root / snf-cyclades-app / synnefo / logic / management / commands / reconcile.py @ 0e9a423f
History | View | Annotate | Download (12.4 kB)
1 |
# Copyright 2011-2012 GRNET S.A. All rights reserved.
|
---|---|
2 |
#
|
3 |
# Redistribution and use in source and binary forms, with or without
|
4 |
# modification, are permitted provided that the following conditions
|
5 |
# are met:
|
6 |
#
|
7 |
# 1. Redistributions of source code must retain the above copyright
|
8 |
# notice, this list of conditions and the following disclaimer.
|
9 |
#
|
10 |
# 2. Redistributions in binary form must reproduce the above copyright
|
11 |
# notice, this list of conditions and the following disclaimer in the
|
12 |
# documentation and/or other materials provided with the distribution.
|
13 |
#
|
14 |
# THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
15 |
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
16 |
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
17 |
# ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
18 |
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
19 |
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
20 |
# OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
21 |
# HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
22 |
# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
23 |
# OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
24 |
# SUCH DAMAGE.
|
25 |
#
|
26 |
# The views and conclusions contained in the software and documentation are
|
27 |
# those of the authors and should not be interpreted as representing official
|
28 |
# policies, either expressed or implied, of GRNET S.A.
|
29 |
#
|
30 |
"""Reconciliation management command
|
31 |
|
32 |
Management command to reconcile the contents of the Synnefo DB with
|
33 |
the state of the Ganeti backend. See docstring on top of
|
34 |
logic/reconciliation.py for a description of reconciliation rules.
|
35 |
|
36 |
"""
|
37 |
import sys |
38 |
import datetime |
39 |
import subprocess |
40 |
|
41 |
from optparse import make_option |
42 |
|
43 |
from django.conf import settings |
44 |
from django.core.management.base import BaseCommand, CommandError |
45 |
|
46 |
from synnefo.db.models import VirtualMachine |
47 |
from synnefo.logic import reconciliation, backend |
48 |
|
49 |
|
50 |
class Command(BaseCommand): |
51 |
can_import_settings = True
|
52 |
|
53 |
help = 'Reconcile contents of Synnefo DB with state of Ganeti backend'
|
54 |
output_transaction = True # The management command runs inside |
55 |
# an SQL transaction
|
56 |
option_list = BaseCommand.option_list + ( |
57 |
make_option('--detect-stale', action='store_true', dest='detect_stale', |
58 |
default=False, help='Detect stale VM entries in DB'), |
59 |
make_option('--detect-orphans', action='store_true', |
60 |
dest='detect_orphans',
|
61 |
default=False, help='Detect orphan instances in Ganeti'), |
62 |
make_option('--detect-unsynced', action='store_true', |
63 |
dest='detect_unsynced',
|
64 |
default=False, help='Detect unsynced operstate between ' + |
65 |
'DB and Ganeti'),
|
66 |
make_option('--detect-build-errors', action='store_true', |
67 |
dest='detect_build_errors', default=False, |
68 |
help='Detect instances with build error'),
|
69 |
make_option('--detect-unsynced-nics', action='store_true', |
70 |
dest='detect_unsynced_nics', default=False, |
71 |
help='Detect unsynced nics between DB and Ganeti'),
|
72 |
make_option('--detect-all', action='store_true', |
73 |
dest='detect_all',
|
74 |
default=False, help='Enable all --detect-* arguments'), |
75 |
make_option('--fix-stale', action='store_true', dest='fix_stale', |
76 |
default=False, help='Fix (remove) stale DB entries in DB'), |
77 |
make_option('--fix-orphans', action='store_true', dest='fix_orphans', |
78 |
default=False, help='Fix (remove) orphan Ganeti VMs'), |
79 |
make_option('--fix-unsynced', action='store_true', dest='fix_unsynced', |
80 |
default=False, help='Fix server operstate in DB, set ' + |
81 |
'from Ganeti'),
|
82 |
make_option('--fix-build-errors', action='store_true', |
83 |
dest='fix_build_errors', default=False, |
84 |
help='Fix (remove) instances with build errors'),
|
85 |
make_option('--fix-unsynced-nics', action='store_true', |
86 |
dest='fix_unsynced_nics', default=False, |
87 |
help='Fix unsynced nics between DB and Ganeti'),
|
88 |
make_option('--fix-all', action='store_true', dest='fix_all', |
89 |
default=False, help='Enable all --fix-* arguments')) |
90 |
|
91 |
def _process_args(self, options): |
92 |
keys_detect = [k for k in options.keys() if k.startswith('detect_')] |
93 |
keys_fix = [k for k in options.keys() if k.startswith('fix_')] |
94 |
|
95 |
if options['detect_all']: |
96 |
for kd in keys_detect: |
97 |
