-
Notifications
You must be signed in to change notification settings - Fork 60
/
worker.py
377 lines (341 loc) · 14.4 KB
/
worker.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
#
# worker.py - Thread that shepherds a job through it execution sequence
#
import threading
import time
import logging
import tempfile
import requests
from requests.packages.urllib3.util.retry import Retry
from requests.adapters import HTTPAdapter
import os
import shutil
from datetime import datetime
from config import Config
#
# Worker - The worker class is very simple and very dumb. The goal is
# to walk through the VMMS interface, track the job's progress, and if
# anything goes wrong, recover cleanly from it.
#
# The issue is that these VMMS functions can block, taking a
# significant amount of time. By running each worker as a thread, each
# worker can spend as much time necessary on its job without blocking
# anything else in the system.
#
class Worker(threading.Thread):
def __init__(self, job, vmms, jobQueue, preallocator, preVM):
threading.Thread.__init__(self)
self.daemon = True
self.job = job
self.vmms = vmms
self.jobQueue = jobQueue
self.preallocator = preallocator
self.preVM = preVM
threading.Thread.__init__(self)
self.log = logging.getLogger("Worker")
#
# Worker helper functions
#
def detachVM(self, return_vm=False, replace_vm=False):
"""detachVM - Detach the VM from this worker. The options are
to return it to the pool's free list (return_vm), destroy it
(not return_vm), and if destroying it, whether to replace it
or not in the pool (replace_vm). The worker must always call
this function before returning.
"""
# job-owned instance, simply destroy after job is completed
if self.job.accessKeyId:
self.vmms.safeDestroyVM(self.job.vm)
elif return_vm:
self.preallocator.freeVM(self.job.vm)
else:
self.vmms.safeDestroyVM(self.job.vm)
if replace_vm:
self.preallocator.createVM(self.job.vm)
# Important: don't remove the VM from the pool until its
# replacement has been created. Otherwise there is a
# potential race where the job manager thinks that the
# pool is empty and creates a spurious vm.
self.preallocator.removeVM(self.job.vm)
def rescheduleJob(self, hdrfile, ret, err):
"""rescheduleJob - Reschedule a job that has failed because
of a system error, such as a VM timing out or a connection
failure.
"""
self.log.error("Job %s:%d failed: %s" % (self.job.name, self.job.id, err))
self.job.appendTrace(
"%s|Job %s:%d failed: %s"
% (datetime.now().ctime(), self.job.name, self.job.id, err)
)
# Try a few times before giving up
if self.job.retries < Config.JOB_RETRIES:
try:
os.remove(hdrfile)
except OSError:
pass
self.detachVM(return_vm=False, replace_vm=True)
self.jobQueue.unassignJob(self.job.id)
# Here is where we give up
else:
self.jobQueue.makeDead(self.job.id, err)
self.appendMsg(
hdrfile,
"Internal error: Unable to complete job after %d tries. Pleae resubmit"
% (Config.JOB_RETRIES),
)
self.appendMsg(
hdrfile,
"Job status: waitVM=%s copyIn=%s runJob=%s copyOut=%s"
% (ret["waitvm"], ret["copyin"], ret["runjob"], ret["copyout"]),
)
self.catFiles(hdrfile, self.job.outputFile)
self.detachVM(return_vm=False, replace_vm=True)
self.notifyServer(self.job)
def appendMsg(self, filename, msg):
"""appendMsg - Append a timestamped Tango message to a file"""
f = open(filename, "a")
f.write("Autograder [%s]: %s\n" % (datetime.now().ctime(), msg))
f.close()
def catFiles(self, f1, f2):
"""catFiles - cat f1 f2 > f2, where f1 is the Tango header
and f2 is the output from the Autodriver
