmpi_args', "-bysocket -bind-to-socket") |
|
153
|
|
|
|
|
154
|
|
|
qos = kwargs.get('qos', None) |
|
155
|
|
|
|
|
156
|
|
|
template = load_template('lsf.jinja') |
|
157
|
|
|
|
|
158
|
|
|
script_str = template.render(my_name = my_name, |
|
|
|
|
|
|
159
|
|
|
my_script = my_script, |
|
|
|
|
|
|
160
|
|
|
my_output = my_output, |
|
|
|
|
|
|
161
|
|
|
user_email = user_email, |
|
|
|
|
|
|
162
|
|
|
queue_name = queue_name, |
|
|
|
|
|
|
163
|
|
|
num_queue_slots = num_queue_slots, |
|
|
|
|
|
|
164
|
|
|
num_tasks = num_tasks, |
|
|
|
|
|
|
165
|
|
|
tpn = tpn, |
|
|
|
|
|
|
166
|
|
|
num_threads_per_task = num_threads_per_task, |
|
|
|
|
|
|
167
|
|
|
num_nodes = num_nodes, |
|
|
|
|
|
|
168
|
|
|
project_name = project_name, |
|
|
|
|
|
|
169
|
|
|
wall_clock = wall_clock, |
|
|
|
|
|
|
170
|
|
|
record_job = record_job, |
|
|
|
|
|
|
171
|
|
|
openmpi_args = openmpi_args, |
|
|
|
|
|
|
172
|
|
|
qos = qos) |
|
|
|
|
|
|
173
|
|
|
|
|
174
|
|
|
return script_str |
|
175
|
|
|
|
|
176
|
|
|
|
|
177
|
|
|
def submit(script_name,immediate, depends_on = None, depends_on_always_run = False): |
|
|
|
|
|
|
178
|
|
|
job_id = None |
|
179
|
|
|
|
|
180
|
|
View Code Duplication |
if depends_on and depends_on_always_run: |
|
|
|
|
|
|
181
|
|
|
with os.popen('bsub -w "ended(%s)" < %s ' % (depends_on, script_name)) as f: |
|
|
|
|
|
|
182
|
|
|
output = f.readline() |
|
183
|
|
|
try: |
|
184
|
|
|
job_id = int(output.split(' ')[1].replace('<','').replace('>','')) |
|
|
|
|
|
|
185
|
|
|
except: |
|
|
|
|
|
|
186
|
|
|
print('Job submission failed: ' + output) |
|
187
|
|
|
elif depends_on is not None: |
|
188
|
|
|
with os.popen('bsub -w "done(%s)" < %s ' % (depends_on, script_name)) as f: |
|
|
|
|
|
|
189
|
|
|
output = f.readline() |
|
190
|
|
|
try: |
|
191
|
|
|
job_id = int(output.split(' ')[1].replace('<','').replace('>','')) |
|
|
|
|
|
|
192
|
|
|
except: |
|
|
|
|
|
|
193
|
|
|
print('Job submission failed: ' + output) |
|
194
|
|
|
else: |
|
195
|
|
|
with os.popen('bsub <'+script_name) as f: |
|
196
|
|
|
output = f.readline() |
|
197
|
|
|
try: |
|
198
|
|
|
job_id = int(output.split(' ')[1].replace('<','').replace('>','')) |
|
|
|
|
|
|
199
|
|
|
except: |
|
|
|
|
|
|
200
|
|
|
print('Job submission failed: ' + output) |
|
201
|
|
|
return job_id |
|
202
|
|
|
|
|
203
|
|
|
|
|
204
|
|
|
def delete(job_id): |
|
205
|
|
|
with os.popen('bkill '+job_id) as f: |
|
206
|
|
|
pass |
|
207
|
|
|
|
|
208
|
|
|
|
|
209
|
|
View Code Duplication |
def status(): |
|
|
|
|
|
|
210
|
|
|
status_dict = {} |
|
211
|
|
|
with os.popen('bjobs -w') as f: |
|
212
|
|
|
try: |
|
213
|
|
|
f.readline() # read header |
|
214
|
|
|
for line in f: |
|
215
|
|
|
new_line = re.sub(' +',' ',line.strip()) |
|
|
|
|
|
|
216
|
|
|
job_id = int(new_line.split(' ')[0]) |
|
217
|
|
|
state = new_line.split(' ')[2] |
|
218
|
|
|
if state == 'RUN': |
|
219
|
|
|
status_dict[job_id] = 'r' |
|
220
|
|
|
else: |
|
221
|
|
|
status_dict[job_id] = state |
|
222
|
|
|
except e: |
|
|
|
|
|
|
223
|
|
|
print(e) |
|
|
|
|
|
|
224
|
|
|
|
|
225
|
|
|
return status_dict |
|
226
|
|
|
|
|
227
|
|
|
|
|
228
|
|
|
def job_stats(job_id): |
|
229
|
|
|
stats_dict = {} |
|
230
|
|
|
with os.popen('bacct -l '+str(job_id)) as f: |
|
231
|
|
|
try: |
|
232
|
|
|
line = f.readline() |
|
