root / qa / rapi-workload.py @ dd2bc9b6
History | View | Annotate | Download (34.6 kB)
1 |
#!/usr/bin/python -u
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2013 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 providing a large amount of RAPI calls to Ganeti.
|
23 |
|
24 |
"""
|
25 |
|
26 |
# pylint: disable=C0103
|
27 |
# due to invalid name
|
28 |
|
29 |
import inspect |
30 |
import sys |
31 |
import types |
32 |
|
33 |
import ganeti.constants as constants |
34 |
from ganeti.rapi.client import GanetiApiError, NODE_EVAC_PRI, NODE_EVAC_SEC |
35 |
|
36 |
import qa_config |
37 |
import qa_error |
38 |
import qa_node |
39 |
import qa_rapi |
40 |
|
41 |
|
42 |
# The purpose of this file is to provide a stable and extensive RAPI workload
|
43 |
# that manipulates the cluster only using RAPI commands, with the assumption
|
44 |
# that an empty cluster was set up beforehand. All the nodes that can be added
|
45 |
# to the cluster should be a part of it, and no instances should be present.
|
46 |
#
|
47 |
# Its intended use is in RAPI compatibility tests, where different versions with
|
48 |
# possibly vastly different QAs must be compared. Running the QA on both
|
49 |
# versions of the cluster will produce RAPI calls, but there is no guarantee
|
50 |
# that they will match, or that functions invoked in between will not change the
|
51 |
# results.
|
52 |
#
|
53 |
# By using only RAPI functions, we are sure to be able to capture and log all
|
54 |
# the changes in cluster state, and be able to compare them afterwards.
|
55 |
#
|
56 |
# The functionality of the QA is still used to generate a functioning,
|
57 |
# RAPI-enabled cluster, and to set up a C{GanetiRapiClient} capable of issuing
|
58 |
# commands to the cluster.
|
59 |
#
|
60 |
# Due to the fact that not all calls issued as a part of the workload might be
|
61 |
# implemented in the different versions of Ganeti, the client does not halt or
|
62 |
# produce a non-zero exit code upon encountering a RAPI error. Instead, it
|
63 |
# reports it and moves on. Any utility comparing the requests should account for
|
64 |
# this.
|
65 |
|
66 |
|
67 |
def MockMethod(*_args, **_kwargs): |
68 |
""" Absorbs all arguments, does nothing, returns None.
|
69 |
|
70 |
"""
|
71 |
return None |
72 |
|
73 |
|
74 |
RAPI_USERNAME = "ganeti-qa"
|
75 |
|
76 |
|
77 |
class GanetiRapiClientWrapper(object): |
78 |
""" Creates and initializes a GanetiRapiClient, and acts as a wrapper invoking
|
79 |
only the methods that the version of the client actually uses.
|
80 |
|
81 |
"""
|
82 |
def __init__(self): |
83 |
self._client = qa_rapi.Setup(RAPI_USERNAME,
|
84 |
qa_rapi.LookupRapiSecret(RAPI_USERNAME)) |
85 |
|
86 |
self._method_invocations = {}
|
87 |
|
88 |
def _RecordMethodInvocation(self, name, arg_dict): |
89 |
""" Records the invocation of a C{GanetiRAPIClient} method, noting the
|
90 |
argument and the method names.
|
91 |
|
92 |
"""
|
93 |
if name not in self._method_invocations: |
94 |
self._method_invocations[name] = set() |
95 |
|
96 |
for named_arg in arg_dict: |
97 |
self._method_invocations[name].add(named_arg)
|
98 |
|
99 |
def _InvokerCreator(self, fn, name): |
100 |
""" Returns an invoker function that will invoke the given function
|
101 |
with any arguments passed to the invoker at a later time, while
|
102 |
catching any specific non-fatal errors we would like to know more
|
103 |
about.
|
104 |
|
105 |
@type fn arbitrary function
|
106 |
@param fn The function to invoke later.
|
107 |
@type name string
|
108 |
@param name The name of the function, for debugging purposes.
|
109 |
@rtype function
|
110 |
|
111 |
"""
|
112 |
def decoratedFn(*args, **kwargs): |
113 |
result = None
|
114 |
try:
|
115 |
print "Using method %s" % name |
116 |
self._RecordMethodInvocation(name, kwargs)
|
117 |
result = fn(*args, **kwargs) |
118 |
except GanetiApiError as e: |
119 |
print "RAPI error while performing function %s : %s" % \ |
120 |
(name, str(e))
|
121 |
return result
|
122 |
|
123 |
return decoratedFn
|
124 |
|
125 |
def __getattr__(self, attr): |
126 |
""" Fetches an attribute from the underlying client if necessary.
|
127 |
|
128 |
"""
|
129 |
# Assuming that this method exposes no public methods of its own,
|
130 |
# and that any private methods are named according to the style
|
131 |
# guide, this will stop infinite loops in attribute fetches.
|
132 |
if attr.startswith("_"): |
133 |
return self.__getattribute__(attr) |
134 |
|
135 |
# We also want to expose non-methods
|
136 |
if hasattr(self._client, attr) and \ |
137 |
not isinstance(getattr(self._client, attr), types.MethodType): |
138 |
return getattr(self._client, attr) |
139 |
|
140 |
try:
|
141 |
return self._InvokerCreator(self._client.__getattribute__(attr), attr) |
142 |
except AttributeError: |
143 |
print "Missing method %s; supplying mock method" % attr |
144 |
return MockMethod
|
145 |
|
146 |
def _OutputMethodInvocationDetails(self): |
147 |
""" Attempts to output as much information as possible about the methods
|
148 |
that have and have not been invoked, including which arguments have not
|
149 |
been used.
