root / test / ganeti.jqueue_unittest.py @ 30c945d0
History | View | Annotate | Download (42.9 kB)
| 1 |
#!/usr/bin/python
|
|---|---|
| 2 |
#
|
| 3 |
|
| 4 |
# Copyright (C) 2010 Google Inc.
|
| 5 |
#
|
| 6 |
# This program is free software; you can redistribute it and/or modify
|
| 7 |
# it under the terms of the GNU General Public License as published by
|
| 8 |
# the Free Software Foundation; either version 2 of the License, or
|
| 9 |
# (at your option) any later version.
|
| 10 |
#
|
| 11 |
# This program is distributed in the hope that it will be useful, but
|
| 12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 14 |
# General Public License for more details.
|
| 15 |
#
|
| 16 |
# You should have received a copy of the GNU General Public License
|
| 17 |
# along with this program; if not, write to the Free Software
|
| 18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
| 19 |
# 02110-1301, USA.
|
| 20 |
|
| 21 |
|
| 22 |
"""Script for testing ganeti.jqueue"""
|
| 23 |
|
| 24 |
import os |
| 25 |
import sys |
| 26 |
import unittest |
| 27 |
import tempfile |
| 28 |
import shutil |
| 29 |
import errno |
| 30 |
import itertools |
| 31 |
|
| 32 |
from ganeti import constants |
| 33 |
from ganeti import utils |
| 34 |
from ganeti import errors |
| 35 |
from ganeti import jqueue |
| 36 |
from ganeti import opcodes |
| 37 |
from ganeti import compat |
| 38 |
from ganeti import mcpu |
| 39 |
|
| 40 |
import testutils |
| 41 |
|
| 42 |
|
| 43 |
class _FakeJob: |
| 44 |
def __init__(self, job_id, status): |
| 45 |
self.id = job_id
|
| 46 |
self._status = status
|
| 47 |
self._log = []
|
| 48 |
|
| 49 |
def SetStatus(self, status): |
| 50 |
self._status = status
|
| 51 |
|
| 52 |
def AddLogEntry(self, msg): |
| 53 |
self._log.append((len(self._log), msg)) |
| 54 |
|
| 55 |
def CalcStatus(self): |
| 56 |
return self._status |
| 57 |
|
| 58 |
def GetInfo(self, fields): |
| 59 |
result = [] |
| 60 |
|
| 61 |
for name in fields: |
| 62 |
if name == "status": |
| 63 |
result.append(self._status)
|
| 64 |
else:
|
| 65 |
raise Exception("Unknown field") |
| 66 |
|
| 67 |
return result
|
| 68 |
|
| 69 |
def GetLogEntries(self, newer_than): |
| 70 |
assert newer_than is None or newer_than >= 0 |
| 71 |
|
| 72 |
if newer_than is None: |
| 73 |
return self._log |
| 74 |
|
| 75 |
return self._log[newer_than:] |
| 76 |
|
| 77 |
|
| 78 |
class TestJobChangesChecker(unittest.TestCase): |
| 79 |
def testStatus(self): |
| 80 |
job = _FakeJob(9094, constants.JOB_STATUS_QUEUED)
|
| 81 |
checker = jqueue._JobChangesChecker(["status"], None, None) |
| 82 |
self.assertEqual(checker(job), ([constants.JOB_STATUS_QUEUED], []))
|
| 83 |
|
| 84 |
job.SetStatus(constants.JOB_STATUS_RUNNING) |
| 85 |
self.assertEqual(checker(job), ([constants.JOB_STATUS_RUNNING], []))
|
| 86 |
|
| 87 |
job.SetStatus(constants.JOB_STATUS_SUCCESS) |
| 88 |
self.assertEqual(checker(job), ([constants.JOB_STATUS_SUCCESS], []))
|
| 89 |
|
| 90 |
# job.id is used by checker
|
| 91 |
self.assertEqual(job.id, 9094) |
| 92 |
|
| 93 |
def testStatusWithPrev(self): |
| 94 |
job = _FakeJob(12807, constants.JOB_STATUS_QUEUED)
|
| 95 |
checker = jqueue._JobChangesChecker(["status"],
|
| 96 |
[constants.JOB_STATUS_QUEUED], None)
|
| 97 |
self.assert_(checker(job) is None) |
| 98 |
|
| 99 |
job.SetStatus(constants.JOB_STATUS_RUNNING) |
| 100 |
self.assertEqual(checker(job), ([constants.JOB_STATUS_RUNNING], []))
|
| 101 |
|
| 102 |
def testFinalStatus(self): |
| 103 |
for status in constants.JOBS_FINALIZED: |
| 104 |
job = _FakeJob(2178711, status)
|
| 105 |
checker = jqueue._JobChangesChecker(["status"], [status], None) |
| 106 |
# There won't be any changes in this status, hence it should signal
|
| 107 |
# a change immediately
|
| 108 |
self.assertEqual(checker(job), ([status], []))
|
| 109 |
|
| 110 |
def testLog(self): |
| 111 |
job = _FakeJob(9094, constants.JOB_STATUS_RUNNING)
|
| 112 |
checker = jqueue._JobChangesChecker(["status"], None, None) |
| 113 |
self.assertEqual(checker(job), ([constants.JOB_STATUS_RUNNING], []))
|
| 114 |
|
| 115 |
job.AddLogEntry("Hello World")
|
| 116 |
(job_info, log_entries) = checker(job) |
| 117 |
self.assertEqual(job_info, [constants.JOB_STATUS_RUNNING])
|
| 118 |
self.assertEqual(log_entries, [[0, "Hello World"]]) |
| 119 |
|
| 120 |
checker2 = jqueue._JobChangesChecker(["status"], job_info, len(log_entries)) |
| 121 |
self.assert_(checker2(job) is None) |
| 122 |
|
| 123 |
job.AddLogEntry("Foo Bar")
|
| 124 |
job.SetStatus(constants.JOB_STATUS_ERROR) |
| 125 |
|
| 126 |
(job_info, log_entries) = checker2(job) |
| 127 |
self.assertEqual(job_info, [constants.JOB_STATUS_ERROR])
|
| 128 |
self.assertEqual(log_entries, [[1, "Foo Bar"]]) |
| 129 |
|
| 130 |
checker3 = jqueue._JobChangesChecker(["status"], None, None) |
| 131 |
(job_info, log_entries) = checker3(job) |
| 132 |
self.assertEqual(job_info, [constants.JOB_STATUS_ERROR])
|
| 133 |
self.assertEqual(log_entries, [[0, "Hello World"], [1, "Foo Bar"]]) |
| 134 |
|
| 135 |
|
| 136 |
class TestJobChangesWaiter(unittest.TestCase): |
| 137 |
def setUp(self): |
| 138 |
self.tmpdir = tempfile.mkdtemp()
|
| 139 |
self.filename = utils.PathJoin(self.tmpdir, "job-1") |
| 140 |
utils.WriteFile(self.filename, data="") |
| 141 |
|
| 142 |
def tearDown(self): |
| 143 |
shutil.rmtree(self.tmpdir)
|
| 144 |
|
| 145 |
def _EnsureNotifierClosed(self, notifier): |
| 146 |
try:
|
| 147 |
os.fstat(notifier._fd) |
| 148 |
except EnvironmentError, err: |
| 149 |
self.assertEqual(err.errno, errno.EBADF)
|
| 150 |
else:
|
| 151 |
self.fail("File descriptor wasn't closed") |
| 152 |
|
| 153 |
def testClose(self): |
| 154 |
for wait in [False, True]: |
| 155 |
waiter = jqueue._JobFileChangesWaiter(self.filename)
|
| 156 |
try:
|
| 157 |
if wait:
|
| 158 |
waiter.Wait(0.001)
|
| 159 |
finally:
|
| 160 |
waiter.Close() |
| 161 |
|
| 162 |
# Ensure file descriptor was closed
|
| 163 |
self._EnsureNotifierClosed(waiter._notifier)
|
| 164 |
|
| 165 |
def testChangingFile(self): |
| 166 |
waiter = jqueue._JobFileChangesWaiter(self.filename)
|
| 167 |
try:
|
| 168 |
self.assertFalse(waiter.Wait(0.1)) |
| 169 |
utils.WriteFile(self.filename, data="changed") |
