root / qa / rapi-workload.py @ 3eea40a0
History | View | Annotate | Download (35.4 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 optparse |
31 |
import sys |
32 |
import types |
33 |
|
34 |
import ganeti.constants as constants |
35 |
from ganeti.rapi.client import GanetiApiError, NODE_EVAC_PRI, NODE_EVAC_SEC |
36 |
|
37 |
import qa_config |
38 |
import qa_error |
39 |
import qa_node |
40 |
import qa_rapi |
41 |
|
42 |
|
43 |
# The purpose of this file is to provide a stable and extensive RAPI workload
|
44 |
# that manipulates the cluster only using RAPI commands, with the assumption
|
45 |
# that an empty cluster was set up beforehand. All the nodes that can be added
|
46 |
# to the cluster should be a part of it, and no instances should be present.
|
47 |
#
|
48 |
# Its intended use is in RAPI compatibility tests, where different versions with
|
49 |
# possibly vastly different QAs must be compared. Running the QA on both
|
50 |
# versions of the cluster will produce RAPI calls, but there is no guarantee
|
51 |
# that they will match, or that functions invoked in between will not change the
|
52 |
# results.
|
53 |
#
|
54 |
# By using only RAPI functions, we are sure to be able to capture and log all
|
55 |
# the changes in cluster state, and be able to compare them afterwards.
|
56 |
#
|
57 |
# The functionality of the QA is still used to generate a functioning,
|
58 |
# RAPI-enabled cluster, and to set up a C{GanetiRapiClient} capable of issuing
|
59 |
# commands to the cluster.
|
60 |
#
|
61 |
# Due to the fact that not all calls issued as a part of the workload might be
|
62 |
# implemented in the different versions of Ganeti, the client does not halt or
|
63 |
# produce a non-zero exit code upon encountering a RAPI error. Instead, it
|
64 |
# reports it and moves on. Any utility comparing the requests should account for
|
65 |
# this.
|
66 |
|
67 |
|
68 |
def MockMethod(*_args, **_kwargs): |
69 |
""" Absorbs all arguments, does nothing, returns None.
|
70 |
|
71 |
"""
|
72 |
return None |
73 |
|
74 |
|
75 |
RAPI_USERNAME = "ganeti-qa"
|
76 |
|
77 |
|
78 |
class GanetiRapiClientWrapper(object): |
79 |
""" Creates and initializes a GanetiRapiClient, and acts as a wrapper invoking
|
80 |
only the methods that the version of the client actually uses.
|
81 |
|
82 |
"""
|
83 |
def __init__(self): |
84 |
self._client = qa_rapi.Setup(RAPI_USERNAME,
|
85 |
qa_rapi.LookupRapiSecret(RAPI_USERNAME)) |
86 |
|
87 |
self._method_invocations = {}
|
88 |
|
89 |
def _RecordMethodInvocation(self, name, arg_dict): |
90 |
""" Records the invocation of a C{GanetiRAPIClient} method, noting the
|
91 |
argument and the method names.
|
92 |
|
93 |
"""
|
94 |
if name not in self._method_invocations: |
95 |
self._method_invocations[name] = set() |
96 |
|
97 |
for named_arg in arg_dict: |
98 |
self._method_invocations[name].add(named_arg)
|
99 |
|
100 |
def _InvokerCreator(self, fn, name): |
101 |
""" Returns an invoker function that will invoke the given function
|
102 |
with any arguments passed to the invoker at a later time, while
|
103 |
catching any specific non-fatal errors we would like to know more
|
104 |
about.
|
105 |
|
106 |
@type fn: arbitrary function
|
107 |
@param fn: The function to invoke later.
|
108 |
@type name: string
|
109 |
@param name: The name of the function, for debugging purposes.
|
110 |
@rtype: function
|
111 |
|
112 |
"""
|
113 |
def decoratedFn(*args, **kwargs): |
114 |
result = None
|
115 |
try:
|
116 |
print "Using method %s" % name |
117 |
self._RecordMethodInvocation(name, kwargs)
|
118 |
result = fn(*args, **kwargs) |
119 |
except GanetiApiError as e: |
120 |
print "RAPI error while performing function %s : %s" % \ |
121 |
(name, str(e))
|
122 |
return result
|
123 |
|
124 |
return decoratedFn
|
125 |
|
126 |
def __getattr__(self, attr): |
127 |
""" Fetches an attribute from the underlying client if necessary.
|
128 |
|
129 |
"""
|
130 |
# Assuming that this method exposes no public methods of its own,
|
131 |
# and that any private methods are named according to the style
|
132 |
# guide, this will stop infinite loops in attribute fetches.
|
133 |
if attr.startswith("_"): |
134 |
return self.__getattribute__(attr) |
135 |
|
136 |
# We also want to expose non-methods
|
137 |
if hasattr(self._client, attr) and \ |
138 |
not isinstance(getattr(self._client, attr), types.MethodType): |
139 |
return getattr(self._client, attr) |
140 |
|
141 |
try:
|
142 |
return self._InvokerCreator(self._client.__getattribute__(attr), attr) |
143 |
except AttributeError: |
144 |
print "Missing method %s; supplying mock method" % attr |
145 |
return MockMethod
|
146 |
|
147 |
def _OutputMethodInvocationDetails(self): |
148 |
""" Attempts to output as much information as possible about the methods
|
149 |
that have and have not been invoked, including which arguments have not
|
150 |
been used.
