__init__.py 9.73 KB
Newer Older
1
2
3

from __future__ import print_function

4
5
6
7
8
9
10
import unittest
import nose

import os
import pkg_resources

import gridtk
11
12
import subprocess, signal
import time
13
14
15

from gridtk.models import Job

16
class GridTKTest(unittest.TestCase):
17
18
19
20
21
22
23
  # This class defines tests for the gridtk

  def setUp(self):
    # Create a temporary directory that will contain all outputs
    import tempfile
    self.temp_dir = tempfile.mkdtemp(prefix='gridtk_test')
    self.log_dir = os.path.join(self.temp_dir, 'logs')
24
25
    self.database = os.path.join(self.temp_dir, 'database.sql3')
    self.scheduler_job = None
26
27
28


  def tearDown(self):
29
30
    # make sure that all scheduler jobs are stopped after exiting
    if self.scheduler_job:
31
      self.scheduler_job.kill()
32
33
34
35
36
37
38
    # Clean up the mess that we created
    import shutil
    shutil.rmtree(self.temp_dir)

  def test01_local(self):
    # This test executes all commands of the local grid manager and asserts that everything is fine

39
40
41
42
43
    # first test, if the '/bin/bash' exists
    bash = '/bin/bash'
    if not os.path.exists(bash):
      raise SkipException("Could not find '%s' which is required to run the test scripts"%bash)

44
45
46
47
48
49
50
    try:

      # first, add some commands to the database
      script_1 = pkg_resources.resource_filename('gridtk.tests', 'test_script.sh')
      script_2 = pkg_resources.resource_filename('gridtk.tests', 'test_array.sh')
      from gridtk.script import jman
      # add a simple script that will write some information to the
51
52
      jman.main(['./bin/jman', '--local', '--database', self.database, 'submit', '--log-dir', self.log_dir, '--name', 'test_1', bash, script_1])
      jman.main(['./bin/jman', '--local', '--database', self.database, 'submit', '--log-dir', self.log_dir, '--name', 'test_2',  '--dependencies', '1', '--parametric', '1-7:2', bash, script_2])
53
54

      # check that the database was created successfully
55
      self.assertTrue(os.path.exists(self.database))
56
57
58
59
60
61
62
63
64
65

      print()
      # test that the list command works (should also work with the "default" grid manager
      jman.main(['./bin/jman', '--database', self.database, 'list', '--job-ids', '1'])
      jman.main(['./bin/jman', '--database', self.database, 'list', '--job-ids', '2', '--print-array-jobs', '--print-dependencies'])

      # get insight into the database
      job_manager = gridtk.local.JobManagerLocal(database=self.database)
      session = job_manager.lock()
      jobs = list(session.query(Job))
66
67
68
69
70
71
      self.assertEqual(len(jobs), 2)
      self.assertEqual(jobs[0].id, 1)
      self.assertEqual(jobs[1].id, 2)
      self.assertEqual(len(jobs[1].array), 4)
      self.assertEqual(jobs[0].status, 'submitted')
      self.assertEqual(jobs[1].status, 'submitted')
72
73
74

      # check that the job dependencies are correct
      waiting = jobs[0].get_jobs_waiting_for_us()
75
76
      self.assertEqual(len(waiting), 1)
      self.assertEqual(waiting[0].id, 2)
77
      waited = jobs[1].get_jobs_we_wait_for()
78
79
      self.assertEqual(len(waited), 1)
      self.assertEqual(waited[0].id, 1)
80
81
82
83
84
85
86
87
88
89
90

      job_manager.unlock()

      # now, start the local execution of the job in a parallel job
      self.scheduler_job = subprocess.Popen(['./bin/jman', '--local', '--database', self.database, 'run-scheduler', '--sleep-time', '5', '--parallel', '2'])

      # sleep some time to assure that the scheduler was able to start the first job
      time.sleep(4)
      # ... and kill the scheduler
      self.scheduler_job.kill()
      self.scheduler_job = None
91

92
93
94
      # now, the first job needs to have status failure, and the second needs to be queued
      session = job_manager.lock()
      jobs = list(session.query(Job))
95
96
97
      self.assertEqual(len(jobs), 2)
      self.assertEqual(jobs[0].status, 'failure')
      self.assertEqual(jobs[1].status, 'queued')
98
      # the result files should not be there yet
99
100
      self.assertFalse(os.path.exists(jobs[0].std_out_file()))
      self.assertFalse(os.path.exists(jobs[0].std_err_file()))
101
      job_manager.unlock()
102
103


104
105
      # reset the job 1
      jman.main(['./bin/jman', '--local', '--database', self.database, 'resubmit', '--job-id', '1', '--running-jobs'])
106

107
108
      # now, start the local execution of the job in a parallel job
      self.scheduler_job = subprocess.Popen(['./bin/jman', '--local', '--database', self.database, 'run-scheduler', '--sleep-time', '5', '--parallel', '2'])
109

110
111
112
113
114
115
116
117
118
      # sleep some time to assure that the scheduler was able to finish the first and start the second job
      time.sleep(9)
      # ... and kill the scheduler
      self.scheduler_job.kill()
      self.scheduler_job = None