233
|
|
|
new_line = re.sub(' +',' ',line.strip()) |
|
|
|
|
|
|
234
|
|
|
stats_dict['wallclock'] = new_line.split(' ')[0] |
|
235
|
|
|
stats_dict['cpu'] = new_line.split(' ')[1] |
|
236
|
|
|
stats_dict['queue'] = new_line.split(' ')[2] |
|
237
|
|
|
stats_dict['mem'] = '-' # float(new_line.split(' ')[4])*int(new_line.split(' ')[3]) |
|
|
|
|
|
|
238
|
|
|
except: |
|
|
|
|
|
|
239
|
|
|
print('LSF: Error reading job stats') |
|
240
|
|
|
|
|
241
|
|
|
return stats_dict |
|
242
|
|
|
|
|
243
|
|
|
|
|
244
|
|
|
def job_stats_enhanced(job_id): |
|
245
|
|
|
""" |
|
246
|
|
|
Get full job and step stats for job_id |
|
247
|
|
|
""" |
|
248
|
|
|
stats_dict = {} |
|
249
|
|
|
with os.popen('bjobs -o "jobid run_time cpu_used queue slots stat exit_code start_time estimated_start_time finish_time delimiter=\'|\'" -noheader '+str(job_id)) as f: |
|
|
|
|
|
|
250
|
|
|
try: |
|
251
|
|
|
line = f.readline() |
|
252
|
|
|
cols = line.split('|') |
|
253
|
|
|
stats_dict['job_id'] = cols[0] |
|
254
|
|
|
if cols[1] != '-': |
|
255
|
|
|
stats_dict['wallclock'] = timedelta(seconds=float(cols[1].split(' ')[0])) |
|
|
|
|
|
|
256
|
|
|
if cols[2] != '-': |
|
257
|
|
|
stats_dict['cpu'] = timedelta(seconds=float(cols[2].split(' ')[0])) |
|
|
|
|
|
|
258
|
|
|
stats_dict['queue'] = cols[3] |
|
259
|
|
|
stats_dict['status'] = cols[5] |
|
260
|
|
|
stats_dict['exit_code'] = cols[6] |
|
261
|
|
|
stats_dict['start'] = cols[7] |
|
262
|
|
|
stats_dict['start_time'] = cols[8] |
|
263
|
|
|
if stats_dict['status'] in ['DONE', 'EXIT']: |
|
264
|
|
|
stats_dict['end'] = cols[9] |
|
265
|
|
|
|
|
266
|
|
|
steps = [] |
|
267
|
|
|
stats_dict['steps'] = steps |
|
268
|
|
|
except: |
|
|
|
|
|
|
269
|
|
|
with os.popen('bhist -l '+str(job_id)) as f: |
|
270
|
|
|
try: |
|
271
|
|
|
output = f.readlines() |
|
272
|
|
|
for line in output: |
|
273
|
|
|
if "Done successfully" in line: |
|
274
|
|
|
stats_dict['status'] = 'DONE' |
|
275
|
|
|
return stats_dict |
|
276
|
|
|
elif "Completed <exit>" in line: |
|
277
|
|
|
stats_dict['status'] = 'EXIT' |
|
278
|
|
|
return stats_dict |
|
279
|
|
|
else: |
|
280
|
|
|
stats_dict['status'] = 'UNKNOWN' |
|
281
|
|
|
except Exception as e: |
|
282
|
|
|
print(e) |
|
283
|
|
|
print('LSF: Error reading job stats') |
|
284
|
|
|
stats_dict['status'] = 'UNKNOWN' |
|
285
|
|
|
return stats_dict |
|
286
|
|
|
|
|
287
|
|
|
|
|
288
|
|
|
def running_stats(job_id): |
|
289
|
|
|
stats_dict = {} |
|
290
|
|
|
with os.popen('bjobs -W '+str(job_id)) as f: |
|
291
|
|
|
try: |
|
292
|
|
|
line = f.readline() |
|
293
|
|
|
new_line = re.sub(' +',' ',line.strip()) |
|
|
|
|
|
|
294
|
|
|
stats_dict['wallclock'] = new_line.split(' ')[0] |
|
|
|
|
|
|
295
|
|
|
except: |
|
|
|
|
|
|
296
|
|
|
pass |
|
297
|
|
|
|
|
298
|
|
|
with os.popen('bjobs -W '+str(job_id)) as f: |
|
299
|
|
|
try: |
|
300
|
|
|
line = f.readline() |
|
301
|
|
|
new_line = re.sub(' +',' ',line.strip()) |
|
|
|
|
|
|
302
|
|
|
ntasks = int(new_line.split(' ')[2]) |
|
303
|
|
|
stats_dict['mem'] = float(new_line.split(' ')[1])*ntasks |
|
|
|
|
|
|
304
|
|
|
stats_dict['cpu'] = float(new_line.split(' ')[0])*ntasks |
|
|
|
|
|
|
305
|
|
|
except: |
|
|
|
|
|
|
306
|
|
|
pass |
|
307
|
|
|
|
|
308
|
|
|
return stats_dict |
|
309
|
|
|
|
zenotech /
MyCluster
Loading history...