|
151 |
|
152 |
"""
|
153 |
print "\nMethod usage:\n" |
154 |
for method in [n for n in dir(self._client) |
155 |
if not n.startswith('_') and |
156 |
isinstance(self.__getattr__(n), types.FunctionType)]: |
157 |
if method not in self._method_invocations: |
158 |
print "Method unused: %s" % method |
159 |
else:
|
160 |
arg_spec, _, _, default_arg_spec = \ |
161 |
inspect.getargspec(getattr(self._client, method)) |
162 |
default_args = [] |
163 |
if default_arg_spec is not None: |
164 |
default_args = arg_spec[-len(default_arg_spec):]
|
165 |
used_arg_set = self._method_invocations[method]
|
166 |
unused_args = [arg for arg in default_args if arg not in used_arg_set] |
167 |
if unused_args:
|
168 |
print "Method %s used, but arguments unused: %s" % \ |
169 |
(method, ", ".join(unused_args))
|
170 |
|
171 |
|
172 |
def Finish(client, fn, *args, **kwargs): |
173 |
""" When invoked with a job-starting RAPI client method, it passes along any
|
174 |
additional arguments and waits until its completion.
|
175 |
|
176 |
@type client: C{GanetiRapiClientWrapper}
|
177 |
@param client: The client wrapper.
|
178 |
@type fn: function
|
179 |
@param fn: A client method returning a job id.
|
180 |
|
181 |
@rtype: tuple of bool, any object
|
182 |
@return: The success status and the result of the operation, if any
|
183 |
|
184 |
"""
|
185 |
possible_job_id = fn(*args, **kwargs) |
186 |
try:
|
187 |
# The job ids are returned as both ints and ints represented by strings.
|
188 |
# This is a pythonic check to see if the content is an int.
|
189 |
int(possible_job_id)
|
190 |
except (ValueError, TypeError): |
191 |
# As a rule of thumb, failures will return None, and other methods are
|
192 |
# expected to return at least something
|
193 |
if possible_job_id is not None: |
194 |
print ("Finish called with a method not producing a job id, " |
195 |
"returning %s" % possible_job_id)
|
196 |
return possible_job_id
|
197 |
|
198 |
success = client.WaitForJobCompletion(possible_job_id) |
199 |
|
200 |
result = client.GetJobStatus(possible_job_id)["opresult"][0] |
201 |
if success:
|
202 |
return success, result
|
203 |
else:
|
204 |
print "Error encountered while performing operation: " |
205 |
print result
|
206 |
return success, None |
207 |
|
208 |
|
209 |
def TestTags(client, get_fn, add_fn, delete_fn, *args): |
210 |
""" Tests whether tagging works.
|
211 |
|
212 |
@type client: C{GanetiRapiClientWrapper}
|
213 |
@param client: The client wrapper.
|
214 |
@type get_fn: function
|
215 |
@param get_fn: A Get*Tags function of the client.
|
216 |
@type add_fn: function
|
217 |
@param add_fn: An Add*Tags function of the client.
|
218 |
@type delete_fn: function
|
219 |
@param delete_fn: A Delete*Tags function of the client.
|
220 |
|
221 |
To allow this method to work for all tagging functions of the client, use
|
222 |
named methods.
|
223 |
|
224 |
"""
|
225 |
get_fn(*args) |
226 |
|
227 |
tags = ["tag1", "tag2", "tag3"] |
228 |
Finish(client, add_fn, *args, tags=tags, dry_run=True)
|
229 |
Finish(client, add_fn, *args, tags=tags) |
230 |
|
231 |
get_fn(*args) |
232 |
|
233 |
Finish(client, delete_fn, *args, tags=tags[:1], dry_run=True) |
234 |
Finish(client, delete_fn, *args, tags=tags[:1])
|
235 |
|
236 |
get_fn(*args) |
237 |
|
238 |
Finish(client, delete_fn, *args, tags=tags[1:])
|
239 |
|
240 |
get_fn(*args) |
241 |
|
242 |
|
243 |
def TestGetters(client): |
244 |
""" Tests the various get functions which only retrieve information about the
|
245 |
cluster.
|
246 |
|
247 |
@type client: C{GanetiRapiClientWrapper}
|
248 |
|
249 |
"""
|
250 |
client.GetVersion() |
251 |
client.GetFeatures() |
252 |
client.GetOperatingSystems() |
253 |
client.GetInfo() |
254 |
client.GetClusterTags() |
255 |
client.GetInstances() |
256 |
client.GetInstances(bulk=True)
|
257 |
client.GetJobs() |
258 |
client.GetJobs(bulk=True)
|
259 |
client.GetNodes() |
260 |
client.GetNodes(bulk=True)
|
261 |
client.GetNetworks() |
262 |
client.GetNetworks(bulk=True)
|
263 |
client.GetGroups() |
264 |
client.GetGroups(bulk=True)
|
265 |
|
266 |
|
267 |
def TestQueries(client, resource_name): |
268 |
""" Finds out which fields are present for a given resource type, and attempts
|
269 |
to retrieve their values for all present resources.
|
270 |
|
271 |
@type client: C{GanetiRapiClientWrapper}
|
272 |
@param client: A wrapped RAPI client.
|
273 |
@type resource_name: string
|
274 |
@param resource_name: The name of the resource to use.
|
275 |
|
276 |
"""
|
277 |
|
278 |
FIELDS_KEY = "fields"
|
279 |
|
280 |
query_res = client.QueryFields(resource_name) |
281 |
|
282 |
if query_res is None or FIELDS_KEY not in query_res or \ |
283 |
len(query_res[FIELDS_KEY]) == 0: |
284 |
return
|
285 |
|
286 |
field_entries = query_res[FIELDS_KEY] |
287 |
|
288 |
fields = map(lambda e: e["name"], field_entries) |
289 |
|
290 |
client.Query(resource_name, fields) |
291 |
|
292 |
|
293 |
def TestQueryFiltering(client, master_name): |
294 |
""" Performs queries by playing around with the only guaranteed resource, the
|
295 |
master node.
|
296 |
|
297 |
@type client: C{GanetiRapiClientWrapper}
|
298 |
@param client: A wrapped RAPI client.
|
299 |
@type master_name: string
|
300 |
@param master_name: The hostname of the master node.