| 170 |
self.assert_(waiter.Wait(60)) |
| 171 |
finally:
|
| 172 |
waiter.Close() |
| 173 |
|
| 174 |
self._EnsureNotifierClosed(waiter._notifier)
|
| 175 |
|
| 176 |
def testChangingFile2(self): |
| 177 |
waiter = jqueue._JobChangesWaiter(self.filename)
|
| 178 |
try:
|
| 179 |
self.assertFalse(waiter._filewaiter)
|
| 180 |
self.assert_(waiter.Wait(0.1)) |
| 181 |
self.assert_(waiter._filewaiter)
|
| 182 |
|
| 183 |
# File waiter is now used, but there have been no changes
|
| 184 |
self.assertFalse(waiter.Wait(0.1)) |
| 185 |
utils.WriteFile(self.filename, data="changed") |
| 186 |
self.assert_(waiter.Wait(60)) |
| 187 |
finally:
|
| 188 |
waiter.Close() |
| 189 |
|
| 190 |
self._EnsureNotifierClosed(waiter._filewaiter._notifier)
|
| 191 |
|
| 192 |
|
| 193 |
class TestWaitForJobChangesHelper(unittest.TestCase): |
| 194 |
def setUp(self): |
| 195 |
self.tmpdir = tempfile.mkdtemp()
|
| 196 |
self.filename = utils.PathJoin(self.tmpdir, "job-2614226563") |
| 197 |
utils.WriteFile(self.filename, data="") |
| 198 |
|
| 199 |
def tearDown(self): |
| 200 |
shutil.rmtree(self.tmpdir)
|
| 201 |
|
| 202 |
def _LoadWaitingJob(self): |
| 203 |
return _FakeJob(2614226563, constants.JOB_STATUS_WAITLOCK) |
| 204 |
|
| 205 |
def _LoadLostJob(self): |
| 206 |
return None |
| 207 |
|
| 208 |
def testNoChanges(self): |
| 209 |
wfjc = jqueue._WaitForJobChangesHelper() |
| 210 |
|
| 211 |
# No change
|
| 212 |
self.assertEqual(wfjc(self.filename, self._LoadWaitingJob, ["status"], |
| 213 |
[constants.JOB_STATUS_WAITLOCK], None, 0.1), |
| 214 |
constants.JOB_NOTCHANGED) |
| 215 |
|
| 216 |
# No previous information
|
| 217 |
self.assertEqual(wfjc(self.filename, self._LoadWaitingJob, |
| 218 |
["status"], None, None, 1.0), |
| 219 |
([constants.JOB_STATUS_WAITLOCK], [])) |
| 220 |
|
| 221 |
def testLostJob(self): |
| 222 |
wfjc = jqueue._WaitForJobChangesHelper() |
| 223 |
self.assert_(wfjc(self.filename, self._LoadLostJob, |
| 224 |
["status"], None, None, 1.0) is None) |
| 225 |
|
| 226 |
|
| 227 |
class TestEncodeOpError(unittest.TestCase): |
| 228 |
def test(self): |
| 229 |
encerr = jqueue._EncodeOpError(errors.LockError("Test 1"))
|
| 230 |
self.assert_(isinstance(encerr, tuple)) |
| 231 |
self.assertRaises(errors.LockError, errors.MaybeRaise, encerr)
|
| 232 |
|
| 233 |
encerr = jqueue._EncodeOpError(errors.GenericError("Test 2"))
|
| 234 |
self.assert_(isinstance(encerr, tuple)) |
| 235 |
self.assertRaises(errors.GenericError, errors.MaybeRaise, encerr)
|
| 236 |
|
| 237 |
encerr = jqueue._EncodeOpError(NotImplementedError("Foo")) |
| 238 |
self.assert_(isinstance(encerr, tuple)) |
| 239 |
self.assertRaises(errors.OpExecError, errors.MaybeRaise, encerr)
|
| 240 |
|
| 241 |
encerr = jqueue._EncodeOpError("Hello World")
|
| 242 |
self.assert_(isinstance(encerr, tuple)) |
| 243 |
self.assertRaises(errors.OpExecError, errors.MaybeRaise, encerr)
|
| 244 |
|
| 245 |
|
| 246 |
class TestQueuedOpCode(unittest.TestCase): |
| 247 |
def testDefaults(self): |
| 248 |
def _Check(op): |
| 249 |
self.assertFalse(hasattr(op.input, "dry_run")) |
| 250 |
self.assertEqual(op.priority, constants.OP_PRIO_DEFAULT)
|
| 251 |
self.assertFalse(op.log)
|
| 252 |
self.assert_(op.start_timestamp is None) |
| 253 |
self.assert_(op.exec_timestamp is None) |
| 254 |
self.assert_(op.end_timestamp is None) |
| 255 |
self.assert_(op.result is None) |
| 256 |
self.assertEqual(op.status, constants.OP_STATUS_QUEUED)
|
| 257 |
|
| 258 |
op1 = jqueue._QueuedOpCode(opcodes.OpTestDelay()) |
| 259 |
_Check(op1) |
| 260 |
op2 = jqueue._QueuedOpCode.Restore(op1.Serialize()) |
| 261 |
_Check(op2) |
| 262 |
self.assertEqual(op1.Serialize(), op2.Serialize())
|
| 263 |
|
| 264 |
def testPriority(self): |
| 265 |
def _Check(op): |
| 266 |
assert constants.OP_PRIO_DEFAULT != constants.OP_PRIO_HIGH, \
|
| 267 |
"Default priority equals high priority; test can't work"
|
| 268 |
self.assertEqual(op.priority, constants.OP_PRIO_HIGH)
|
| 269 |
self.assertEqual(op.status, constants.OP_STATUS_QUEUED)
|
| 270 |
|
| 271 |
inpop = opcodes.OpGetTags(priority=constants.OP_PRIO_HIGH) |
| 272 |
op1 = jqueue._QueuedOpCode(inpop) |
| 273 |
_Check(op1) |
| 274 |
op2 = jqueue._QueuedOpCode.Restore(op1.Serialize()) |
| 275 |
_Check(op2) |
| 276 |
self.assertEqual(op1.Serialize(), op2.Serialize())
|
| 277 |
|
| 278 |
|
| 279 |
class TestQueuedJob(unittest.TestCase): |
| 280 |
def test(self): |
| 281 |
self.assertRaises(errors.GenericError, jqueue._QueuedJob,
|
| 282 |
None, 1, []) |
| 283 |
|
| 284 |
def testDefaults(self): |
| 285 |
job_id = 4260
|
| 286 |
ops = [ |
| 287 |
opcodes.OpGetTags(), |
| 288 |
opcodes.OpTestDelay(), |
| 289 |
] |
| 290 |
|
| 291 |
def _Check(job): |
| 292 |
self.assertEqual(job.id, job_id)
|
| 293 |
self.assertEqual(job.log_serial, 0) |
| 294 |
self.assert_(job.received_timestamp)
|
| 295 |
self.assert_(job.start_timestamp is None) |
| 296 |
self.assert_(job.end_timestamp is None) |
| 297 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 298 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT)
|
| 299 |
self.assert_(repr(job).startswith("<")) |
| 300 |
self.assertEqual(len(job.ops), len(ops)) |
| 301 |
self.assert_(compat.all(inp.__getstate__() == op.input.__getstate__()
|
| 302 |
for (inp, op) in zip(ops, job.ops))) |
| 303 |
self.assertRaises(errors.OpExecError, job.GetInfo,
|
| 304 |
["unknown-field"])
|
| 305 |
self.assertEqual(job.GetInfo(["summary"]), |
| 306 |
[[op.input.Summary() for op in job.ops]]) |
| 307 |
|
| 308 |
job1 = jqueue._QueuedJob(None, job_id, ops)
|
| 309 |
_Check(job1) |
| 310 |
job2 = jqueue._QueuedJob.Restore(None, job1.Serialize())
|
| 311 |
_Check(job2) |
| 312 |
self.assertEqual(job1.Serialize(), job2.Serialize())
|
| 313 |
|
| 314 |
def testPriority(self): |
| 315 |
job_id = 4283
|
| 316 |
ops = [ |
| 317 |
opcodes.OpGetTags(priority=constants.OP_PRIO_DEFAULT), |
| 318 |
opcodes.OpTestDelay(), |
| 319 |
] |
| 320 |
|
| 321 |
def _Check(job): |
| 322 |
self.assertEqual(job.id, job_id)
|
| 323 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 324 |
self.assert_(repr(job).startswith("<")) |
| 325 |
|
| 326 |
job = jqueue._QueuedJob(None, job_id, ops)
|
| 327 |
_Check(job) |