|
150 |
|
151 |
"""
|
152 |
print "\nMethod usage:\n" |
153 |
for method in [n for n in dir(self._client) |
154 |
if not n.startswith('_') and |
155 |
isinstance(self.__getattr__(n), types.FunctionType)]: |
156 |
if method not in self._method_invocations: |
157 |
print "Method unused: %s" % method |
158 |
else:
|
159 |
arg_spec, _, _, default_arg_spec = \ |
160 |
inspect.getargspec(getattr(self._client, method)) |
161 |
default_args = [] |
162 |
if default_arg_spec is not None: |
163 |
default_args = arg_spec[-len(default_arg_spec):]
|
164 |
used_arg_set = self._method_invocations[method]
|
165 |
unused_args = [arg for arg in default_args if arg not in used_arg_set] |
166 |
if unused_args:
|
167 |
print "Method %s used, but arguments unused: %s" % \ |
168 |
(method, ", ".join(unused_args))
|
169 |
|
170 |
|
171 |
def Finish(client, fn, *args, **kwargs): |
172 |
""" When invoked with a job-starting RAPI client method, it passes along any
|
173 |
additional arguments and waits until its completion.
|
174 |
|
175 |
@type client C{GanetiRapiClientWrapper}
|
176 |
@param client The client wrapper.
|
177 |
@type fn function
|
178 |
@param fn A client method returning a job id.
|
179 |
|
180 |
@rtype tuple of bool, any object
|
181 |
@return The success status and the result of the operation, if any
|
182 |
|
183 |
"""
|
184 |
possible_job_id = fn(*args, **kwargs) |
185 |
try:
|
186 |
# The job ids are returned as both ints and ints represented by strings.
|
187 |
# This is a pythonic check to see if the content is an int.
|
188 |
int(possible_job_id)
|
189 |
except (ValueError, TypeError): |
190 |
# As a rule of thumb, failures will return None, and other methods are
|
191 |
# expected to return at least something
|
192 |
if possible_job_id is not None: |
193 |
print ("Finish called with a method not producing a job id, " |
194 |
"returning %s" % possible_job_id)
|
195 |
return possible_job_id
|
196 |
|
197 |
success = client.WaitForJobCompletion(possible_job_id) |
198 |
|
199 |
result = client.GetJobStatus(possible_job_id)["opresult"][0] |
200 |
if success:
|
201 |
return success, result
|
202 |
else:
|
203 |
print "Error encountered while performing operation: " |
204 |
print result
|
205 |
return success, None |
206 |
|
207 |
|
208 |
def TestTags(client, get_fn, add_fn, delete_fn, *args): |
209 |
""" Tests whether tagging works.
|
210 |
|
211 |
@type client C{GanetiRapiClientWrapper}
|
212 |
@param client The client wrapper.
|
213 |
@type get_fn function
|
214 |
@param get_fn A Get*Tags function of the client.
|
215 |
@type add_fn function
|
216 |
@param add_fn An Add*Tags function of the client.
|
217 |
@type delete_fn function
|
218 |
@param delete_fn A Delete*Tags function of the client.
|
219 |
|
220 |
To allow this method to work for all tagging functions of the client, use
|
221 |
named methods.
|
222 |
|
223 |
"""
|
224 |
get_fn(*args) |
225 |
|
226 |
tags = ["tag1", "tag2", "tag3"] |
227 |
Finish(client, add_fn, *args, tags=tags, dry_run=True)
|
228 |
Finish(client, add_fn, *args, tags=tags) |
229 |
|
230 |
get_fn(*args) |
231 |
|
232 |
Finish(client, delete_fn, *args, tags=tags[:1], dry_run=True) |
233 |
Finish(client, delete_fn, *args, tags=tags[:1])
|
234 |
|
235 |
get_fn(*args) |
236 |
|
237 |
Finish(client, delete_fn, *args, tags=tags[1:])
|
238 |
|
239 |
get_fn(*args) |
240 |
|
241 |
|
242 |
def TestGetters(client): |
243 |
""" Tests the various get functions which only retrieve information about the
|
244 |
cluster.
|
245 |
|
246 |
@type client C{GanetiRapiClientWrapper}
|
247 |
|
248 |
"""
|
249 |
client.GetVersion() |
250 |
client.GetFeatures() |
251 |
client.GetOperatingSystems() |
252 |
client.GetInfo() |
253 |
client.GetClusterTags() |
254 |
client.GetInstances() |
255 |
client.GetInstances(bulk=True)
|
256 |
client.GetJobs() |
257 |
client.GetJobs(bulk=True)
|
258 |
client.GetNodes() |
259 |
client.GetNodes(bulk=True)
|
260 |
client.GetNetworks() |
261 |
client.GetNetworks(bulk=True)
|
262 |
client.GetGroups() |
263 |
client.GetGroups(bulk=True)
|
264 |
|
265 |
|
266 |
def TestQueries(client, resource_name): |
267 |
""" Finds out which fields are present for a given resource type, and attempts
|
268 |
to retrieve their values for all present resources.
|
269 |
|
270 |
@type client C{GanetiRapiClientWrapper}
|
271 |
@param client A wrapped RAPI client.
|
272 |
@type resource_name string
|
273 |
@param resource_name The name of the resource to use.
|
274 |
|
275 |
"""
|
276 |
|
277 |
FIELDS_KEY = "fields"
|
278 |
|
279 |
query_res = client.QueryFields(resource_name) |
280 |
|
281 |
if query_res is None or FIELDS_KEY not in query_res or \ |
282 |
len(query_res[FIELDS_KEY]) == 0: |
283 |
return
|
284 |
|
285 |
field_entries = query_res[FIELDS_KEY] |
286 |
|
287 |
fields = map(lambda e: e["name"], field_entries) |
288 |
|
289 |
client.Query(resource_name, fields) |
290 |
|
291 |
|
292 |
def TestQueryFiltering(client, master_name): |
293 |
""" Performs queries by playing around with the only guaranteed resource, the
|
294 |
master node.