|
301 |
|
302 |
"""
|
303 |
client.Query("node", ["name"], |
304 |
["|",
|
305 |
["=", "name", master_name], |
306 |
[">", "dtotal", 0], |
307 |
]) |
308 |
|
309 |
client.Query("instance", ["name"], |
310 |
["|",
|
311 |
["=", "name", "NonexistentInstance"], |
312 |
[">", "oper_ram", 0], |
313 |
]) |
314 |
|
315 |
|
316 |
def RemoveAllInstances(client): |
317 |
""" Queries for a list of instances, then removes them all.
|
318 |
|
319 |
@type client: C{GanetiRapiClientWrapper}
|
320 |
@param client: A wrapped RAPI client.
|
321 |
|
322 |
"""
|
323 |
instances = client.GetInstances() |
324 |
for inst in instances: |
325 |
Finish(client, client.DeleteInstance, inst) |
326 |
|
327 |
instances = client.GetInstances() |
328 |
assert len(instances) == 0 |
329 |
|
330 |
|
331 |
def TestSingleInstance(client, instance_name, alternate_name, node_one, |
332 |
node_two): |
333 |
""" Creates an instance, performs operations involving it, and then deletes
|
334 |
it.
|
335 |
|
336 |
@type client: C{GanetiRapiClientWrapper}
|
337 |
@param client: A wrapped RAPI client.
|
338 |
@type instance_name: string
|
339 |
@param instance_name: The hostname to use.
|
340 |
@type instance_name: string
|
341 |
@param instance_name: Another valid hostname to use.
|
342 |
@type node_one: string
|
343 |
@param node_one: A node on which an instance can be added.
|
344 |
@type node_two: string
|
345 |
@param node_two: Another node on which an instance can be added.
|
346 |
|
347 |
"""
|
348 |
|
349 |
# Check that a dry run works, use string with size and unit
|
350 |
Finish(client, client.CreateInstance, |
351 |
"create", instance_name, "plain", [{"size":"1gb"}], [], dry_run=True, |
352 |
os="debian-image", pnode=node_one)
|
353 |
|
354 |
# Another dry run, numeric size, should work, but still a dry run
|
355 |
Finish(client, client.CreateInstance, |
356 |
"create", instance_name, "plain", [{"size": "1000"}], [{}], |
357 |
dry_run=True, os="debian-image", pnode=node_one) |
358 |
|
359 |
# Create a smaller instance, and delete it immediately
|
360 |
Finish(client, client.CreateInstance, |
361 |
"create", instance_name, "plain", [{"size":800}], [{}], |
362 |
os="debian-image", pnode=node_one)
|
363 |
|
364 |
Finish(client, client.DeleteInstance, instance_name) |
365 |
|
366 |
# Create one instance to use in further tests
|
367 |
Finish(client, client.CreateInstance, |
368 |
"create", instance_name, "plain", [{"size":1200}], [{}], |
369 |
os="debian-image", pnode=node_one)
|
370 |
|
371 |
client.GetInstance(instance_name) |
372 |
|
373 |
Finish(client, client.GetInstanceInfo, instance_name) |
374 |
|
375 |
Finish(client, client.GetInstanceInfo, instance_name, static=True)
|
376 |
|
377 |
TestQueries(client, "instance")
|
378 |
|
379 |
TestTags(client, client.GetInstanceTags, client.AddInstanceTags, |
380 |
client.DeleteInstanceTags, instance_name) |
381 |
|
382 |
Finish(client, client.GrowInstanceDisk, |
383 |
instance_name, 0, 100, wait_for_sync=True) |
384 |
|
385 |
Finish(client, client.RebootInstance, |
386 |
instance_name, "soft", ignore_secondaries=True, dry_run=True, |
387 |
reason="Hulk smash gently!")
|
388 |
|
389 |
Finish(client, client.ShutdownInstance, |
390 |
instance_name, dry_run=True, no_remember=False, |
391 |
reason="Hulk smash hard!")
|
392 |
|
393 |
Finish(client, client.StartupInstance, |
394 |
instance_name, dry_run=True, no_remember=False, |
395 |
reason="Not hard enough!")
|
396 |
|
397 |
Finish(client, client.RebootInstance, |
398 |
instance_name, "soft", ignore_secondaries=True, dry_run=False) |
399 |
|
400 |
Finish(client, client.ShutdownInstance, |
401 |
instance_name, dry_run=False, no_remember=False) |
402 |
|
403 |
Finish(client, client.ModifyInstance, |
404 |
instance_name, disk_template="drbd", remote_node=node_two)
|
405 |
|
406 |
Finish(client, client.ModifyInstance, |
407 |
instance_name, disk_template="plain")
|
408 |
|
409 |
Finish(client, client.RenameInstance, |
410 |
instance_name, alternate_name, ip_check=True, name_check=True) |
411 |
|
412 |
Finish(client, client.RenameInstance, alternate_name, instance_name) |
413 |
|
414 |
Finish(client, client.DeactivateInstanceDisks, instance_name) |
415 |
|
416 |
Finish(client, client.ActivateInstanceDisks, instance_name) |
417 |
|
418 |
# Note that the RecreateInstanceDisks command will always fail, as there is
|
419 |
# no way to induce the necessary prerequisites (removal of LV) via RAPI.
|
420 |
# Keeping it around allows us to at least know that it still exists.
|
421 |
Finish(client, client.RecreateInstanceDisks, |
422 |
instance_name, [0], [node_one])
|
423 |
|
424 |
Finish(client, client.StartupInstance, |
425 |
instance_name, dry_run=False, no_remember=False) |
426 |
|
427 |
client.GetInstanceConsole(instance_name) |
428 |
|
429 |
Finish(client, client.ReinstallInstance, |
430 |
instance_name, os=None, no_startup=False, osparams={}) |
431 |
|
432 |
Finish(client, client.DeleteInstance, instance_name, dry_run=True)
|
433 |
|
434 |
Finish(client, client.DeleteInstance, instance_name) |
435 |
|
436 |
|
437 |
def MarkUnmarkNode(client, node, state): |
438 |
""" Given a certain node state, marks a node as being in that state, and then
|
439 |
unmarks it.