| 328 |
self.assert_(compat.all(op.priority == constants.OP_PRIO_DEFAULT
|
| 329 |
for op in job.ops)) |
| 330 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT)
|
| 331 |
|
| 332 |
# Increase first
|
| 333 |
job.ops[0].priority -= 1 |
| 334 |
_Check(job) |
| 335 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT - 1) |
| 336 |
|
| 337 |
# Mark opcode as finished
|
| 338 |
job.ops[0].status = constants.OP_STATUS_SUCCESS
|
| 339 |
_Check(job) |
| 340 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT)
|
| 341 |
|
| 342 |
# Increase second
|
| 343 |
job.ops[1].priority -= 10 |
| 344 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT - 10) |
| 345 |
|
| 346 |
# Test increasing first
|
| 347 |
job.ops[0].status = constants.OP_STATUS_RUNNING
|
| 348 |
job.ops[0].priority -= 19 |
| 349 |
self.assertEqual(job.CalcPriority(), constants.OP_PRIO_DEFAULT - 20) |
| 350 |
|
| 351 |
def testCalcStatus(self): |
| 352 |
def _Queued(ops): |
| 353 |
# The default status is "queued"
|
| 354 |
self.assert_(compat.all(op.status == constants.OP_STATUS_QUEUED
|
| 355 |
for op in ops)) |
| 356 |
|
| 357 |
def _Waitlock1(ops): |
| 358 |
ops[0].status = constants.OP_STATUS_WAITLOCK
|
| 359 |
|
| 360 |
def _Waitlock2(ops): |
| 361 |
ops[0].status = constants.OP_STATUS_SUCCESS
|
| 362 |
ops[1].status = constants.OP_STATUS_SUCCESS
|
| 363 |
ops[2].status = constants.OP_STATUS_WAITLOCK
|
| 364 |
|
| 365 |
def _Running(ops): |
| 366 |
ops[0].status = constants.OP_STATUS_SUCCESS
|
| 367 |
ops[1].status = constants.OP_STATUS_RUNNING
|
| 368 |
for op in ops[2:]: |
| 369 |
op.status = constants.OP_STATUS_QUEUED |
| 370 |
|
| 371 |
def _Canceling1(ops): |
| 372 |
ops[0].status = constants.OP_STATUS_SUCCESS
|
| 373 |
ops[1].status = constants.OP_STATUS_SUCCESS
|
| 374 |
for op in ops[2:]: |
| 375 |
op.status = constants.OP_STATUS_CANCELING |
| 376 |
|
| 377 |
def _Canceling2(ops): |
| 378 |
for op in ops: |
| 379 |
op.status = constants.OP_STATUS_CANCELING |
| 380 |
|
| 381 |
def _Canceled(ops): |
| 382 |
for op in ops: |
| 383 |
op.status = constants.OP_STATUS_CANCELED |
| 384 |
|
| 385 |
def _Error1(ops): |
| 386 |
for idx, op in enumerate(ops): |
| 387 |
if idx > 3: |
| 388 |
op.status = constants.OP_STATUS_ERROR |
| 389 |
else:
|
| 390 |
op.status = constants.OP_STATUS_SUCCESS |
| 391 |
|
| 392 |
def _Error2(ops): |
| 393 |
for op in ops: |
| 394 |
op.status = constants.OP_STATUS_ERROR |
| 395 |
|
| 396 |
def _Success(ops): |
| 397 |
for op in ops: |
| 398 |
op.status = constants.OP_STATUS_SUCCESS |
| 399 |
|
| 400 |
tests = {
|
| 401 |
constants.JOB_STATUS_QUEUED: [_Queued], |
| 402 |
constants.JOB_STATUS_WAITLOCK: [_Waitlock1, _Waitlock2], |
| 403 |
constants.JOB_STATUS_RUNNING: [_Running], |
| 404 |
constants.JOB_STATUS_CANCELING: [_Canceling1, _Canceling2], |
| 405 |
constants.JOB_STATUS_CANCELED: [_Canceled], |
| 406 |
constants.JOB_STATUS_ERROR: [_Error1, _Error2], |
| 407 |
constants.JOB_STATUS_SUCCESS: [_Success], |
| 408 |
} |
| 409 |
|
| 410 |
def _NewJob(): |
| 411 |
job = jqueue._QueuedJob(None, 1, |
| 412 |
[opcodes.OpTestDelay() for _ in range(10)]) |
| 413 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 414 |
self.assert_(compat.all(op.status == constants.OP_STATUS_QUEUED
|
| 415 |
for op in job.ops)) |
| 416 |
return job
|
| 417 |
|
| 418 |
for status in constants.JOB_STATUS_ALL: |
| 419 |
sttests = tests[status] |
| 420 |
assert sttests
|
| 421 |
for fn in sttests: |
| 422 |
job = _NewJob() |
| 423 |
fn(job.ops) |
| 424 |
self.assertEqual(job.CalcStatus(), status)
|
| 425 |
|
| 426 |
|
| 427 |
class _FakeQueueForProc: |
| 428 |
def __init__(self): |
| 429 |
self._acquired = False |
| 430 |
self._updates = []
|
| 431 |
|
| 432 |
def IsAcquired(self): |
| 433 |
return self._acquired |
| 434 |
|
| 435 |
def GetNextUpdate(self): |
| 436 |
return self._updates.pop(0) |
| 437 |
|
| 438 |
def acquire(self, shared=0): |
| 439 |
assert shared == 1 |
| 440 |
self._acquired = True |
| 441 |
|
| 442 |
def release(self): |
| 443 |
assert self._acquired |
| 444 |
self._acquired = False |
| 445 |
|
| 446 |
def UpdateJobUnlocked(self, job, replicate=True): |
| 447 |
assert self._acquired, "Lock not acquired while updating job" |
| 448 |
self._updates.append((job, bool(replicate))) |
| 449 |
|
| 450 |
|
| 451 |
class _FakeExecOpCodeForProc: |
| 452 |
def __init__(self, queue, before_start, after_start): |
| 453 |
self._queue = queue
|
| 454 |
self._before_start = before_start
|
| 455 |
self._after_start = after_start
|
| 456 |
|
| 457 |
def __call__(self, op, cbs, timeout=None, priority=None): |
| 458 |
assert isinstance(op, opcodes.OpTestDummy) |
| 459 |
assert not self._queue.IsAcquired(), \ |
| 460 |
"Queue lock not released when executing opcode"
|
| 461 |
|
| 462 |
if self._before_start: |
| 463 |
self._before_start(timeout, priority)
|
| 464 |
|
| 465 |
cbs.NotifyStart() |
| 466 |
|
| 467 |
if self._after_start: |
| 468 |
self._after_start(op, cbs)
|
| 469 |
|
| 470 |
# Check again after the callbacks
|
| 471 |
assert not self._queue.IsAcquired() |
| 472 |
|
| 473 |
if op.fail:
|
| 474 |
raise errors.OpExecError("Error requested (%s)" % op.result) |
| 475 |
|
| 476 |
return op.result
|
| 477 |
|
| 478 |
|
| 479 |
class _JobProcessorTestUtils: |
| 480 |
def _CreateJob(self, queue, job_id, ops): |
| 481 |
job = jqueue._QueuedJob(queue, job_id, ops) |
| 482 |
self.assertFalse(job.start_timestamp)
|
| 483 |
self.assertFalse(job.end_timestamp)
|
| 484 |
self.assertEqual(len(ops), len(job.ops)) |
| 485 |
self.assert_(compat.all(op.input == inp
|
| 486 |
for (op, inp) in zip(job.ops, ops))) |
| 487 |
self.assertEqual(job.GetInfo(["ops"]), [[op.__getstate__() for op in ops]]) |
| 488 |
return job
|
| 489 |
|
| 490 |
|
| 491 |
class TestJobProcessor(unittest.TestCase, _JobProcessorTestUtils): |
| 492 |
def _GenericCheckJob(self, job): |
| 493 |
assert compat.all(isinstance(op.input, opcodes.OpTestDummy) |
| 494 |
for op in job.ops) |
| 495 |
|
| 496 |
self.assertEqual(job.GetInfo(["opstart", "opexec", "opend"]), |
| 497 |
[[op.start_timestamp for op in job.ops], |
| 498 |
[op.exec_timestamp for op in job.ops], |
| 499 |
[op.end_timestamp for op in job.ops]]) |
| 500 |