      # Job 1 and two array jobs of job two should be finished now, the other two still need to be queued
      session = job_manager.lock()
      jobs = list(session.query(Job))
119
120
121
122
123
124
125
126
      self.assertEqual(len(jobs), 2)
      self.assertEqual(jobs[0].status, 'failure')
      self.assertEqual(jobs[1].status, 'executing')
      self.assertEqual(jobs[1].array[0].status, 'failure')
      self.assertEqual(jobs[1].array[0].result, 1)
      self.assertEqual(jobs[1].array[1].status, 'success')
      self.assertEqual(jobs[1].array[1].result, 0)
      self.assertEqual(len([a for a in jobs[1].array if a.status == 'queued']), 2)
127
128
129
130
131
      out_file = jobs[0].std_out_file()
      err_file = jobs[0].std_err_file()
      job_manager.unlock()

      # the result files of the first job should now be there
132
133
134
135
      self.assertTrue(os.path.isfile(out_file))
      self.assertTrue(os.path.isfile(err_file))
      self.assertEqual(open(out_file).read().rstrip(), 'This is a text message to std-out')
      self.assertEqual(open(err_file).read().rstrip(), 'This is a text message to std-err')
136
137
138
139

      # resubmit all jobs
      jman.main(['./bin/jman', '--local', '--database', self.database, 'resubmit', '--running-jobs'])
      # check that the log files have been cleaned
140
141
      self.assertFalse(os.path.exists(out_file))
      self.assertFalse(os.path.exists(err_file))
142
      # ... but the log dir still exists
143
      self.assertTrue(os.path.exists(self.log_dir))
144
145
146
147
148
149
150
151

      # now, let the scheduler run all jobs
      self.scheduler_job = subprocess.Popen(['./bin/jman', '--local', '--database', self.database, 'run-scheduler', '--sleep-time', '1', '--parallel', '2', '--die-when-finished'])
      # and wait for the job to finish (the timeout argument to Popen only exists from python 3.3 onwards)
      self.scheduler_job.wait()
      self.scheduler_job = None

      # check that all output files are generated again
152
153
154
155
      self.assertTrue(os.path.isfile(out_file))
      self.assertTrue(os.path.isfile(err_file))
      self.assertEqual(open(out_file).read().rstrip(), 'This is a text message to std-out')
      self.assertEqual(open(err_file).read().rstrip(), 'This is a text message to std-err')
156
157
158

      # check that exactly four output and four error files have been created
      files = os.listdir(self.log_dir)
159
      self.assertEqual(len(files), 10)
160
      for i in range(1,8,2):
161
162
        self.assertTrue('test_2.o2.%d'%i in files)
        self.assertTrue('test_2.e2.%d'%i in files)
163
164
165
166

      # check that all array jobs are finished now
      session = job_manager.lock()
      jobs = list(session.query(Job))
167
168
169
170
      self.assertEqual(len(jobs), 2)
      self.assertEqual(jobs[1].status, 'failure')
      self.assertEqual(jobs[1].array[0].status, 'failure')
      self.assertEqual(jobs[1].array[0].result, 1)
171
      for i in range(1,4):
172
173
        self.assertEqual(jobs[1].array[i].status, 'success')
        self.assertEqual(jobs[1].array[i].result, 0)
174
175
176
177
178
179
180
181
182
183
184
185
186
187
      job_manager.unlock()

      print()
      # test that the list command still works
      jman.main(['./bin/jman', '--database', self.database, 'list', '--print-array-jobs'])

      print()
      # test that the report command works
      jman.main(['./bin/jman', '--database', self.database, 'report'])

      # clean-up
      jman.main(['./bin/jman', '--local', '--database', self.database, 'delete'])

      # check that the database and the log files are gone
188
      self.assertEqual(len(os.listdir(self.temp_dir)), 0)
189
190

      # add the scripts again, but this time with the --stop-on-failure option
191
192
      jman.main(['./bin/jman', '--local', '--database', self.database, 'submit', '--log-dir', self.log_dir, '--name', 'test_1', '--stop-on-failure', bash, script_1])
      jman.main(['./bin/jman', '--local', '--database', self.database, 'submit', '--log-dir', self.log_dir, '--name', 'test_2',  '--dependencies', '1', '--parametric', '1-7:2', '--stop-on-failure', bash, script_2])
193
194
195
196
197
198
199
200

      # and execute them, but without writing the log files
      self.scheduler_job = subprocess.Popen(['./bin/jman', '--local', '--database', self.database, 'run-scheduler', '--sleep-time', '0.1', '--parallel', '2', '--die-when-finished', '--no-log-files'])
      # and wait for the job to finish (the timeout argument to Popen only exists from python 3.3 onwards)
      self.scheduler_job.wait()
      self.scheduler_job = None

      # assert that the log files are not there
201
202
      self.assertFalse(os.path.isfile(out_file))
      self.assertFalse(os.path.isfile(err_file))
203
204
205
206
207


      # check that all array jobs are finished now
      session = job_manager.lock()
      jobs = list(session.query(Job))
208
209
210
211
212
      self.assertEqual(len(jobs), 2)
      self.assertEqual(jobs[0].status, 'failure')
      self.assertEqual(jobs[0].result, 255)
      self.assertEqual(jobs[1].status, 'failure')
      self.assertTrue(jobs[1].result is None)
213
214
215
216
217
218
219
220
      job_manager.unlock()

      # and clean up again
      jman.main(['./bin/jman', '--local', '--database', self.database, 'delete'])

    except KeyboardInterrupt:
      # make sure that the keyboard interrupt is captured and the mess is cleaned up (i.e. by calling tearDown)
      pass
221

222
223
224
225
226

  def test02_grid(self):
    # Tests the functionality of the grid toolkit in the grid
    raise nose.plugins.skip.SkipTest("This test is not yet implemented. If you find a proper ways to test the grid functionality, please go ahead and implement the test.")