|
440 |
|
441 |
@type client: C{GanetiRapiClientWrapper}
|
442 |
@param client: A wrapped RAPI client.
|
443 |
@type node: string
|
444 |
@type state: string
|
445 |
|
446 |
"""
|
447 |
# pylint: disable=W0142
|
448 |
Finish(client, client.ModifyNode, node, **{state: True})
|
449 |
Finish(client, client.ModifyNode, node, **{state: False})
|
450 |
# pylint: enable=W0142
|
451 |
|
452 |
|
453 |
def TestNodeOperations(client, non_master_node): |
454 |
""" Tests various operations related to nodes only
|
455 |
|
456 |
@type client: C{GanetiRapiClientWrapper}
|
457 |
@param client: A wrapped RAPI client.
|
458 |
@type non_master_node: string
|
459 |
@param non_master_node: The name of a non-master node in the cluster.
|
460 |
|
461 |
"""
|
462 |
|
463 |
client.GetNode(non_master_node) |
464 |
|
465 |
old_role = client.GetNodeRole(non_master_node) |
466 |
|
467 |
# Should fail
|
468 |
Finish(client, client.SetNodeRole, |
469 |
non_master_node, "master", False, auto_promote=True) |
470 |
|
471 |
Finish(client, client.SetNodeRole, |
472 |
non_master_node, "regular", False, auto_promote=True) |
473 |
|
474 |
Finish(client, client.SetNodeRole, |
475 |
non_master_node, "master-candidate", False, auto_promote=True) |
476 |
|
477 |
Finish(client, client.SetNodeRole, |
478 |
non_master_node, "drained", False, auto_promote=True) |
479 |
|
480 |
Finish(client, client.SetNodeRole, |
481 |
non_master_node, old_role, False, auto_promote=True) |
482 |
|
483 |
Finish(client, client.PowercycleNode, |
484 |
non_master_node, force=False)
|
485 |
|
486 |
storage_units_fields = [ |
487 |
"name", "allocatable", "free", "node", "size", "type", "used", |
488 |
] |
489 |
|
490 |
for storage_type in constants.STS_REPORT: |
491 |
success, storage_units = Finish(client, client.GetNodeStorageUnits, |
492 |
non_master_node, storage_type, |
493 |
",".join(storage_units_fields))
|
494 |
|
495 |
if success and len(storage_units) > 0 and len(storage_units[0]) > 0: |
496 |
# Name is the first entry of the first result, allocatable the other
|
497 |
unit_name = storage_units[0][0] |
498 |
Finish(client, client.ModifyNodeStorageUnits, |
499 |
non_master_node, storage_type, unit_name, |
500 |
allocatable=not storage_units[0][1]) |
501 |
Finish(client, client.ModifyNodeStorageUnits, |
502 |
non_master_node, storage_type, unit_name, |
503 |
allocatable=storage_units[0][1]) |
504 |
Finish(client, client.RepairNodeStorageUnits, |
505 |
non_master_node, storage_type, unit_name) |
506 |
|
507 |
MarkUnmarkNode(client, non_master_node, "drained")
|
508 |
MarkUnmarkNode(client, non_master_node, "powered")
|
509 |
MarkUnmarkNode(client, non_master_node, "offline")
|
510 |
|
511 |
TestQueries(client, "node")
|
512 |
|
513 |
|
514 |
def TestGroupOperations(client, node, another_node): |
515 |
""" Tests various operations related to groups only.
|
516 |
|
517 |
@type client: C{GanetiRapiClientWrapper}
|
518 |
@param client: A Ganeti RAPI client to use.
|
519 |
@type node: string
|
520 |
@param node: The name of a node in the cluster.
|
521 |
@type another_node: string
|
522 |
@param another_node: The name of another node in the cluster.
|
523 |
|
524 |
"""
|
525 |
|
526 |
DEFAULT_GROUP_NAME = constants.INITIAL_NODE_GROUP_NAME |
527 |
TEST_GROUP_NAME = "TestGroup"
|
528 |
ALTERNATE_GROUP_NAME = "RenamedTestGroup"
|
529 |
|
530 |
Finish(client, client.CreateGroup, |
531 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED, |
532 |
dry_run=True)
|
533 |
|
534 |
Finish(client, client.CreateGroup, |
535 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED) |
536 |
|
537 |
client.GetGroup(TEST_GROUP_NAME) |
538 |
|
539 |
TestQueries(client, "group")
|
540 |
|
541 |
TestTags(client, client.GetGroupTags, client.AddGroupTags, |
542 |
client.DeleteGroupTags, TEST_GROUP_NAME) |
543 |
|
544 |
Finish(client, client.ModifyGroup, |
545 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED, |
546 |
depends=None)
|
547 |
|
548 |
Finish(client, client.AssignGroupNodes, |
549 |
TEST_GROUP_NAME, [node, another_node], force=False, dry_run=True) |
550 |
|
551 |
Finish(client, client.AssignGroupNodes, |
552 |
TEST_GROUP_NAME, [another_node], force=False)
|
553 |
|
554 |
Finish(client, client.RenameGroup, |
555 |
TEST_GROUP_NAME, ALTERNATE_GROUP_NAME) |
556 |
|
557 |
Finish(client, client.RenameGroup, |
558 |
ALTERNATE_GROUP_NAME, TEST_GROUP_NAME) |
559 |
|
560 |
Finish(client, client.AssignGroupNodes, |
561 |
DEFAULT_GROUP_NAME, [another_node], force=False)
|
562 |
|
563 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME, dry_run=True)
|
564 |
|
565 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME) |
566 |
|
567 |
|
568 |
def TestNetworkConnectDisconnect(client, network_name, mode, link): |
569 |
""" Test connecting and disconnecting the network to a new node group.
|
570 |
|
571 |
@type network_name: string
|
572 |
@param network_name: The name of an existing and unconnected network.