self.assertEqual(job.GetInfo(["received_ts", "start_ts", "end_ts"]), |
| 501 |
[job.received_timestamp, |
| 502 |
job.start_timestamp, |
| 503 |
job.end_timestamp]) |
| 504 |
self.assert_(job.start_timestamp)
|
| 505 |
self.assert_(job.end_timestamp)
|
| 506 |
self.assertEqual(job.start_timestamp, job.ops[0].start_timestamp) |
| 507 |
|
| 508 |
def testSuccess(self): |
| 509 |
queue = _FakeQueueForProc() |
| 510 |
|
| 511 |
for (job_id, opcount) in [(25351, 1), (6637, 3), |
| 512 |
(24644, 10), (32207, 100)]: |
| 513 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 514 |
for i in range(opcount)] |
| 515 |
|
| 516 |
# Create job
|
| 517 |
job = self._CreateJob(queue, job_id, ops)
|
| 518 |
|
| 519 |
def _BeforeStart(timeout, priority): |
| 520 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 521 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 522 |
self.assertFalse(queue.IsAcquired())
|
| 523 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 524 |
self.assertFalse(job.cur_opctx)
|
| 525 |
|
| 526 |
def _AfterStart(op, cbs): |
| 527 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 528 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 529 |
|
| 530 |
self.assertFalse(queue.IsAcquired())
|
| 531 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_RUNNING)
|
| 532 |
self.assertFalse(job.cur_opctx)
|
| 533 |
|
| 534 |
# Job is running, cancelling shouldn't be possible
|
| 535 |
(success, _) = job.Cancel() |
| 536 |
self.assertFalse(success)
|
| 537 |
|
| 538 |
opexec = _FakeExecOpCodeForProc(queue, _BeforeStart, _AfterStart) |
| 539 |
|
| 540 |
for idx in range(len(ops)): |
| 541 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 542 |
result = jqueue._JobProcessor(queue, opexec, job)() |
| 543 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 544 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 545 |
if idx == len(ops) - 1: |
| 546 |
# Last opcode
|
| 547 |
self.assert_(result)
|
| 548 |
else:
|
| 549 |
self.assertFalse(result)
|
| 550 |
|
| 551 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 552 |
self.assert_(job.start_timestamp)
|
| 553 |
self.assertFalse(job.end_timestamp)
|
| 554 |
|
| 555 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 556 |
|
| 557 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_SUCCESS)
|
| 558 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_SUCCESS]) |
| 559 |
self.assertEqual(job.GetInfo(["opresult"]), |
| 560 |
[[op.input.result for op in job.ops]]) |
| 561 |
self.assertEqual(job.GetInfo(["opstatus"]), |
| 562 |
[len(job.ops) * [constants.OP_STATUS_SUCCESS]])
|
| 563 |
self.assert_(compat.all(op.start_timestamp and op.end_timestamp |
| 564 |
for op in job.ops)) |
| 565 |
|
| 566 |
self._GenericCheckJob(job)
|
| 567 |
|
| 568 |
# Finished jobs can't be processed any further
|
| 569 |
self.assertRaises(errors.ProgrammerError,
|
| 570 |
jqueue._JobProcessor(queue, opexec, job)) |
| 571 |
|
| 572 |
def testOpcodeError(self): |
| 573 |
queue = _FakeQueueForProc() |
| 574 |
|
| 575 |
testdata = [ |
| 576 |
(17077, 1, 0, 0), |
| 577 |
(1782, 5, 2, 2), |
| 578 |
(18179, 10, 9, 9), |
| 579 |
(4744, 10, 3, 8), |
| 580 |
(23816, 100, 39, 45), |
| 581 |
] |
| 582 |
|
| 583 |
for (job_id, opcount, failfrom, failto) in testdata: |
| 584 |
# Prepare opcodes
|
| 585 |
ops = [opcodes.OpTestDummy(result="Res%s" % i,
|
| 586 |
fail=(failfrom <= i and
|
| 587 |
i <= failto)) |
| 588 |
for i in range(opcount)] |
| 589 |
|
| 590 |
# Create job
|
| 591 |
job = self._CreateJob(queue, job_id, ops)
|
| 592 |
|
| 593 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 594 |
|
| 595 |
for idx in range(len(ops)): |
| 596 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 597 |
result = jqueue._JobProcessor(queue, opexec, job)() |
| 598 |
# queued to waitlock
|
| 599 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 600 |
# waitlock to running
|
| 601 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 602 |
# Opcode result
|
| 603 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 604 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 605 |
|
| 606 |
if idx in (failfrom, len(ops) - 1): |
| 607 |
# Last opcode
|
| 608 |
self.assert_(result)
|
| 609 |
break
|
| 610 |
|
| 611 |
self.assertFalse(result)
|
| 612 |
|
| 613 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 614 |
|
| 615 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 616 |
|
| 617 |
# Check job status
|
| 618 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_ERROR)
|
| 619 |
self.assertEqual(job.GetInfo(["id"]), [job_id]) |
| 620 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_ERROR]) |
| 621 |
|
| 622 |
# Check opcode status
|
| 623 |
data = zip(job.ops,
|
| 624 |
job.GetInfo(["opstatus"])[0], |
| 625 |
job.GetInfo(["opresult"])[0]) |
| 626 |
|
| 627 |
for idx, (op, opstatus, opresult) in enumerate(data): |
| 628 |
if idx < failfrom:
|
| 629 |
assert not op.input.fail |
| 630 |
self.assertEqual(opstatus, constants.OP_STATUS_SUCCESS)
|
| 631 |
self.assertEqual(opresult, op.input.result)
|
| 632 |
elif idx <= failto:
|
| 633 |
assert op.input.fail
|
| 634 |
self.assertEqual(opstatus, constants.OP_STATUS_ERROR)
|
| 635 |
self.assertRaises(errors.OpExecError, errors.MaybeRaise, opresult)
|
| 636 |
else:
|
| 637 |
assert not op.input.fail |
| 638 |
self.assertEqual(opstatus, constants.OP_STATUS_ERROR)
|
| 639 |
self.assertRaises(errors.OpExecError, errors.MaybeRaise, opresult)
|
| 640 |
|
| 641 |
self.assert_(compat.all(op.start_timestamp and op.end_timestamp |
| 642 |
for op in job.ops[:failfrom])) |
| 643 |
|
| 644 |
self._GenericCheckJob(job)
|
| 645 |
|
| 646 |
# Finished jobs can't be processed any further
|
| 647 |
self.assertRaises(errors.ProgrammerError,
|
| 648 |
jqueue._JobProcessor(queue, opexec, job)) |
| 649 |
|
| 650 |
def testCancelWhileInQueue(self): |
| 651 |
queue = _FakeQueueForProc() |
| 652 |
|
| 653 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 654 |
for i in range(5)] |
| 655 |
|
| 656 |
# Create job
|
| 657 |
job_id = 17045
|
| 658 |
job = self._CreateJob(queue, job_id, ops)
|
| 659 |
|
| 660 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 661 |
|
| 662 |
# Mark as cancelled
|