"""
self.appendMsg(f1, "Here is the output from the autograder:\n---")
(wfd, tmpname) = tempfile.mkstemp(dir=os.path.dirname(f2))
wf = os.fdopen(wfd, "ab")
with open(f1, "rb") as f1fd:
shutil.copyfileobj(f1fd, wf)
# f2 may not exist if autograder failed
try:
with open(f2, "rb") as f2fd:
shutil.copyfileobj(f2fd, wf)
except OSError:
pass
wf.close()
os.rename(tmpname, f2)
os.remove(f1)
def notifyServer(self, job):
try:
if job.notifyURL:
outputFileName = job.outputFile.split("/")[-1] # get filename from path
fh = open(job.outputFile, "rb")
files = {"file": str(fh.read(), errors="ignore")}
hdrs = {"Filename": outputFileName}
self.log.debug("Sending request to %s" % job.notifyURL)
with requests.session() as s:
# urllib3 retry, allow POST to be retried, use backoffs
r = Retry(total=10, allowed_methods=False, backoff_factor=1)
s.mount("http://", HTTPAdapter(max_retries=r))
s.mount("https://", HTTPAdapter(max_retries=r))
response = s.post(
job.notifyURL, files=files, headers=hdrs, verify=False
)
self.log.info(
"Response from callback to %s:%s"
% (job.notifyURL, response.content)
)
fh.close()
except Exception as e:
self.log.debug("Error in notifyServer: %s" % str(e))
#
# Main worker function
#
def run(self):
"""run - Step a job through its execution sequence"""
try:
# Hash of return codes for each step
ret = {}
ret["waitvm"] = None
ret["copyin"] = None
ret["runjob"] = None
ret["copyout"] = None
self.log.debug("Run worker")
vm = None
# Header message for user
hdrfile = tempfile.mktemp()
self.appendMsg(hdrfile, "Received job %s:%d" % (self.job.name, self.job.id))
# Assigning job to a preallocated VM
if self.preVM: # self.preVM:
self.log.debug("Assigning job to preallocated VM")
self.job.makeVM(self.preVM)
self.log.info(
"Assigned job %s:%d existing VM %s"
% (
self.job.name,
self.job.id,
self.vmms.instanceName(self.preVM.id, self.preVM.name),
)
)
self.job.appendTrace(
"%s|Assigned job %s:%d existing VM %s"
% (
datetime.now().ctime(),
self.job.name,
self.job.id,
self.vmms.instanceName(self.preVM.id, self.preVM.name),
)
)
self.log.debug("Assigned job to preallocated VM")
# Assigning job to a new VM
else:
self.log.debug("Assigning job to a new VM")
self.job.syncRemote()
self.job.vm.id = self.job.id
self.job.updateRemote()
self.log.info(
"Assigned job %s:%d new VM %s"
% (
self.job.name,
self.job.id,
self.vmms.instanceName(self.job.vm.id, self.job.vm.name),
)
)
self.job.appendTrace(
"%s|Assigned job %s:%d new VM %s"
% (
datetime.now().ctime(),
self.job.name,
self.job.id,
self.vmms.instanceName(self.job.vm.id, self.job.vm.name),
)
)
# Host name returned from EC2 is stored in the vm object
self.vmms.initializeVM(self.job.vm)
self.log.debug("Asigned job to a new VM")
vm = self.job.vm
# Wait for the instance to be ready
self.log.debug(
"Job %s:%d waiting for VM %s"
% (self.job.name, self.job.id, self.vmms.instanceName(vm.id, vm.name))
)
self.job.appendTrace(
"%s|Job %s:%d waiting for VM %s"
% (
datetime.now().ctime(),
self.job.name,
self.job.id,
self.vmms.instanceName(vm.id, vm.name),
)
)
self.log.debug("Waiting for VM")
ret["waitvm"] = self.vmms.waitVM(vm, Config.WAITVM_TIMEOUT)
self.log.debug("Waited for VM")
# If the instance did not become ready in a reasonable
# amount of time, then reschedule the job, detach the VM,
# and exit worker
if ret["waitvm"] == -1:
Config.waitvm_timeouts += 1
self.rescheduleJob(
hdrfile,
ret,
"Internal error: waitVM timeout after %d secs"
% Config.WAITVM_TIMEOUT,
)
# Thread Exit after waitVM timeout
return
self.log.info(