|
573 |
@type mode: string
|
574 |
@param mode: The network mode.
|
575 |
@type link: string
|
576 |
@param link: The network link.
|
577 |
|
578 |
"""
|
579 |
# For testing the connect/disconnect calls, a group is needed
|
580 |
TEST_GROUP_NAME = "TestGroup"
|
581 |
Finish(client, client.CreateGroup, |
582 |
TEST_GROUP_NAME, alloc_policy=constants.ALLOC_POLICY_PREFERRED) |
583 |
|
584 |
Finish(client, client.ConnectNetwork, |
585 |
network_name, TEST_GROUP_NAME, mode, link, dry_run=True)
|
586 |
|
587 |
Finish(client, client.ConnectNetwork, |
588 |
network_name, TEST_GROUP_NAME, mode, link) |
589 |
|
590 |
Finish(client, client.DisconnectNetwork, |
591 |
network_name, TEST_GROUP_NAME, dry_run=True)
|
592 |
|
593 |
Finish(client, client.DisconnectNetwork, |
594 |
network_name, TEST_GROUP_NAME) |
595 |
|
596 |
# Clean up the group
|
597 |
Finish(client, client.DeleteGroup, TEST_GROUP_NAME) |
598 |
|
599 |
|
600 |
def TestNetworks(client): |
601 |
""" Add some networks of different sizes, using RFC5737 addresses like in the
|
602 |
QA.
|
603 |
|
604 |
"""
|
605 |
|
606 |
NETWORK_NAME = "SurelyCertainlyNonexistentNetwork"
|
607 |
|
608 |
Finish(client, client.CreateNetwork, |
609 |
NETWORK_NAME, "192.0.2.0/30", tags=[], dry_run=True) |
610 |
|
611 |
Finish(client, client.CreateNetwork, |
612 |
NETWORK_NAME, "192.0.2.0/30", tags=[])
|
613 |
|
614 |
client.GetNetwork(NETWORK_NAME) |
615 |
|
616 |
TestTags(client, client.GetNetworkTags, client.AddNetworkTags, |
617 |
client.DeleteNetworkTags, NETWORK_NAME) |
618 |
|
619 |
Finish(client, client.ModifyNetwork, |
620 |
NETWORK_NAME, mac_prefix=None)
|
621 |
|
622 |
TestQueries(client, "network")
|
623 |
|
624 |
default_nicparams = qa_config.get("default-nicparams", None) |
625 |
|
626 |
# The entry might not be present in the QA config
|
627 |
if default_nicparams is not None: |
628 |
mode = default_nicparams.get("mode", None) |
629 |
link = default_nicparams.get("link", None) |
630 |
if mode is not None and link is not None: |
631 |
TestNetworkConnectDisconnect(client, NETWORK_NAME, mode, link) |
632 |
|
633 |
# Clean up the network
|
634 |
Finish(client, client.DeleteNetwork, |
635 |
NETWORK_NAME, dry_run=True)
|
636 |
|
637 |
Finish(client, client.DeleteNetwork, NETWORK_NAME) |
638 |
|
639 |
|
640 |
def CreateDRBDInstance(client, node_one, node_two, instance_name): |
641 |
""" Creates a DRBD-enabled instance on the given nodes.
|
642 |
|
643 |
"""
|
644 |
Finish(client, client.CreateInstance, |
645 |
"create", instance_name, "drbd", [{"size": "1000"}], [{}], |
646 |
os="debian-image", pnode=node_one, snode=node_two)
|
647 |
|
648 |
|
649 |
def TestInstanceMigrations(client, node_one, node_two, node_three, |
650 |
instance_name): |
651 |
""" Test various operations related to migrating instances.
|
652 |
|
653 |
@type node_one: string
|
654 |
@param node_one: The name of a node in the cluster.
|
655 |
@type node_two: string
|
656 |
@param node_two: The name of another node in the cluster.
|
657 |
@type node_three: string
|
658 |
@param node_three: The name of yet another node in the cluster.
|
659 |
@type instance_name: string
|
660 |
@param instance_name: An instance name that can be used.
|
661 |
|
662 |
"""
|
663 |
|
664 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
665 |
Finish(client, client.FailoverInstance, instance_name) |
666 |
Finish(client, client.DeleteInstance, instance_name) |
667 |
|
668 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
669 |
Finish(client, client.EvacuateNode, |
670 |
node_two, early_release=False, mode=NODE_EVAC_SEC,
|
671 |
remote_node=node_three) |
672 |
Finish(client, client.DeleteInstance, instance_name) |
673 |
|
674 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
675 |
Finish(client, client.EvacuateNode, |
676 |
node_one, early_release=False, mode=NODE_EVAC_PRI, iallocator="hail") |
677 |
Finish(client, client.DeleteInstance, instance_name) |
678 |
|
679 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
680 |
Finish(client, client.MigrateInstance, |
681 |
instance_name, cleanup=True, target_node=node_two)
|
682 |
Finish(client, client.DeleteInstance, instance_name) |
683 |
|
684 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
685 |
Finish(client, client.MigrateNode, |
686 |
node_one, iallocator="hail", mode="non-live") |
687 |
Finish(client, client.DeleteInstance, instance_name) |
688 |
|
689 |
CreateDRBDInstance(client, node_one, node_two, instance_name) |
690 |
Finish(client, client.MigrateNode, |
691 |
node_one, target_node=node_two, mode="non-live")
|
692 |
Finish(client, client.DeleteInstance, instance_name) |
693 |
|
694 |
|
695 |
def ExtractAllNicInformationPossible(nics, replace_macs=True): |
696 |
""" Extracts NIC information as a dictionary.
|
697 |
|
698 |
@type nics: list of tuples of varying structure
|
699 |
@param nics: The network interfaces, as received from the instance info RAPI
|
700 |
call.
|
701 |
|
702 |
@rtype: list of dict
|
703 |
@return: Dictionaries of NIC information.