| 663 |
(success, _) = job.Cancel() |
| 664 |
self.assert_(success)
|
| 665 |
|
| 666 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 667 |
|
| 668 |
self.assert_(compat.all(op.status == constants.OP_STATUS_CANCELED
|
| 669 |
for op in job.ops)) |
| 670 |
|
| 671 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 672 |
self.assert_(jqueue._JobProcessor(queue, opexec, job)())
|
| 673 |
|
| 674 |
# Check result
|
| 675 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_CANCELED)
|
| 676 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_CANCELED]) |
| 677 |
self.assertFalse(job.start_timestamp)
|
| 678 |
self.assert_(job.end_timestamp)
|
| 679 |
self.assertFalse(compat.any(op.start_timestamp or op.end_timestamp |
| 680 |
for op in job.ops)) |
| 681 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 682 |
[[constants.OP_STATUS_CANCELED for _ in job.ops], |
| 683 |
["Job canceled by request" for _ in job.ops]]) |
| 684 |
|
| 685 |
def testCancelWhileWaitlockInQueue(self): |
| 686 |
queue = _FakeQueueForProc() |
| 687 |
|
| 688 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 689 |
for i in range(5)] |
| 690 |
|
| 691 |
# Create job
|
| 692 |
job_id = 8645
|
| 693 |
job = self._CreateJob(queue, job_id, ops)
|
| 694 |
|
| 695 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 696 |
|
| 697 |
job.ops[0].status = constants.OP_STATUS_WAITLOCK
|
| 698 |
|
| 699 |
assert len(job.ops) == 5 |
| 700 |
|
| 701 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 702 |
|
| 703 |
# Mark as cancelling
|
| 704 |
(success, _) = job.Cancel() |
| 705 |
self.assert_(success)
|
| 706 |
|
| 707 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 708 |
|
| 709 |
self.assert_(compat.all(op.status == constants.OP_STATUS_CANCELING
|
| 710 |
for op in job.ops)) |
| 711 |
|
| 712 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 713 |
self.assert_(jqueue._JobProcessor(queue, opexec, job)())
|
| 714 |
|
| 715 |
# Check result
|
| 716 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_CANCELED)
|
| 717 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_CANCELED]) |
| 718 |
self.assertFalse(job.start_timestamp)
|
| 719 |
self.assert_(job.end_timestamp)
|
| 720 |
self.assertFalse(compat.any(op.start_timestamp or op.end_timestamp |
| 721 |
for op in job.ops)) |
| 722 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 723 |
[[constants.OP_STATUS_CANCELED for _ in job.ops], |
| 724 |
["Job canceled by request" for _ in job.ops]]) |
| 725 |
|
| 726 |
def testCancelWhileWaitlock(self): |
| 727 |
queue = _FakeQueueForProc() |
| 728 |
|
| 729 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 730 |
for i in range(5)] |
| 731 |
|
| 732 |
# Create job
|
| 733 |
job_id = 11009
|
| 734 |
job = self._CreateJob(queue, job_id, ops)
|
| 735 |
|
| 736 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 737 |
|
| 738 |
def _BeforeStart(timeout, priority): |
| 739 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 740 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 741 |
self.assertFalse(queue.IsAcquired())
|
| 742 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 743 |
|
| 744 |
# Mark as cancelled
|
| 745 |
(success, _) = job.Cancel() |
| 746 |
self.assert_(success)
|
| 747 |
|
| 748 |
self.assert_(compat.all(op.status == constants.OP_STATUS_CANCELING
|
| 749 |
for op in job.ops)) |
| 750 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 751 |
|
| 752 |
def _AfterStart(op, cbs): |
| 753 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 754 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 755 |
self.assertFalse(queue.IsAcquired())
|
| 756 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_RUNNING)
|
| 757 |
|
| 758 |
opexec = _FakeExecOpCodeForProc(queue, _BeforeStart, _AfterStart) |
| 759 |
|
| 760 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 761 |
self.assert_(jqueue._JobProcessor(queue, opexec, job)())
|
| 762 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 763 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 764 |
|
| 765 |
# Check result
|
| 766 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_CANCELED)
|
| 767 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_CANCELED]) |
| 768 |
self.assert_(job.start_timestamp)
|
| 769 |
self.assert_(job.end_timestamp)
|
| 770 |
self.assertFalse(compat.all(op.start_timestamp and op.end_timestamp |
| 771 |
for op in job.ops)) |
| 772 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 773 |
[[constants.OP_STATUS_CANCELED for _ in job.ops], |
| 774 |
["Job canceled by request" for _ in job.ops]]) |
| 775 |
|
| 776 |
def testCancelWhileWaitlockWithTimeout(self): |
| 777 |
queue = _FakeQueueForProc() |
| 778 |
|
| 779 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 780 |
for i in range(5)] |
| 781 |
|
| 782 |
# Create job
|
| 783 |
job_id = 24314
|
| 784 |
job = self._CreateJob(queue, job_id, ops)
|
| 785 |
|
| 786 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 787 |
|
| 788 |
def _BeforeStart(timeout, priority): |
| 789 |
self.assertFalse(queue.IsAcquired())
|
| 790 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 791 |
|
| 792 |
# Mark as cancelled
|
| 793 |
(success, _) = job.Cancel() |
| 794 |
self.assert_(success)
|
| 795 |
|
| 796 |
self.assert_(compat.all(op.status == constants.OP_STATUS_CANCELING
|
| 797 |
for op in job.ops)) |
| 798 |
|
| 799 |
# Fake an acquire attempt timing out
|
| 800 |
raise mcpu.LockAcquireTimeout()
|
| 801 |
|
| 802 |
def _AfterStart(op, cbs): |
| 803 |
self.fail("Should not reach this") |
| 804 |
|
| 805 |
opexec = _FakeExecOpCodeForProc(queue, _BeforeStart, _AfterStart) |
| 806 |
|
| 807 |
self.assert_(jqueue._JobProcessor(queue, opexec, job)())
|
| 808 |
|
| 809 |
# Check result
|
| 810 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_CANCELED)
|
| 811 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_CANCELED]) |
| 812 |
self.assert_(job.start_timestamp)
|
| 813 |
self.assert_(job.end_timestamp)
|
| 814 |
self.assertFalse(compat.all(op.start_timestamp and op.end_timestamp |
| 815 |
for op in job.ops)) |
| 816 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 817 |
[[constants.OP_STATUS_CANCELED for _ in job.ops], |