|
295 |
|
296 |
@type client C{GanetiRapiClientWrapper}
|
297 |
@param client A wrapped RAPI client.
|
298 |
@type master_name string
|
299 |
@param master_name The hostname of the master node.
|
300 |
|
301 |
"""
|
302 |
client.Query("node", ["name"], |
303 |
["|",
|
304 |
["=", "name", master_name], |
305 |
[">", "dtotal", 0], |
306 |
]) |
307 |
|
308 |
client.Query("instance", ["name"], |
309 |
["|",
|
310 |
["=", "name", "NonexistentInstance"], |
311 |
[">", "oper_ram", 0], |
312 |
]) |
313 |
|
314 |
|
315 |
def RemoveAllInstances(client): |
316 |
""" Queries for a list of instances, then removes them all.
|
317 |
|
318 |
@type client C{GanetiRapiClientWrapper}
|
319 |
@param client A wrapped RAPI client.
|
320 |
|
321 |
"""
|
322 |
instances = client.GetInstances() |
323 |
for inst in instances: |
324 |
Finish(client, client.DeleteInstance, inst) |
325 |
|
326 |
instances = client.GetInstances() |
327 |
assert len(instances) == 0 |
328 |
|
329 |
|
330 |
def TestSingleInstance(client, instance_name, alternate_name, node_one, |
331 |
node_two): |
332 |
""" Creates an instance, performs operations involving it, and then deletes
|
333 |
it.
|
334 |
|
335 |
@type client C{GanetiRapiClientWrapper}
|
336 |
@param client A wrapped RAPI client.
|
337 |
@type instance_name string
|
338 |
@param instance_name The hostname to use.
|
339 |
@type instance_name string
|
340 |
@param instance_name Another valid hostname to use.
|
341 |
@type node_one string
|
342 |
@param node_one A node on which an instance can be added.
|
343 |
@type node_two string
|
344 |
@param node_two A node on which an instance can be added.
|
345 |
|
346 |
"""
|
347 |
|
348 |
# Check that a dry run works, use string with size and unit
|
349 |
Finish(client, client.CreateInstance, |
350 |
"create", instance_name, "plain", [{"size":"1gb"}], [], dry_run=True, |
351 |
os="debian-image", pnode=node_one)
|
352 |
|
353 |
# Another dry run, numeric size, should work, but still a dry run
|
354 |
Finish(client, client.CreateInstance, |
355 |
"create", instance_name, "plain", [{"size": "1000"}], [{}], |
356 |
dry_run=True, os="debian-image", pnode=node_one) |
357 |
|
358 |
# Create a smaller instance, and delete it immediately
|
359 |
Finish(client, client.CreateInstance, |
360 |
"create", instance_name, "plain", [{"size":800}], [{}], |
361 |
os="debian-image", pnode=node_one)
|
362 |
|
363 |
Finish(client, client.DeleteInstance, instance_name) |
364 |
|
365 |
# Create one instance to use in further tests
|
366 |
Finish(client, client.CreateInstance, |
367 |
"create", instance_name, "plain", [{"size":1200}], [{}], |
368 |
os="debian-image", pnode=node_one)
|
369 |
|
370 |
client.GetInstance(instance_name) |
371 |
|
372 |
Finish(client, client.GetInstanceInfo, instance_name) |
373 |
|
374 |
Finish(client, client.GetInstanceInfo, instance_name, static=True)
|
375 |
|
376 |
TestQueries(client, "instance")
|
377 |
|
378 |
TestTags(client, client.GetInstanceTags, client.AddInstanceTags, |
379 |
client.DeleteInstanceTags, instance_name) |
380 |
|
381 |
Finish(client, client.GrowInstanceDisk, |
382 |
instance_name, 0, 100, wait_for_sync=True) |
383 |
|
384 |
Finish(client, client.RebootInstance, |
385 |
instance_name, "soft", ignore_secondaries=True, dry_run=True, |
386 |
reason="Hulk smash gently!")
|
387 |
|
388 |
Finish(client, client.ShutdownInstance, |
389 |
instance_name, dry_run=True, no_remember=False, |
390 |
reason="Hulk smash hard!")
|
391 |
|
392 |
Finish(client, client.StartupInstance, |
393 |
instance_name, dry_run=True, no_remember=False, |
394 |
reason="Not hard enough!")
|
395 |
|
396 |
Finish(client, client.RebootInstance, |
397 |
instance_name, "soft", ignore_secondaries=True, dry_run=False) |
398 |
|
399 |
Finish(client, client.ShutdownInstance, |
400 |
instance_name, dry_run=False, no_remember=False) |
401 |
|
402 |
Finish(client, client.ModifyInstance, |
403 |
instance_name, disk_template="drbd", remote_node=node_two)
|
404 |
|
405 |
Finish(client, client.ModifyInstance, |
406 |
instance_name, disk_template="plain")
|
407 |
|
408 |
Finish(client, client.RenameInstance, |
409 |
instance_name, alternate_name, ip_check=True, name_check=True) |
410 |
|
411 |
Finish(client, client.RenameInstance, alternate_name, instance_name) |
412 |
|
413 |
Finish(client, client.DeactivateInstanceDisks, instance_name) |
414 |
|
415 |
Finish(client, client.ActivateInstanceDisks, instance_name) |
416 |
|
417 |
# Note that the RecreateInstanceDisks command will always fail, as there is
|
418 |
# no way to induce the necessary prerequisites (removal of LV) via RAPI.
|
419 |
# Keeping it around allows us to at least know that it still exists.
|
420 |
Finish(client, client.RecreateInstanceDisks, |
421 |
instance_name, [0], [node_one])
|
422 |
|
423 |
Finish(client, client.StartupInstance, |
424 |
instance_name, dry_run=False, no_remember=False) |
425 |
|
426 |
client.GetInstanceConsole(instance_name) |
427 |
|
428 |
Finish(client, client.ReinstallInstance, |
429 |
instance_name, os=None, no_startup=False, osparams={}) |
430 |
|
431 |
Finish(client, client.DeleteInstance, instance_name, dry_run=True)
|
432 |
|
433 |
Finish(client, client.DeleteInstance, instance_name) |
434 |
|
435 |
|
436 |
def MarkUnmarkNode(client, node, state): |
437 |
""" Given a certain node state, marks a node as being in that state, and then
|
438 |
unmarks it.