|
704 |
|
705 |
The NIC information is returned in a different format across versions, and to
|
706 |
try and see if the execution of commands is still compatible, this function
|
707 |
attempts to grab all the info that it can.
|
708 |
|
709 |
"""
|
710 |
|
711 |
desired_entries = [ |
712 |
constants.INIC_IP, |
713 |
constants.INIC_MAC, |
714 |
constants.INIC_MODE, |
715 |
constants.INIC_LINK, |
716 |
constants.INIC_VLAN, |
717 |
constants.INIC_NETWORK, |
718 |
constants.INIC_NAME, |
719 |
] |
720 |
|
721 |
nic_dicts = [] |
722 |
for nic_index in range(len(nics)): |
723 |
nic_raw_data = nics[nic_index] |
724 |
|
725 |
# Fill dictionary with None-s as defaults
|
726 |
nic_dict = dict([(key, None) for key in desired_entries]) |
727 |
|
728 |
try:
|
729 |
# The 2.6 format
|
730 |
ip, mac, mode, link = nic_raw_data |
731 |
except ValueError: |
732 |
# If there is yet another ValueError here, let it go through as it is
|
733 |
# legitimate - we are out of versions
|
734 |
|
735 |
# The 2.11 format
|
736 |
nic_name, _, ip, mac, mode, link, vlan, network, _ = nic_raw_data |
737 |
nic_dict[constants.INIC_VLAN] = vlan |
738 |
nic_dict[constants.INIC_NETWORK] = network |
739 |
nic_dict[constants.INIC_NAME] = nic_name |
740 |
|
741 |
# These attributes will be present in either version
|
742 |
nic_dict[constants.INIC_IP] = ip |
743 |
nic_dict[constants.INIC_MAC] = mac |
744 |
nic_dict[constants.INIC_MODE] = mode |
745 |
nic_dict[constants.INIC_LINK] = link |
746 |
|
747 |
# Very simple mac generation, which should work as the setup cluster should
|
748 |
# have no mac prefix restrictions in the default network, and there is a
|
749 |
# hard and reasonable limit of only 8 NICs
|
750 |
if replace_macs:
|
751 |
nic_dict[constants.INIC_MAC] = "00:00:00:00:00:%02x" % nic_index
|
752 |
|
753 |
nic_dicts.append(nic_dict) |
754 |
|
755 |
return nic_dicts
|
756 |
|
757 |
|
758 |
def MoveInstance(client, src_instance, dst_instance, src_node, dst_node): |
759 |
""" Moves a single instance, compatible with 2.6.
|
760 |
|
761 |
@rtype: bool
|
762 |
@return: Whether the instance was moved successfully
|
763 |
|
764 |
"""
|
765 |
success, inst_info_all = Finish(client, client.GetInstanceInfo, |
766 |
src_instance.name, static=True)
|
767 |
|
768 |
if not success or src_instance.name not in inst_info_all: |
769 |
raise Exception("Did not find the source instance information!") |
770 |
|
771 |
inst_info = inst_info_all[src_instance.name] |
772 |
|
773 |
# Try to extract NICs first, as this is the operation most likely to fail
|
774 |
try:
|
775 |
nic_info = ExtractAllNicInformationPossible(inst_info["nics"])
|
776 |
except ValueError: |
777 |
# Without the NIC info, there is very little we can do
|
778 |
return False |
779 |
|
780 |
NIC_COMPONENTS_26 = [ |
781 |
constants.INIC_IP, |
782 |
constants.INIC_MAC, |
783 |
constants.INIC_MODE, |
784 |
constants.INIC_LINK, |
785 |
] |
786 |
|
787 |
nic_converter = lambda old: dict((k, old[k]) for k in NIC_COMPONENTS_26) |
788 |
nics = map(nic_converter, nic_info)
|
789 |
|
790 |
# Prepare the parameters
|
791 |
disks = [] |
792 |
for idisk in inst_info["disks"]: |
793 |
odisk = { |
794 |
constants.IDISK_SIZE: idisk["size"],
|
795 |
constants.IDISK_MODE: idisk["mode"],
|
796 |
} |
797 |
|
798 |
spindles = idisk.get("spindles")
|
799 |
if spindles is not None: |
800 |
odisk[constants.IDISK_SPINDLES] = spindles |
801 |
|
802 |
# Disk name may be present, but must not be supplied in 2.6!
|
803 |
disks.append(odisk) |
804 |
|
805 |
# With all the parameters properly prepared, try the export
|
806 |
success, exp_info = Finish(client, client.PrepareExport, |
807 |
src_instance.name, constants.EXPORT_MODE_REMOTE) |
808 |
|
809 |
if not success: |
810 |
# The instance will still have to be deleted
|
811 |
return False |
812 |
|
813 |
success, _ = Finish(client, client.CreateInstance, |
814 |
constants.INSTANCE_REMOTE_IMPORT, dst_instance.name, |
815 |
inst_info["disk_template"], disks, nics,
|
816 |
os=inst_info["os"],
|
817 |
pnode=dst_node.primary, |
818 |
snode=src_node.primary, # Ignored as no DRBD
|
819 |
start=(inst_info["config_state"] == "up"), |
820 |
ip_check=False,
|
821 |
iallocator=inst_info.get("iallocator", None), |
822 |
hypervisor=inst_info["hypervisor"],
|
823 |
source_handshake=exp_info["handshake"],
|
824 |
source_x509_ca=exp_info["x509_ca"],
|
825 |
source_instance_name=inst_info["name"],
|
826 |
beparams=inst_info["be_instance"],
|
827 |
hvparams=inst_info["hv_instance"],
|
828 |
osparams=inst_info["os_instance"])
|
829 |
|
830 |
return success
|
831 |
|
832 |
|
833 |
def CreateInstanceForMoveTest(client, node, instance): |
834 |
""" Creates a single shutdown instance to move about in tests.