| 818 |
["Job canceled by request" for _ in job.ops]]) |
| 819 |
|
| 820 |
def testCancelWhileRunning(self): |
| 821 |
# Tests canceling a job with finished opcodes and more, unprocessed ones
|
| 822 |
queue = _FakeQueueForProc() |
| 823 |
|
| 824 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 825 |
for i in range(3)] |
| 826 |
|
| 827 |
# Create job
|
| 828 |
job_id = 28492
|
| 829 |
job = self._CreateJob(queue, job_id, ops)
|
| 830 |
|
| 831 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 832 |
|
| 833 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 834 |
|
| 835 |
# Run one opcode
|
| 836 |
self.assertFalse(jqueue._JobProcessor(queue, opexec, job)())
|
| 837 |
|
| 838 |
# Job goes back to queued
|
| 839 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 840 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 841 |
[[constants.OP_STATUS_SUCCESS, |
| 842 |
constants.OP_STATUS_QUEUED, |
| 843 |
constants.OP_STATUS_QUEUED], |
| 844 |
["Res0", None, None]]) |
| 845 |
|
| 846 |
# Mark as cancelled
|
| 847 |
(success, _) = job.Cancel() |
| 848 |
self.assert_(success)
|
| 849 |
|
| 850 |
# Try processing another opcode (this will actually cancel the job)
|
| 851 |
self.assert_(jqueue._JobProcessor(queue, opexec, job)())
|
| 852 |
|
| 853 |
# Check result
|
| 854 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_CANCELED)
|
| 855 |
self.assertEqual(job.GetInfo(["id"]), [job_id]) |
| 856 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_CANCELED]) |
| 857 |
self.assertEqual(job.GetInfo(["opstatus", "opresult"]), |
| 858 |
[[constants.OP_STATUS_SUCCESS, |
| 859 |
constants.OP_STATUS_CANCELED, |
| 860 |
constants.OP_STATUS_CANCELED], |
| 861 |
["Res0", "Job canceled by request", |
| 862 |
"Job canceled by request"]])
|
| 863 |
|
| 864 |
def testPartiallyRun(self): |
| 865 |
# Tests calling the processor on a job that's been partially run before the
|
| 866 |
# program was restarted
|
| 867 |
queue = _FakeQueueForProc() |
| 868 |
|
| 869 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 870 |
|
| 871 |
for job_id, successcount in [(30697, 1), (2552, 4), (12489, 9)]: |
| 872 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 873 |
for i in range(10)] |
| 874 |
|
| 875 |
# Create job
|
| 876 |
job = self._CreateJob(queue, job_id, ops)
|
| 877 |
|
| 878 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 879 |
|
| 880 |
for _ in range(successcount): |
| 881 |
self.assertFalse(jqueue._JobProcessor(queue, opexec, job)())
|
| 882 |
|
| 883 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 884 |
self.assertEqual(job.GetInfo(["opstatus"]), |
| 885 |
[[constants.OP_STATUS_SUCCESS |
| 886 |
for _ in range(successcount)] + |
| 887 |
[constants.OP_STATUS_QUEUED |
| 888 |
for _ in range(len(ops) - successcount)]]) |
| 889 |
|
| 890 |
self.assert_(job.ops_iter)
|
| 891 |
|
| 892 |
# Serialize and restore (simulates program restart)
|
| 893 |
newjob = jqueue._QueuedJob.Restore(queue, job.Serialize()) |
| 894 |
self.assertFalse(newjob.ops_iter)
|
| 895 |
self._TestPartial(newjob, successcount)
|
| 896 |
|
| 897 |
def _TestPartial(self, job, successcount): |
| 898 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 899 |
self.assertEqual(job.start_timestamp, job.ops[0].start_timestamp) |
| 900 |
|
| 901 |
queue = _FakeQueueForProc() |
| 902 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 903 |
|
| 904 |
for remaining in reversed(range(len(job.ops) - successcount)): |
| 905 |
result = jqueue._JobProcessor(queue, opexec, job)() |
| 906 |
|
| 907 |
if remaining == 0: |
| 908 |
# Last opcode
|
| 909 |
self.assert_(result)
|
| 910 |
break
|
| 911 |
|
| 912 |
self.assertFalse(result)
|
| 913 |
|
| 914 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 915 |
|
| 916 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_SUCCESS)
|
| 917 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_SUCCESS]) |
| 918 |
self.assertEqual(job.GetInfo(["opresult"]), |
| 919 |
[[op.input.result for op in job.ops]]) |
| 920 |
self.assertEqual(job.GetInfo(["opstatus"]), |
| 921 |
[[constants.OP_STATUS_SUCCESS for _ in job.ops]]) |
| 922 |
self.assert_(compat.all(op.start_timestamp and op.end_timestamp |
| 923 |
for op in job.ops)) |
| 924 |
|
| 925 |
self._GenericCheckJob(job)
|
| 926 |
|
| 927 |
# Finished jobs can't be processed any further
|
| 928 |
self.assertRaises(errors.ProgrammerError,
|
| 929 |
jqueue._JobProcessor(queue, opexec, job)) |
| 930 |
|
| 931 |
# ... also after being restored
|
| 932 |
job2 = jqueue._QueuedJob.Restore(queue, job.Serialize()) |
| 933 |
self.assertRaises(errors.ProgrammerError,
|
| 934 |
jqueue._JobProcessor(queue, opexec, job2)) |
| 935 |
|
| 936 |
def testProcessorOnRunningJob(self): |
| 937 |
ops = [opcodes.OpTestDummy(result="result", fail=False)] |
| 938 |
|
| 939 |
queue = _FakeQueueForProc() |
| 940 |
opexec = _FakeExecOpCodeForProc(queue, None, None) |
| 941 |
|
| 942 |
# Create job
|
| 943 |
job = self._CreateJob(queue, 9571, ops) |
| 944 |
|
| 945 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 946 |
|
| 947 |
job.ops[0].status = constants.OP_STATUS_RUNNING
|
| 948 |
|
| 949 |
assert len(job.ops) == 1 |
| 950 |
|
| 951 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_RUNNING)
|
| 952 |
|
| 953 |
# Calling on running job must fail
|
| 954 |
self.assertRaises(errors.ProgrammerError,
|
| 955 |
jqueue._JobProcessor(queue, opexec, job)) |
| 956 |
|
| 957 |
def testLogMessages(self): |
| 958 |
# Tests the "Feedback" callback function
|
| 959 |
queue = _FakeQueueForProc() |
| 960 |
|
| 961 |
messages = {
|
| 962 |
1: [
|
| 963 |
(None, "Hello"), |
| 964 |
(None, "World"), |
| 965 |
(constants.ELOG_MESSAGE, "there"),
|
| 966 |
], |
| 967 |
4: [
|
| 968 |
(constants.ELOG_JQUEUE_TEST, (1, 2, 3)), |
| 969 |
(constants.ELOG_JQUEUE_TEST, ("other", "type")), |
| 970 |
], |
| 971 |
} |
| 972 |
ops = [opcodes.OpTestDummy(result="Logtest%s" % i, fail=False, |
| 973 |
messages=messages.get(i, [])) |
| 974 |
for i in range(5)] |
| 975 |
|
| 976 |
# Create job
|
| 977 |
job = self._CreateJob(queue, 29386, ops) |
| 978 |
|
| 979 |
def _BeforeStart(timeout, priority): |