|
439 |
|
440 |
@type client C{GanetiRapiClientWrapper}
|
441 |
@param client A wrapped RAPI client.
|
442 |
@type node string
|
443 |
@type state string
|
444 |
|
445 |
"""
|
446 |
# pylint: disable=W0142
|
447 |
Finish(client, client.ModifyNode, node, **{state: True})
|
448 |
Finish(client, client.ModifyNode, node, **{state: False})
|
449 |
# pylint: enable=W0142
|
450 |
|
451 |
|
452 |
def TestNodeOperations(client, non_master_node): |
453 |
""" Tests various operations related to nodes only
|
454 |
|
455 |
@type client C{GanetiRapiClientWrapper}
|
456 |
@param client A wrapped RAPI client.
|
457 |
@type non_master_node string
|
458 |
@param non_master_node The name of a non-master node in the cluster.
|
459 |
|
460 |
"""
|
461 |
|
462 |
client.GetNode(non_master_node) |
463 |
|
464 |
old_role = client.GetNodeRole(non_master_node) |
465 |
|
466 |
# Should fail
|
467 |
Finish(client, client.SetNodeRole, |
468 |
non_master_node, "master", False, auto_promote=True) |
469 |
|
470 |
Finish(client, client.SetNodeRole, |
471 |
non_master_node, "regular", False, auto_promote=True) |
472 |
|
473 |
Finish(client, client.SetNodeRole, |
474 |
non_master_node, "master-candidate", False, auto_promote=True) |
475 |
|
476 |
Finish(client, client.SetNodeRole, |
477 |
non_master_node, "drained", False, auto_promote=True) |
478 |
|
479 |
Finish(client, client.SetNodeRole, |
480 |
non_master_node, old_role, False, auto_promote=True) |
481 |
|
482 |
Finish(client, client.PowercycleNode, |
483 |
non_master_node, force=False)
|
484 |
|
485 |
storage_units_fields = [ |
486 |
"name", "allocatable", "free", "node", "size", "type", "used", |
487 |
] |
488 |
|
489 |
for storage_type in constants.STS_REPORT: |
490 |
success, storage_units = Finish(client, client.GetNodeStorageUnits, |
491 |
non_master_node, storage_type, |
492 |
",".join(storage_units_fields))
|
493 |
|
494 |
if success and len(storage_units) > 0 and len(storage_units[0]) > 0: |
495 |
# Name is the first entry of the first result, allocatable the other
|
496 |
unit_name = storage_units[0][0] |
497 |
Finish(client, client.ModifyNodeStorageUnits, |
498 |
non_master_node, storage_type, unit_name, |
499 |
allocatable=not storage_units[0][1]) |
500 |
Finish(client, client.ModifyNodeStorageUnits, |
501 |
non_master_node, storage_type, unit_name, |
502 |
allocatable=storage_units[0][1]) |
503 |
Finish(client, client.RepairNodeStorageUnits, |
504 |
non_master_node, storage_type, unit_name) |
505 |
|
506 |
MarkUnmarkNode(client, non_master_node, "drained")
|
507 |
MarkUnmarkNode(client, non_master_node, "powered")
|
508 |
MarkUnmarkNode(client, non_master_node, "offline")
|
509 |
|
510 |
TestQueries(client, "node")
|
511 |
|
512 |
|
513 |
def TestGroupOperations(client, node, another_node): |
514 |
""" Tests various operations related to groups only.
|
515 |
|
516 |
@type client C{GanetiRapiClientWrapper}
|
517 |
@param client A Ganeti RAPI client to use.
|
518 |
@type node string
|
519 |
@param node The name of a node in the cluster.
|
520 |
@type another_node string
|
521 |
@param another_node The name of another node in the cluster.
|
522 |
|
523 |
"""
|
524 |
|
525 |
DEFAULT_GROUP_NAME = constants.INITIAL_NODE_GROUP_NAME |
526 |
TEST_GROUP_NAME = "TestGroup"
|
527 |
ALTERNATE_GROUP_NAME = "RenamedTestGroup"
|
528 |
|
529 |
Finish(client, client.CreateGroup, |
530 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED, |
531 |
dry_run=True)
|
532 |
|
533 |
Finish(client, client.CreateGroup, |
534 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED) |
535 |
|
536 |
client.GetGroup(TEST_GROUP_NAME) |
537 |
|
538 |
TestQueries(client, "group")
|
539 |
|
540 |
TestTags(client, client.GetGroupTags, client.AddGroupTags, |
541 |
client.DeleteGroupTags, TEST_GROUP_NAME) |
542 |
|
543 |
Finish(client, client.ModifyGroup, |
544 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED, |
545 |
depends=None)
|
546 |
|
547 |
Finish(client, client.AssignGroupNodes, |
548 |
TEST_GROUP_NAME, [node, another_node], force=False, dry_run=True) |
549 |
|
550 |
Finish(client, client.AssignGroupNodes, |
551 |
TEST_GROUP_NAME, [another_node], force=False)
|
552 |
|
553 |
Finish(client, client.RenameGroup, |
554 |
TEST_GROUP_NAME, ALTERNATE_GROUP_NAME) |
555 |
|
556 |
Finish(client, client.RenameGroup, |
557 |
ALTERNATE_GROUP_NAME, TEST_GROUP_NAME) |
558 |
|
559 |
Finish(client, client.AssignGroupNodes, |
560 |
DEFAULT_GROUP_NAME, [another_node], force=False)
|
561 |
|
562 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME, dry_run=True)
|
563 |
|
564 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME) |
565 |
|
566 |
|
567 |
def TestNetworkConnectDisconnect(client, network_name, mode, link): |
568 |
""" Test connecting and disconnecting the network to a new node group.