|
835 |
|
836 |
@type node: C{_QaNode}
|
837 |
@param node: A node configuration object.
|
838 |
@type instance: C{_QaInstance}
|
839 |
@param instance: An instance configuration object.
|
840 |
|
841 |
"""
|
842 |
Finish(client, client.CreateInstance, |
843 |
"create", instance.name, "plain", [{"size": "2000"}], [{}], |
844 |
os="debian-image", pnode=node.primary)
|
845 |
|
846 |
Finish(client, client.ShutdownInstance, |
847 |
instance.name, dry_run=False, no_remember=False) |
848 |
|
849 |
|
850 |
def Test26InstanceMove(client, node_one, node_two, instance_to_create, |
851 |
new_instance): |
852 |
""" Tests instance moves using commands that work in 2.6.
|
853 |
|
854 |
"""
|
855 |
|
856 |
# First create the instance to move
|
857 |
CreateInstanceForMoveTest(client, node_one, instance_to_create) |
858 |
|
859 |
# The cleanup should be conditional on operation success
|
860 |
if MoveInstance(client, instance_to_create, new_instance, node_one, node_two):
|
861 |
Finish(client, client.DeleteInstance, new_instance.name) |
862 |
else:
|
863 |
Finish(client, client.DeleteInstance, instance_to_create.name) |
864 |
|
865 |
|
866 |
def Test211InstanceMove(client, node_one, node_two, instance_to_create, |
867 |
new_instance): |
868 |
""" Tests instance moves using the QA-provided move test.
|
869 |
|
870 |
"""
|
871 |
|
872 |
# First create the instance to move
|
873 |
CreateInstanceForMoveTest(client, node_one, instance_to_create) |
874 |
|
875 |
instance_to_create.SetDiskTemplate("plain")
|
876 |
|
877 |
try:
|
878 |
qa_rapi.TestInterClusterInstanceMove(instance_to_create, new_instance, |
879 |
[node_one], node_two, |
880 |
perform_checks=False)
|
881 |
except qa_error.Error:
|
882 |
# A failure is sad, but requires no special actions to be undertaken
|
883 |
pass
|
884 |
|
885 |
# Try to delete the instance when done - either the move has failed, or
|
886 |
# a double move was performed - the instance to delete is one and the same
|
887 |
Finish(client, client.DeleteInstance, instance_to_create.name) |
888 |
|
889 |
|
890 |
def TestInstanceMoves(client, node_one, node_two, instance_to_create, |
891 |
new_instance): |
892 |
""" Performs two types of instance moves, one compatible with 2.6, the other
|
893 |
with 2.11.
|
894 |
|
895 |
@type node_one: C{_QaNode}
|
896 |
@param node_one: A node configuration object.
|
897 |
@type node_two: C{_QaNode}
|
898 |
@param node_two: A node configuration object.
|
899 |
@type instance_to_create: C{_QaInstance}
|
900 |
@param instance_to_create: An instance configuration object.
|
901 |
@type new_instance: C{_QaInstance}
|
902 |
@param new_instance: An instance configuration object.
|
903 |
|
904 |
"""
|
905 |
|
906 |
Test26InstanceMove(client, node_one, node_two, instance_to_create, |
907 |
new_instance) |
908 |
Test211InstanceMove(client, node_one, node_two, instance_to_create, |
909 |
new_instance) |
910 |
|
911 |
|
912 |
def TestClusterParameterModification(client): |
913 |
""" Try to modify some of the cluster parameters using RAPI.
|
914 |
|
915 |
"""
|
916 |
cluster_info = client.GetInfo() |
917 |
|
918 |
# Each attribute has several safe choices we can use
|
919 |
ATTRIBUTES_TO_MODIFY = [ |
920 |
("default_iallocator", ["hail", ""]), # Use "" to reset |
921 |
("candidate_pool_size", [1, 5]), |
922 |
("maintain_node_health", [True, False]), |
923 |
] |
924 |
|
925 |
for attribute, options in ATTRIBUTES_TO_MODIFY: |
926 |
current_value = cluster_info[attribute] |
927 |
|
928 |
if current_value in options: |
929 |
value_to_use = options[1 - options.index(current_value)]
|
930 |
else:
|
931 |
value_to_use = options[0]
|
932 |
|
933 |
#pylint: disable=W0142
|
934 |
Finish(client, client.ModifyCluster, **{attribute: value_to_use}) |
935 |
Finish(client, client.ModifyCluster, **{attribute: current_value}) |
936 |
#pylint: enable=W0142
|
937 |
|
938 |
|
939 |
def TestJobCancellation(client, node_one, node_two, instance_one, instance_two): |
940 |
""" Test if jobs can be cancelled.
|
941 |
|
942 |
@type node_one: string
|
943 |
@param node_one: The name of a node in the cluster.
|
944 |
@type node_two: string
|
945 |
@param node_two: The name of another node in the cluster.
|
946 |
@type instance_one: string
|
947 |
@param instance_one: An available instance name.
|
948 |
@type instance_two: string
|
949 |
@param instance_two: An available instance name.
|
950 |
|
951 |
"""
|
952 |
|
953 |
# Just in case, remove all previously present instances
|
954 |
RemoveAllInstances(client) |
955 |
|
956 |
# Let us issue a job that is sure to both succeed and last for a while
|
957 |
running_job = client.CreateInstance("create", instance_one, "drbd", |
958 |
[{"size": "5000"}], [{}], |
959 |
os="debian-image", pnode=node_one,
|
960 |
snode=node_two) |
961 |
|
962 |
# And immediately afterwards, another very similar one
|
963 |
job_to_cancel = client.CreateInstance("create", instance_two, "drbd", |
964 |
[{"size": "5000"}], [{}], |
965 |
os="debian-image", pnode=node_one,
|
966 |
snode=node_two) |
967 |
|
968 |
# Try to cancel, which should fail as the job is already running
|
969 |
success, msg = client.CancelJob(running_job) |
970 |
if success:
|
971 |
print "Job succeeded: this should not have happened as it is running!" |
972 |
print "Message: %s" % msg |
973 |
|
974 |
success, msg = client.CancelJob(job_to_cancel) |
975 |
if not success: |
976 |
print "Job failed: this was unexpected as it was not a dry run" |
977 |
print "Message: %s" % msg |
978 |
|
979 |
# And wait for the proper job
|
980 |
client.WaitForJobCompletion(running_job) |
981 |
|
982 |
# Remove all the leftover instances, success or no success
|
983 |
RemoveAllInstances(client) |
984 |
|
985 |
|
986 |
def Workload(client): |
987 |
""" The actual RAPI workload used for tests.