| 980 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 981 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 982 |
self.assertFalse(queue.IsAcquired())
|
| 983 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 984 |
|
| 985 |
def _AfterStart(op, cbs): |
| 986 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 987 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 988 |
self.assertFalse(queue.IsAcquired())
|
| 989 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_RUNNING)
|
| 990 |
|
| 991 |
self.assertRaises(AssertionError, cbs.Feedback, |
| 992 |
"too", "many", "arguments") |
| 993 |
|
| 994 |
for (log_type, msg) in op.messages: |
| 995 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 996 |
if log_type:
|
| 997 |
cbs.Feedback(log_type, msg) |
| 998 |
else:
|
| 999 |
cbs.Feedback(msg) |
| 1000 |
# Check for job update without replication
|
| 1001 |
self.assertEqual(queue.GetNextUpdate(), (job, False)) |
| 1002 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 1003 |
|
| 1004 |
opexec = _FakeExecOpCodeForProc(queue, _BeforeStart, _AfterStart) |
| 1005 |
|
| 1006 |
for remaining in reversed(range(len(job.ops))): |
| 1007 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 1008 |
result = jqueue._JobProcessor(queue, opexec, job)() |
| 1009 |
self.assertEqual(queue.GetNextUpdate(), (job, True)) |
| 1010 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 1011 |
|
| 1012 |
if remaining == 0: |
| 1013 |
# Last opcode
|
| 1014 |
self.assert_(result)
|
| 1015 |
break
|
| 1016 |
|
| 1017 |
self.assertFalse(result)
|
| 1018 |
|
| 1019 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 1020 |
|
| 1021 |
self.assertRaises(IndexError, queue.GetNextUpdate) |
| 1022 |
|
| 1023 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_SUCCESS)
|
| 1024 |
self.assertEqual(job.GetInfo(["opresult"]), |
| 1025 |
[[op.input.result for op in job.ops]]) |
| 1026 |
|
| 1027 |
logmsgcount = sum(len(m) for m in messages.values()) |
| 1028 |
|
| 1029 |
self._CheckLogMessages(job, logmsgcount)
|
| 1030 |
|
| 1031 |
# Serialize and restore (simulates program restart)
|
| 1032 |
newjob = jqueue._QueuedJob.Restore(queue, job.Serialize()) |
| 1033 |
self._CheckLogMessages(newjob, logmsgcount)
|
| 1034 |
|
| 1035 |
# Check each message
|
| 1036 |
prevserial = -1
|
| 1037 |
for idx, oplog in enumerate(job.GetInfo(["oplog"])[0]): |
| 1038 |
for (serial, timestamp, log_type, msg) in oplog: |
| 1039 |
(exptype, expmsg) = messages.get(idx).pop(0)
|
| 1040 |
if exptype:
|
| 1041 |
self.assertEqual(log_type, exptype)
|
| 1042 |
else:
|
| 1043 |
self.assertEqual(log_type, constants.ELOG_MESSAGE)
|
| 1044 |
self.assertEqual(expmsg, msg)
|
| 1045 |
self.assert_(serial > prevserial)
|
| 1046 |
prevserial = serial |
| 1047 |
|
| 1048 |
def _CheckLogMessages(self, job, count): |
| 1049 |
# Check serial
|
| 1050 |
self.assertEqual(job.log_serial, count)
|
| 1051 |
|
| 1052 |
# No filter
|
| 1053 |
self.assertEqual(job.GetLogEntries(None), |
| 1054 |
[entry for entries in job.GetInfo(["oplog"])[0] if entries |
| 1055 |
for entry in entries]) |
| 1056 |
|
| 1057 |
# Filter with serial
|
| 1058 |
assert count > 3 |
| 1059 |
self.assert_(job.GetLogEntries(3)) |
| 1060 |
self.assertEqual(job.GetLogEntries(3), |
| 1061 |
[entry for entries in job.GetInfo(["oplog"])[0] if entries |
| 1062 |
for entry in entries][3:]) |
| 1063 |
|
| 1064 |
# No log message after highest serial
|
| 1065 |
self.assertFalse(job.GetLogEntries(count))
|
| 1066 |
self.assertFalse(job.GetLogEntries(count + 3)) |
| 1067 |
|
| 1068 |
|
| 1069 |
class _FakeTimeoutStrategy: |
| 1070 |
def __init__(self, timeouts): |
| 1071 |
self.timeouts = timeouts
|
| 1072 |
self.attempts = 0 |
| 1073 |
self.last_timeout = None |
| 1074 |
|
| 1075 |
def NextAttempt(self): |
| 1076 |
self.attempts += 1 |
| 1077 |
if self.timeouts: |
| 1078 |
timeout = self.timeouts.pop(0) |
| 1079 |
else:
|
| 1080 |
timeout = None
|
| 1081 |
self.last_timeout = timeout
|
| 1082 |
return timeout
|
| 1083 |
|
| 1084 |
|
| 1085 |
class TestJobProcessorTimeouts(unittest.TestCase, _JobProcessorTestUtils): |
| 1086 |
def setUp(self): |
| 1087 |
self.queue = _FakeQueueForProc()
|
| 1088 |
self.job = None |
| 1089 |
self.curop = None |
| 1090 |
self.opcounter = None |
| 1091 |
self.timeout_strategy = None |
| 1092 |
self.retries = 0 |
| 1093 |
self.prev_tsop = None |
| 1094 |
self.prev_prio = None |
| 1095 |
self.prev_status = None |
| 1096 |
self.lock_acq_prio = None |
| 1097 |
self.gave_lock = None |
| 1098 |
self.done_lock_before_blocking = False |
| 1099 |
|
| 1100 |
def _BeforeStart(self, timeout, priority): |
| 1101 |
job = self.job
|
| 1102 |
|
| 1103 |
# If status has changed, job must've been written
|
| 1104 |
if self.prev_status != self.job.ops[self.curop].status: |
| 1105 |
self.assertEqual(self.queue.GetNextUpdate(), (job, True)) |
| 1106 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1107 |
|
| 1108 |
self.assertFalse(self.queue.IsAcquired()) |
| 1109 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 1110 |
|
| 1111 |
ts = self.timeout_strategy
|
| 1112 |
|
| 1113 |
self.assert_(timeout is None or isinstance(timeout, (int, float))) |
| 1114 |
self.assertEqual(timeout, ts.last_timeout)
|
| 1115 |
self.assertEqual(priority, job.ops[self.curop].priority) |
| 1116 |
|
| 1117 |
self.gave_lock = True |
| 1118 |
self.lock_acq_prio = priority
|
| 1119 |
|
| 1120 |
if (self.curop == 3 and |
| 1121 |
job.ops[self.curop].priority == constants.OP_PRIO_HIGHEST + 3): |
| 1122 |
# Give locks before running into blocking acquire
|
| 1123 |
assert self.retries == 7 |
| 1124 |
self.retries = 0 |
| 1125 |
self.done_lock_before_blocking = True |
| 1126 |
return
|
| 1127 |
|
| 1128 |
if self.retries > 0: |
| 1129 |
self.assert_(timeout is not None) |
| 1130 |
self.retries -= 1 |
| 1131 |
self.gave_lock = False |
| 1132 |
raise mcpu.LockAcquireTimeout()
|
| 1133 |
|
| 1134 |
if job.ops[self.curop].priority == constants.OP_PRIO_HIGHEST: |
| 1135 |
assert self.retries == 0, "Didn't exhaust all retries at highest priority" |
| 1136 |
assert not ts.timeouts |
| 1137 |
self.assert_(timeout is None) |
| 1138 |
|
| 1139 |
def _AfterStart(self, op, cbs): |
| 1140 |
job = self.job