|
569 |
|
570 |
@type network_name string
|
571 |
@param network_name The name of an existing and unconnected network.
|
572 |
@type mode string
|
573 |
@param mode The network mode.
|
574 |
@type link string
|
575 |
@param link The network link.
|
576 |
|
577 |
"""
|
578 |
# For testing the connect/disconnect calls, a group is needed
|
579 |
TEST_GROUP_NAME = "TestGroup"
|
580 |
Finish(client, client.CreateGroup, |
581 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED) |
582 |
|
583 |
Finish(client, client.ConnectNetwork, |
584 |
network_name, TEST_GROUP_NAME, mode, link, dry_run=True)
|
585 |
|
586 |
Finish(client, client.ConnectNetwork, |
587 |
network_name, TEST_GROUP_NAME, mode, link) |
588 |
|
589 |
Finish(client, client.DisconnectNetwork, |
590 |
network_name, TEST_GROUP_NAME, dry_run=True)
|
591 |
|
592 |
Finish(client, client.DisconnectNetwork, |
593 |
network_name, TEST_GROUP_NAME) |
594 |
|
595 |
# Clean up the group
|
596 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME) |
597 |
|
598 |
|
599 |
def TestNetworks(client): |
600 |
""" Add some networks of different sizes, using RFC5737 addresses like in the
|
601 |
QA.
|
602 |
|
603 |
"""
|
604 |
|
605 |
NETWORK_NAME = "SurelyCertainlyNonexistentNetwork"
|
606 |
|
607 |
Finish(client, client.CreateNetwork, |
608 |
NETWORK_NAME, "192.0.2.0/30", tags=[], dry_run=True) |
609 |
|
610 |
Finish(client, client.CreateNetwork, |
611 |
NETWORK_NAME, "192.0.2.0/30", tags=[])
|
612 |
|
613 |
client.GetNetwork(NETWORK_NAME) |
614 |
|
615 |
TestTags(client, client.GetNetworkTags, client.AddNetworkTags, |
616 |
client.DeleteNetworkTags, NETWORK_NAME) |
617 |
|
618 |
Finish(client, client.ModifyNetwork, |
619 |
NETWORK_NAME, mac_prefix=None)
|
620 |
|
621 |
TestQueries(client, "network")
|
622 |
|
623 |
default_nicparams = qa_config.get("default-nicparams", None) |
624 |
|
625 |
# The entry might not be present in the QA config
|
626 |
if default_nicparams is not None: |
627 |
mode = default_nicparams.get("mode", None) |
628 |
link = default_nicparams.get("link", None) |
629 |
if mode is not None and link is not None: |
630 |
TestNetworkConnectDisconnect(client, NETWORK_NAME, mode, link) |
631 |
|
632 |
# Clean up the network
|
633 |
Finish(client, client.DeleteNetwork, |
634 |
NETWORK_NAME, dry_run=True)
|
635 |
|
636 |
Finish(client, client.DeleteNetwork, NETWORK_NAME) |
637 |
|
638 |
|
639 |
def CreateDRBDInstance(client, node_one, node_two, instance_name): |
640 |
""" Creates a DRBD-enabled instance on the given nodes.
|
641 |
|
642 |
"""
|
643 |
Finish(client, client.CreateInstance, |
644 |
"create", instance_name, "drbd", [{"size": "1000"}], [{}], |
645 |
os="debian-image", pnode=node_one, snode=node_two)
|
646 |
|
647 |
|
648 |
def TestInstanceMigrations(client, node_one, node_two, node_three, |
649 |
instance_name): |
650 |
""" Test various operations related to migrating instances.
|
651 |
|
652 |
@type node_one string
|
653 |
@param node_one The name of a node in the cluster.
|
654 |
@type node_two string
|
655 |
@param node_two The name of another node in the cluster.
|
656 |
@type node_three string
|
657 |
@param node_three The name of yet another node in the cluster.
|
658 |
@type instance_name string
|
659 |
@param instance_name An instance name that can be used.
|
660 |
|
661 |
"""
|
662 |
|
663 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
664 |
Finish(client, client.FailoverInstance, instance_name) |
665 |
Finish(client, client.DeleteInstance, instance_name) |
666 |
|
667 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
668 |
Finish(client, client.EvacuateNode, |
669 |
node_two, early_release=False, mode=NODE_EVAC_SEC,
|
670 |
remote_node=node_three) |
671 |
Finish(client, client.DeleteInstance, instance_name) |
672 |
|
673 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
674 |
Finish(client, client.EvacuateNode, |
675 |
node_one, early_release=False, mode=NODE_EVAC_PRI, iallocator="hail") |
676 |
Finish(client, client.DeleteInstance, instance_name) |
677 |
|
678 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
679 |
Finish(client, client.MigrateInstance, |
680 |
instance_name, cleanup=True, target_node=node_two)
|
681 |
Finish(client, client.DeleteInstance, instance_name) |
682 |
|
683 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
684 |
Finish(client, client.MigrateNode, |
685 |
node_one, iallocator="hail", mode="non-live") |
686 |
Finish(client, client.DeleteInstance, instance_name) |
687 |
|
688 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
689 |
Finish(client, client.MigrateNode, |
690 |
node_one, target_node=node_two, mode="non-live")
|
691 |
Finish(client, client.DeleteInstance, instance_name) |
692 |
|
693 |
|
694 |
def ExtractAllNicInformationPossible(nics, replace_macs=True): |
695 |
""" Extracts NIC information as a dictionary.