|
988 |
|
989 |
@type client: C{GanetiRapiClientWrapper}
|
990 |
@param client: A wrapped RAPI client.
|
991 |
|
992 |
"""
|
993 |
|
994 |
# First just the simple information retrievals
|
995 |
TestGetters(client) |
996 |
|
997 |
# Then the only remaining function which is parameter-free
|
998 |
Finish(client, client.RedistributeConfig) |
999 |
|
1000 |
# Try changing the cluster parameters
|
1001 |
TestClusterParameterModification(client) |
1002 |
|
1003 |
TestTags(client, client.GetClusterTags, client.AddClusterTags, |
1004 |
client.DeleteClusterTags) |
1005 |
|
1006 |
# Generously assume the master is present
|
1007 |
node = qa_config.AcquireNode() |
1008 |
TestTags(client, client.GetNodeTags, client.AddNodeTags, |
1009 |
client.DeleteNodeTags, node.primary) |
1010 |
node.Release() |
1011 |
|
1012 |
# Instance tests
|
1013 |
|
1014 |
# First remove all instances the QA might have created
|
1015 |
RemoveAllInstances(client) |
1016 |
|
1017 |
nodes = qa_config.AcquireManyNodes(2)
|
1018 |
instances = qa_config.AcquireManyInstances(2)
|
1019 |
TestSingleInstance(client, instances[0].name, instances[1].name, |
1020 |
nodes[0].primary, nodes[1].primary) |
1021 |
qa_config.ReleaseManyInstances(instances) |
1022 |
qa_config.ReleaseManyNodes(nodes) |
1023 |
|
1024 |
# Test all the queries which involve resources that do not have functions
|
1025 |
# of their own
|
1026 |
TestQueries(client, "lock")
|
1027 |
TestQueries(client, "job")
|
1028 |
TestQueries(client, "export")
|
1029 |
|
1030 |
node = qa_config.AcquireNode(exclude=qa_config.GetMasterNode()) |
1031 |
TestNodeOperations(client, node.primary) |
1032 |
TestQueryFiltering(client, node.primary) |
1033 |
node.Release() |
1034 |
|
1035 |
nodes = qa_config.AcquireManyNodes(2)
|
1036 |
TestGroupOperations(client, nodes[0].primary, nodes[1].primary) |
1037 |
qa_config.ReleaseManyNodes(nodes) |
1038 |
|
1039 |
TestNetworks(client) |
1040 |
|
1041 |
nodes = qa_config.AcquireManyNodes(3)
|
1042 |
instance = qa_config.AcquireInstance() |
1043 |
TestInstanceMigrations(client, nodes[0].primary, nodes[1].primary, |
1044 |
nodes[2].primary, instance.name)
|
1045 |
instance.Release() |
1046 |
qa_config.ReleaseManyNodes(nodes) |
1047 |
|
1048 |
nodes = qa_config.AcquireManyNodes(2)
|
1049 |
instances = qa_config.AcquireManyInstances(2)
|
1050 |
TestInstanceMoves(client, nodes[0], nodes[1], instances[0], instances[1]) |
1051 |
TestJobCancellation(client, nodes[0].primary, nodes[1].primary, |
1052 |
instances[0].name, instances[1].name) |
1053 |
qa_config.ReleaseManyInstances(instances) |
1054 |
qa_config.ReleaseManyNodes(nodes) |
1055 |
|
1056 |
|
1057 |
def Main(): |
1058 |
parser = optparse.OptionParser(usage="%prog [options] <config-file>")
|
1059 |
parser.add_option("--yes-do-it", dest="yes_do_it", |
1060 |
action="store_true",
|
1061 |
help="Really execute the tests")
|
1062 |
parser.add_option("--show-invocations", dest="show_invocations", |
1063 |
action="store_true",
|
1064 |
help="Show which client methods have and have not been "
|
1065 |
"called")
|
1066 |
(opts, args) = parser.parse_args() |
1067 |
|
1068 |
if not opts.yes_do_it: |
1069 |
print ("Executing this script irreversibly destroys any Ganeti\n" |
1070 |
"configuration on all nodes involved. If you really want\n"
|
1071 |
"to start testing, supply the --yes-do-it option.")
|
1072 |
sys.exit(1)
|
1073 |
|
1074 |
qa_config.Load(args[0])
|
1075 |
|
1076 |
# Only the master will be present after a fresh QA cluster setup, so we have
|
1077 |
# to invoke this to get all the other nodes.
|
1078 |
qa_node.TestNodeAddAll() |
1079 |
|
1080 |
client = GanetiRapiClientWrapper() |
1081 |
|
1082 |
Workload(client) |
1083 |
|
1084 |
qa_node.TestNodeRemoveAll() |
1085 |
|
1086 |
# The method invoked has the naming of the protected method, and pylint does
|
1087 |
# not like this. Disabling the warning is healthier than explicitly adding and
|
1088 |
# maintaining an exception for this method in the wrapper.
|
1089 |
if opts.show_invocations:
|
1090 |
# pylint: disable=W0212
|
1091 |
client._OutputMethodInvocationDetails() |
1092 |
# pylint: enable=W0212
|
1093 |
|
1094 |
if __name__ == "__main__": |
1095 |
Main() |