|
| 1141 |
|
| 1142 |
# Setting to "running" requires an update
|
| 1143 |
self.assertEqual(self.queue.GetNextUpdate(), (job, True)) |
| 1144 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1145 |
|
| 1146 |
self.assertFalse(self.queue.IsAcquired()) |
| 1147 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_RUNNING)
|
| 1148 |
|
| 1149 |
# Job is running, cancelling shouldn't be possible
|
| 1150 |
(success, _) = job.Cancel() |
| 1151 |
self.assertFalse(success)
|
| 1152 |
|
| 1153 |
def _NextOpcode(self): |
| 1154 |
self.curop = self.opcounter.next() |
| 1155 |
self.prev_prio = self.job.ops[self.curop].priority |
| 1156 |
self.prev_status = self.job.ops[self.curop].status |
| 1157 |
|
| 1158 |
def _NewTimeoutStrategy(self): |
| 1159 |
job = self.job
|
| 1160 |
|
| 1161 |
self.assertEqual(self.retries, 0) |
| 1162 |
|
| 1163 |
if self.prev_tsop == self.curop: |
| 1164 |
# Still on the same opcode, priority must've been increased
|
| 1165 |
self.assertEqual(self.prev_prio, job.ops[self.curop].priority + 1) |
| 1166 |
|
| 1167 |
if self.curop == 1: |
| 1168 |
# Normal retry
|
| 1169 |
timeouts = range(10, 31, 10) |
| 1170 |
self.retries = len(timeouts) - 1 |
| 1171 |
|
| 1172 |
elif self.curop == 2: |
| 1173 |
# Let this run into a blocking acquire
|
| 1174 |
timeouts = range(11, 61, 12) |
| 1175 |
self.retries = len(timeouts) |
| 1176 |
|
| 1177 |
elif self.curop == 3: |
| 1178 |
# Wait for priority to increase, but give lock before blocking acquire
|
| 1179 |
timeouts = range(12, 100, 14) |
| 1180 |
self.retries = len(timeouts) |
| 1181 |
|
| 1182 |
self.assertFalse(self.done_lock_before_blocking) |
| 1183 |
|
| 1184 |
elif self.curop == 4: |
| 1185 |
self.assert_(self.done_lock_before_blocking) |
| 1186 |
|
| 1187 |
# Timeouts, but no need to retry
|
| 1188 |
timeouts = range(10, 31, 10) |
| 1189 |
self.retries = 0 |
| 1190 |
|
| 1191 |
elif self.curop == 5: |
| 1192 |
# Normal retry
|
| 1193 |
timeouts = range(19, 100, 11) |
| 1194 |
self.retries = len(timeouts) |
| 1195 |
|
| 1196 |
else:
|
| 1197 |
timeouts = [] |
| 1198 |
self.retries = 0 |
| 1199 |
|
| 1200 |
assert len(job.ops) == 10 |
| 1201 |
assert self.retries <= len(timeouts) |
| 1202 |
|
| 1203 |
ts = _FakeTimeoutStrategy(timeouts) |
| 1204 |
|
| 1205 |
self.timeout_strategy = ts
|
| 1206 |
self.prev_tsop = self.curop |
| 1207 |
self.prev_prio = job.ops[self.curop].priority |
| 1208 |
|
| 1209 |
return ts
|
| 1210 |
|
| 1211 |
def testTimeout(self): |
| 1212 |
ops = [opcodes.OpTestDummy(result="Res%s" % i, fail=False) |
| 1213 |
for i in range(10)] |
| 1214 |
|
| 1215 |
# Create job
|
| 1216 |
job_id = 15801
|
| 1217 |
job = self._CreateJob(self.queue, job_id, ops) |
| 1218 |
self.job = job
|
| 1219 |
|
| 1220 |
self.opcounter = itertools.count(0) |
| 1221 |
|
| 1222 |
opexec = _FakeExecOpCodeForProc(self.queue, self._BeforeStart, |
| 1223 |
self._AfterStart)
|
| 1224 |
tsf = self._NewTimeoutStrategy
|
| 1225 |
|
| 1226 |
self.assertFalse(self.done_lock_before_blocking) |
| 1227 |
|
| 1228 |
while True: |
| 1229 |
proc = jqueue._JobProcessor(self.queue, opexec, job,
|
| 1230 |
_timeout_strategy_factory=tsf) |
| 1231 |
|
| 1232 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1233 |
|
| 1234 |
if self.curop is not None: |
| 1235 |
self.prev_status = self.job.ops[self.curop].status |
| 1236 |
|
| 1237 |
self.lock_acq_prio = None |
| 1238 |
|
| 1239 |
result = proc(_nextop_fn=self._NextOpcode)
|
| 1240 |
assert self.curop is not None |
| 1241 |
|
| 1242 |
if result or self.gave_lock: |
| 1243 |
# Got lock and/or job is done, result must've been written
|
| 1244 |
self.assertFalse(job.cur_opctx)
|
| 1245 |
self.assertEqual(self.queue.GetNextUpdate(), (job, True)) |
| 1246 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1247 |
self.assertEqual(self.lock_acq_prio, job.ops[self.curop].priority) |
| 1248 |
self.assert_(job.ops[self.curop].exec_timestamp) |
| 1249 |
|
| 1250 |
if result:
|
| 1251 |
self.assertFalse(job.cur_opctx)
|
| 1252 |
break
|
| 1253 |
|
| 1254 |
self.assertFalse(result)
|
| 1255 |
|
| 1256 |
if self.curop == 0: |
| 1257 |
self.assertEqual(job.ops[self.curop].start_timestamp, |
| 1258 |
job.start_timestamp) |
| 1259 |
|
| 1260 |
if self.gave_lock: |
| 1261 |
# Opcode finished, but job not yet done
|
| 1262 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_QUEUED)
|
| 1263 |
else:
|
| 1264 |
# Did not get locks
|
| 1265 |
self.assert_(job.cur_opctx)
|
| 1266 |
self.assertEqual(job.cur_opctx._timeout_strategy._fn,
|
| 1267 |
self.timeout_strategy.NextAttempt)
|
| 1268 |
self.assertFalse(job.ops[self.curop].exec_timestamp) |
| 1269 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_WAITLOCK)
|
| 1270 |
|
| 1271 |
# If priority has changed since acquiring locks, the job must've been
|
| 1272 |
# updated
|
| 1273 |
if self.lock_acq_prio != job.ops[self.curop].priority: |
| 1274 |
self.assertEqual(self.queue.GetNextUpdate(), (job, True)) |
| 1275 |
|
| 1276 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1277 |
|
| 1278 |
self.assert_(job.start_timestamp)
|
| 1279 |
self.assertFalse(job.end_timestamp)
|
| 1280 |
|
| 1281 |
self.assertEqual(self.curop, len(job.ops) - 1) |
| 1282 |
self.assertEqual(self.job, job) |
| 1283 |
self.assertEqual(self.opcounter.next(), len(job.ops)) |
| 1284 |
self.assert_(self.done_lock_before_blocking) |
| 1285 |
|
| 1286 |
self.assertRaises(IndexError, self.queue.GetNextUpdate) |
| 1287 |
self.assertEqual(job.CalcStatus(), constants.JOB_STATUS_SUCCESS)
|
| 1288 |
self.assertEqual(job.GetInfo(["status"]), [constants.JOB_STATUS_SUCCESS]) |
| 1289 |
self.assertEqual(job.GetInfo(["opresult"]), |
| 1290 |
[[op.input.result for op in job.ops]]) |
| 1291 |
self.assertEqual(job.GetInfo(["opstatus"]), |
| 1292 |
[len(job.ops) * [constants.OP_STATUS_SUCCESS]])
|
| 1293 |
self.assert_(compat.all(op.start_timestamp and op.end_timestamp |
| 1294 |
for op in job.ops)) |
| 1295 |
|
| 1296 |
# Finished jobs can't be processed any further
|
| 1297 |
self.assertRaises(errors.ProgrammerError,
|
| 1298 |
jqueue._JobProcessor(self.queue, opexec, job))
|
| 1299 |
|
| 1300 |
|
| 1301 |
if __name__ == "__main__": |
| 1302 |
testutils.GanetiTestProgram() |