|
696 |
|
697 |
@type nics list of tuples of varying structure
|
698 |
@param nics The network interfaces, as received from the instance info RAPI
|
699 |
call.
|
700 |
|
701 |
@rtype list of dict
|
702 |
@return Dictionaries of NIC information.
|
703 |
|
704 |
The NIC information is returned in a different format across versions, and to
|
705 |
try and see if the execution of commands is still compatible, this function
|
706 |
attempts to grab all the info that it can.
|
707 |
|
708 |
"""
|
709 |
|
710 |
desired_entries = [ |
711 |
constants.INIC_IP, |
712 |
constants.INIC_MAC, |
713 |
constants.INIC_MODE, |
714 |
constants.INIC_LINK, |
715 |
constants.INIC_VLAN, |
716 |
constants.INIC_NETWORK, |
717 |
constants.INIC_NAME, |
718 |
] |
719 |
|
720 |
nic_dicts = [] |
721 |
for nic_index in range(len(nics)): |
722 |
nic_raw_data = nics[nic_index] |
723 |
|
724 |
# Fill dictionary with None-s as defaults
|
725 |
nic_dict = dict([(key, None) for key in desired_entries]) |
726 |
|
727 |
try:
|
728 |
# The 2.6 format
|
729 |
ip, mac, mode, link = nic_raw_data |
730 |
except ValueError: |
731 |
# If there is yet another ValueError here, let it go through as it is
|
732 |
# legitimate - we are out of versions
|
733 |
|
734 |
# The 2.11 format
|
735 |
nic_name, _, ip, mac, mode, link, vlan, network, _ = nic_raw_data |
736 |
nic_dict[constants.INIC_VLAN] = vlan |
737 |
nic_dict[constants.INIC_NETWORK] = network |
738 |
nic_dict[constants.INIC_NAME] = nic_name |
739 |
|
740 |
# These attributes will be present in either version
|
741 |
nic_dict[constants.INIC_IP] = ip |
742 |
nic_dict[constants.INIC_MAC] = mac |
743 |
nic_dict[constants.INIC_MODE] = mode |
744 |
nic_dict[constants.INIC_LINK] = link |
745 |
|
746 |
# Very simple mac generation, which should work as the setup cluster should
|
747 |
# have no mac prefix restrictions in the default network, and there is a
|
748 |
# hard and reasonable limit of only 8 NICs
|
749 |
if replace_macs:
|
750 |
nic_dict[constants.INIC_MAC] = "00:00:00:00:00:%02x" % nic_index
|
751 |
|
752 |
nic_dicts.append(nic_dict) |
753 |
|
754 |
return nic_dicts
|
755 |
|
756 |
|
757 |
def MoveInstance(client, src_instance, dst_instance, src_node, dst_node): |
758 |
""" Moves a single instance, compatible with 2.6.
|
759 |
|
760 |
@rtype bool
|
761 |
@return Whether the instance was moved successfully
|
762 |
|
763 |
"""
|
764 |
success, inst_info_all = Finish(client, client.GetInstanceInfo, |
765 |
src_instance.name, static=True)
|
766 |
|
767 |
if not success or src_instance.name not in inst_info_all: |
768 |
raise Exception("Did not find the source instance information!") |
769 |
|
770 |
inst_info = inst_info_all[src_instance.name] |
771 |
|
772 |
# Try to extract NICs first, as this is the operation most likely to fail
|
773 |
try:
|
774 |
nic_info = ExtractAllNicInformationPossible(inst_info["nics"])
|
775 |
except ValueError: |
776 |
# Without the NIC info, there is very little we can do
|
777 |
return False |
778 |
|
779 |
NIC_COMPONENTS_26 = [ |
780 |
constants.INIC_IP, |
781 |
constants.INIC_MAC, |
782 |
constants.INIC_MODE, |
783 |
constants.INIC_LINK, |
784 |
] |
785 |
|
786 |
nic_converter = lambda old: dict((k, old[k]) for k in NIC_COMPONENTS_26) |
787 |
nics = map(nic_converter, nic_info)
|
788 |
|
789 |
# Prepare the parameters
|
790 |
disks = [] |
791 |
for idisk in inst_info["disks"]: |
792 |
odisk = { |
793 |
constants.IDISK_SIZE: idisk["size"],
|
794 |
constants.IDISK_MODE: idisk["mode"],
|
795 |
} |
796 |
|
797 |
spindles = idisk.get("spindles")
|
798 |
if spindles is not None: |
799 |
odisk[constants.IDISK_SPINDLES] = spindles |
800 |
|
801 |
# Disk name may be present, but must not be supplied in 2.6!
|
802 |
disks.append(odisk) |
803 |
|
804 |
# With all the parameters properly prepared, try the export
|
805 |
success, exp_info = Finish(client, client.PrepareExport, |
806 |
src_instance.name, constants.EXPORT_MODE_REMOTE) |
807 |
|
808 |
if not success: |
809 |
# The instance will still have to be deleted
|
810 |
return False |
811 |
|
812 |
success, _ = Finish(client, client.CreateInstance, |
813 |
constants.INSTANCE_REMOTE_IMPORT, dst_instance.name, |
814 |
inst_info["disk_template"], disks, nics,
|
815 |
os=inst_info["os"],
|
816 |
pnode=dst_node.primary, |
817 |
snode=src_node.primary, # Ignored as no DRBD
|
818 |
start=(inst_info["config_state"] == "up"), |
819 |
ip_check=False,
|
820 |
iallocator=inst_info.get("iallocator", None), |
821 |
hypervisor=inst_info["hypervisor"],
|
822 |
source_handshake=exp_info["handshake"],
|
823 |
source_x509_ca=exp_info["x509_ca"],
|
824 |
source_instance_name=inst_info["name"],
|
825 |
beparams=inst_info["be_instance"],
|
826 |
hvparams=inst_info["hv_instance"],
|
827 |
osparams=inst_info["os_instance"])
|
828 |
|
829 |
return success
|
830 |
|
831 |
|
832 |
def CreateInstanceForMoveTest(client, node, instance): |
833 |
""" Creates a single shutdown instance to move about in tests.