"VM %s ready for job %s:%d"
% (self.vmms.instanceName(vm.id, vm.name), self.job.name, self.job.id)
)
self.job.appendTrace(
"%s|VM %s ready for job %s:%d"
% (
datetime.now().ctime(),
self.vmms.instanceName(vm.id, vm.name),
self.job.name,
self.job.id,
)
)
# Copy input files to VM
ret["copyin"] = self.vmms.copyIn(vm, self.job.input)
if ret["copyin"] != 0:
Config.copyin_errors += 1
self.log.info(
"Input copied for job %s:%d [status=%d]"
% (self.job.name, self.job.id, ret["copyin"])
)
self.job.appendTrace(
"%s|Input copied for job %s:%d [status=%d]"
% (datetime.now().ctime(), self.job.name, self.job.id, ret["copyin"])
)
# Run the job on the virtual machine
ret["runjob"] = self.vmms.runJob(
vm,
self.job.timeout,
self.job.maxOutputFileSize,
self.job.disableNetwork,
)
if ret["runjob"] != 0:
Config.runjob_errors += 1
if ret["runjob"] == -1:
Config.runjob_timeouts += 1
self.log.info(
"Job %s:%d executed [status=%s]"
% (self.job.name, self.job.id, ret["runjob"])
)
self.job.appendTrace(
"%s|Job %s:%d executed [status=%s]"
% (datetime.now().ctime(), self.job.name, self.job.id, ret["runjob"])
)
# Copy the output back.
ret["copyout"] = self.vmms.copyOut(vm, self.job.outputFile)
if ret["copyout"] != 0:
Config.copyout_errors += 1
self.log.info(
"Output copied for job %s:%d [status=%d]"
% (self.job.name, self.job.id, ret["copyout"])
)
self.job.appendTrace(
"%s|Output copied for job %s:%d [status=%d]"
% (datetime.now().ctime(), self.job.name, self.job.id, ret["copyout"])
)
# Job termination. Notice that Tango considers
# things like runjob timeouts and makefile errors to be
# normal termination and doesn't reschedule the job.
self.log.info("Success: job %s:%d finished" % (self.job.name, self.job.id))
# Move the job from the live queue to the dead queue
# with an explanatory message
msg = "Success: Autodriver returned normally"
(returnVM, replaceVM) = (True, False)
if ret["copyin"] != 0:
msg = "Error: Copy in to VM failed (status=%d)" % (ret["copyin"])
elif ret["runjob"] != 0:
if ret["runjob"] == 1: # This should never happen
msg = "Error: Autodriver usage error (status=%d)" % (ret["runjob"])
elif ret["runjob"] == 2:
msg = "Error: Job timed out after %d seconds" % (self.job.timeout)
elif ret["runjob"] == 3: # EXIT_OSERROR in Autodriver
# Abnormal job termination (Autodriver encountered an OS
# error). Assume that the VM is damaged. Destroy this VM
# and do not retry the job since the job may have damaged
# the VM.
msg = "Error: OS error while running job on VM"
(returnVM, replaceVM) = (False, True)
else: # This should never happen
msg = "Error: Unknown autodriver error (status=%d)" % (
ret["runjob"]
)
elif ret["copyout"] != 0:
msg += "Error: Copy out from VM failed (status=%d)" % (ret["copyout"])
self.jobQueue.makeDead(self.job.id, msg)
# Update the text that users see in the autograder output file
self.appendMsg(hdrfile, msg)
self.catFiles(hdrfile, self.job.outputFile)
# Thread exit after termination
self.detachVM(return_vm=returnVM, replace_vm=replaceVM)
self.notifyServer(self.job)
return
#
# Exception: ec2CallError - Raised by ec2Call()
#
except Exception as err:
self.log.exception("Internal Error")
self.appendMsg(self.job.outputFile, "Internal Error: %s" % err)
# if vm is set, then the normal job assignment completed,
# and detachVM can be run
# if vm is not set but self.preVM is set, we still need
# to return the VM, but have to initialize self.job.vm first
if self.preVM and not vm:
vm = self.job.vm = self.preVM
if vm:
self.detachVM(return_vm=False, replace_vm=True)