|
834 |
|
835 |
@type node C{_QaNode}
|
836 |
@param node A node configuration object.
|
837 |
@type instance C{_QaInstance}
|
838 |
@param instance An instance configuration object.
|
839 |
|
840 |
"""
|
841 |
Finish(client, client.CreateInstance, |
842 |
"create", instance.name, "plain", [{"size": "2000"}], [{}], |
843 |
os="debian-image", pnode=node.primary)
|
844 |
|
845 |
Finish(client, client.ShutdownInstance, |
846 |
instance.name, dry_run=False, no_remember=False) |
847 |
|
848 |
|
849 |
def Test26InstanceMove(client, node_one, node_two, instance_to_create, |
850 |
new_instance): |
851 |
""" Tests instance moves using commands that work in 2.6.
|
852 |
|
853 |
"""
|
854 |
|
855 |
# First create the instance to move
|
856 |
CreateInstanceForMoveTest(client, node_one, instance_to_create) |
857 |
|
858 |
# The cleanup should be conditional on operation success
|
859 |
if MoveInstance(client, instance_to_create, new_instance, node_one, node_two):
|
860 |
Finish(client, client.DeleteInstance, new_instance.name) |
861 |
else:
|
862 |
Finish(client, client.DeleteInstance, instance_to_create.name) |
863 |
|
864 |
|
865 |
def Test211InstanceMove(client, node_one, node_two, instance_to_create, |
866 |
new_instance): |
867 |
""" Tests instance moves using the QA-provided move test.
|
868 |
|
869 |
"""
|
870 |
|
871 |
# First create the instance to move
|
872 |
CreateInstanceForMoveTest(client, node_one, instance_to_create) |
873 |
|
874 |
instance_to_create.SetDiskTemplate("plain")
|
875 |
|
876 |
try:
|
877 |
qa_rapi.TestInterClusterInstanceMove(instance_to_create, new_instance, |
878 |
[node_one], node_two, |
879 |
perform_checks=False)
|
880 |
except qa_error.Error:
|
881 |
# A failure is sad, but requires no special actions to be undertaken
|
882 |
pass
|
883 |
|
884 |
# Try to delete the instance when done - either the move has failed, or
|
885 |
# a double move was performed - the instance to delete is one and the same
|
886 |
Finish(client, client.DeleteInstance, instance_to_create.name) |
887 |
|
888 |
|
889 |
def TestInstanceMoves(client, node_one, node_two, instance_to_create, |
890 |
new_instance): |
891 |
""" Performs two types of instance moves, one compatible with 2.6, the other
|
892 |
with 2.11.
|
893 |
|
894 |
@type node_one C{_QaNode}
|
895 |
@param node_one A node configuration object.
|
896 |
@type node_two C{_QaNode}
|
897 |
@param node_two A node configuration object.
|
898 |
@type instance_to_create C{_QaInstance}
|
899 |
@param instance_to_create An instance configuration object.
|
900 |
@type new_instance C{_QaInstance}
|
901 |
@param new_instance An instance configuration object.
|
902 |
|
903 |
"""
|
904 |
|
905 |
Test26InstanceMove(client, node_one, node_two, instance_to_create, |
906 |
new_instance) |
907 |
Test211InstanceMove(client, node_one, node_two, instance_to_create, |
908 |
new_instance) |
909 |
|
910 |
|
911 |
def TestClusterParameterModification(client): |
912 |
""" Try to modify some of the cluster parameters using RAPI.
|
913 |
|
914 |
"""
|
915 |
cluster_info = client.GetInfo() |
916 |
|
917 |
# Each attribute has several safe choices we can use
|
918 |
ATTRIBUTES_TO_MODIFY = [ |
919 |
("default_iallocator", ["hail", ""]), # Use "" to reset |
920 |
("candidate_pool_size", [1, 5]), |
921 |
("maintain_node_health", [True, False]), |
922 |
] |
923 |
|
924 |
for attribute, options in ATTRIBUTES_TO_MODIFY: |
925 |
current_value = cluster_info[attribute] |
926 |
|
927 |
if current_value in options: |
928 |
value_to_use = options[1 - options.index(current_value)]
|
929 |
else:
|
930 |
value_to_use = options[0]
|
931 |
|
932 |
#pylint: disable=W0142
|
933 |
Finish(client, client.ModifyCluster, **{attribute: value_to_use}) |
934 |
Finish(client, client.ModifyCluster, **{attribute: current_value}) |
935 |
#pylint: enable=W0142
|
936 |
|
937 |
|
938 |
def TestJobCancellation(client, node_one, node_two, instance_one, instance_two): |
939 |
""" Test if jobs can be cancelled.
|
940 |
|
941 |
@type node_one string
|
942 |
@param node_one The name of a node in the cluster.
|
943 |
@type node_two string
|
944 |
@param node_two The name of a node in the cluster.
|
945 |
@type instance_one string
|
946 |
@param instance_one An available instance name.
|
947 |
@type instance_two string
|
948 |
@param instance_two An available instance name.
|
949 |
|
950 |
"""
|
951 |
|
952 |
# Just in case, remove all previously present instances
|
953 |
RemoveAllInstances(client) |
954 |
|
955 |
# Let us issue a job that is sure to both succeed and last for a while
|
956 |
running_job = client.CreateInstance("create", instance_one, "drbd", |
957 |
[{"size": "5000"}], [{}], |
958 |
os="debian-image", pnode=node_one,
|
959 |
snode=node_two) |
960 |
|
961 |
# And immediately afterwards, another very similar one
|
962 |
job_to_cancel = client.CreateInstance("create", instance_two, "drbd", |
963 |
[{"size": "5000"}], [{}], |
964 |
os="debian-image", pnode=node_one,
|
965 |
snode=node_two) |
966 |
|
967 |
# Try to cancel, which should fail as the job is already running
|
968 |
success, msg = client.CancelJob(running_job) |
969 |
if success:
|
970 |
print "Job succeeded: this should not have happened as it is running!" |
971 |
print "Message: %s" % msg |
972 |
|
973 |
success, msg = client.CancelJob(job_to_cancel) |
974 |
if not success: |
975 |
print "Job failed: this was unexpected as it was not a dry run" |
976 |
print "Message: %s" % msg |
977 |
|
978 |
# And wait for the proper job
|
979 |
client.WaitForJobCompletion(running_job) |
980 |
|
981 |
# Remove all the leftover instances, success or no success
|
982 |
RemoveAllInstances(client) |
983 |
|
984 |
|
985 |
def Workload(client): |
986 |
""" The actual RAPI workload used for tests.
|
987 |
|
988 |
@type client C{GanetiRapiClientWrapper}
|
989 |
@param client A wrapped RAPI client.
|
990 |
|
991 |
"""
|
992 |
|
993 |
# First just the simple information retrievals
|
994 |
TestGetters(client) |
995 |
|
996 |
# Then the only remaining function which is parameter-free
|
997 |
Finish(client, client.RedistributeConfig) |
998 |
|
999 |
# Try changing the cluster parameters
|
1000 |
TestClusterParameterModification(client) |
1001 |
|
1002 |
TestTags(client, client.GetClusterTags, client.AddClusterTags, |
1003 |
client.DeleteClusterTags) |
1004 |
|
1005 |
# Generously assume the master is present
|
1006 |
node = qa_config.AcquireNode() |
1007 |
TestTags(client, client.GetNodeTags, client.AddNodeTags, |
1008 |
client.DeleteNodeTags, node.primary) |
1009 |
node.Release() |
1010 |
|
1011 |
# Instance tests
|
1012 |
|
1013 |
# First remove all instances the QA might have created
|
1014 |
RemoveAllInstances(client) |
1015 |
|
1016 |
nodes = qa_config.AcquireManyNodes(2)
|
1017 |
instances = qa_config.AcquireManyInstances(2)
|
1018 |
TestSingleInstance(client, instances[0].name, instances[1].name, |
1019 |
nodes[0].primary, nodes[1].primary) |
1020 |
qa_config.ReleaseManyInstances(instances) |
1021 |
qa_config.ReleaseManyNodes(nodes) |
1022 |
|
1023 |
# Test all the queries which involve resources that do not have functions
|
1024 |
# of their own
|
1025 |
TestQueries(client, "lock")
|
1026 |
TestQueries(client, "job")
|
1027 |
TestQueries(client, "export")
|
1028 |
|
1029 |
node = qa_config.AcquireNode(exclude=qa_config.GetMasterNode()) |
1030 |
TestNodeOperations(client, node.primary) |
1031 |
TestQueryFiltering(client, node.primary) |
1032 |
node.Release() |
1033 |
|
1034 |
nodes = qa_config.AcquireManyNodes(2)
|
1035 |
TestGroupOperations(client, nodes[0].primary, nodes[1].primary) |
1036 |
qa_config.ReleaseManyNodes(nodes) |
1037 |
|
1038 |
TestNetworks(client) |
1039 |
|
1040 |
nodes = qa_config.AcquireManyNodes(3)
|
1041 |
instance = qa_config.AcquireInstance() |
1042 |
TestInstanceMigrations(client, nodes[0].primary, nodes[1].primary, |
1043 |
nodes[2].primary, instance.name)
|
1044 |
instance.Release() |
1045 |
qa_config.ReleaseManyNodes(nodes) |
1046 |
|
1047 |
nodes = qa_config.AcquireManyNodes(2)
|
1048 |
instances = qa_config.AcquireManyInstances(2)
|
1049 |
TestInstanceMoves(client, nodes[0], nodes[1], instances[0], instances[1]) |
1050 |
TestJobCancellation(client, nodes[0].primary, nodes[1].primary, |
1051 |
instances[0].name, instances[1].name) |
1052 |
qa_config.ReleaseManyInstances(instances) |
1053 |
qa_config.ReleaseManyNodes(nodes) |
1054 |
|
1055 |
|
1056 |
def Usage(): |
1057 |
sys.stderr.write("Usage:\n\trapi-workload.py qa-config-file")
|
1058 |
|
1059 |
|
1060 |
def Main(): |
1061 |
if len(sys.argv) < 2: |
1062 |
Usage() |
1063 |
|
1064 |
qa_config.Load(sys.argv[1])
|
1065 |
|
1066 |
# Only the master will be present after a fresh QA cluster setup, so we have
|
1067 |
# to invoke this to get all the other nodes.
|
1068 |
qa_node.TestNodeAddAll() |
1069 |
|
1070 |
client = GanetiRapiClientWrapper() |
1071 |
|
1072 |
Workload(client) |
1073 |
|
1074 |
qa_node.TestNodeRemoveAll() |
1075 |
|
1076 |
# The method invoked has the naming of the protected method, and pylint does
|
1077 |
# not like this. Disabling the warning is healthier than explicitly adding and
|
1078 |
# maintaining an exception for this method in the wrapper.
|
1079 |
# pylint: disable=W0212
|
1080 |
client._OutputMethodInvocationDetails() |
1081 |
# pylint: enable=W0212
|
1082 |
|
1083 |
if __name__ == "__main__": |
1084 |
Main() |