options[kd] = True
|
98 |
if options['fix_all']: |
99 |
for kf in keys_fix: |
100 |
options[kf] = True
|
101 |
|
102 |
if not reduce(lambda x, y: x or y, |
103 |
map(lambda x: options[x], keys_detect)): |
104 |
raise CommandError("At least one of --detect-* must be specified") |
105 |
|
106 |
for kf in keys_fix: |
107 |
kd = kf.replace('fix_', 'detect_', 1) |
108 |
if (options[kf] and not options[kd]): |
109 |
raise CommandError("Cannot use --%s without corresponding " |
110 |
"--%s argument" % (kf, kd))
|
111 |
|
112 |
def handle(self, **options): |
113 |
verbosity = int(options['verbosity']) |
114 |
self._process_args(options)
|
115 |
|
116 |
D = reconciliation.get_servers_from_db() |
117 |
G = reconciliation.get_instances_from_ganeti() |
118 |
|
119 |
DBNics = reconciliation.get_nics_from_db() |
120 |
GNics = reconciliation.get_nics_from_ganeti() |
121 |
#
|
122 |
# Detect problems
|
123 |
#
|
124 |
if options['detect_stale']: |
125 |
stale = reconciliation.stale_servers_in_db(D, G) |
126 |
if len(stale) > 0: |
127 |
print >> sys.stderr, "Found the following stale server IDs: " |
128 |
print " " + "\n ".join( |
129 |
[str(x) for x in stale]) |
130 |
elif verbosity == 2: |
131 |
print >> sys.stderr, "Found no stale server IDs in DB." |
132 |
|
133 |
if options['detect_orphans']: |
134 |
orphans = reconciliation.orphan_instances_in_ganeti(D, G) |
135 |
if len(orphans) > 0: |
136 |
print >> sys.stderr, "Found orphan Ganeti instances with IDs: " |
137 |
print " " + "\n ".join( |
138 |
[str(x) for x in orphans]) |
139 |
elif verbosity == 2: |
140 |
print >> sys.stderr, "Found no orphan Ganeti instances." |
141 |
|
142 |
if options['detect_unsynced']: |
143 |
unsynced = reconciliation.unsynced_operstate(D, G) |
144 |
if len(unsynced) > 0: |
145 |
print >> sys.stderr, "The operstate of the following server" \ |
146 |
" IDs is out-of-sync:"
|
147 |
print " " + "\n ".join( |
148 |
["%d is %s in DB, %s in Ganeti" %
|
149 |
(x[0], x[1], ('UP' if x[2] else 'DOWN')) |
150 |
for x in unsynced]) |
151 |
elif verbosity == 2: |
152 |
print >> sys.stderr, "The operstate of all servers is in sync." |
153 |
|
154 |
if options['detect_build_errors']: |
155 |
build_errors = reconciliation.instances_with_build_errors(D, G) |
156 |
if len(build_errors) > 0: |
157 |
print >> sys.stderr, "The os for the following server IDs was "\ |
158 |
"not build successfully:"
|
159 |
print " " + "\n ".join( |
160 |
["%d" % x for x in build_errors]) |
161 |
elif verbosity == 2: |
162 |
print >> sys.stderr, "Found no instances with build errors." |
163 |
|
164 |
if options['detect_unsynced_nics']: |
165 |
def pretty_print_nics(nics): |
166 |
if not nics: |
167 |
print ''.ljust(18) + 'None' |
168 |
for index, info in nics.items(): |
169 |
print ''.ljust(18) + 'nic/' + str(index) + ': MAC: %s, IP: %s, Network: %s' % \ |
170 |
(info['mac'], info['ipv4'], info['network']) |
171 |
|
172 |
unsynced_nics = reconciliation.unsynced_nics(DBNics, GNics) |
173 |
if len(unsynced_nics) > 0: |
174 |
print >> sys.stderr, "The nics of servers with the folloing ID's "\ |
175 |
"are unsynced:"
|
176 |
for id, nics in unsynced_nics.items(): |
177 |
print ''.ljust(2) + '%6d:' % id |
178 |
print ''.ljust(8) + '%8s:' % 'DB' |
179 |
pretty_print_nics(nics[0])
|
180 |
print ''.ljust(8) + '%8s:' % 'Ganeti' |
181 |
pretty_print_nics(nics[1])
|
182 |
elif verbosity == 2: |
183 |
print >> sys.stderr, "All instance nics are synced." |
184 |
|
185 |
#
|
186 |
# Then fix them
|
187 |
#
|
188 |
if options['fix_stale'] and len(stale) > 0: |
189 |
print >> sys.stderr, \
|
190 |
"Simulating successful Ganeti removal for %d " \
|
191 |
"servers in the DB:" % len(stale) |
192 |
for vm in VirtualMachine.objects.filter(pk__in=stale): |
193 |
event_time = datetime.datetime.now() |
194 |
backend.process_op_status(vm=vm, etime=event_time, jobid=-0,
|
195 |
opcode='OP_INSTANCE_REMOVE', status='success', |
196 |
logmsg='Reconciliation: simulated Ganeti event')
|
197 |
print >> sys.stderr, " ...done" |
198 |
|
199 |
if options['fix_orphans'] and len(orphans) > 0: |
200 |
print >> sys.stderr, \
|
201 |
"Issuing OP_INSTANCE_REMOVE for %d Ganeti instances:" % \
|
202 |
len(orphans)
|
203 |
for id in orphans: |
204 |
vm = VirtualMachine.objects.get(pk=id)
|
205 |
vm.client.DeleteInstance('%s%s' %
|
206 |
(settings.BACKEND_PREFIX_ID, str(id))) |
207 |
print >> sys.stderr, " ...done" |
208 |
|
209 |
if options['fix_unsynced'] and len(unsynced) > 0: |
210 |
print >> sys.stderr, "Setting the state of %d out-of-sync VMs:" % \ |
211 |
len(unsynced)
|
212 |
for id, db_state, ganeti_up in unsynced: |
213 |
vm = VirtualMachine.objects.get(pk=id)
|
214 |
opcode = "OP_INSTANCE_REBOOT" if ganeti_up \ |
215 |
else "OP_INSTANCE_SHUTDOWN" |
216 |
event_time = datetime.datetime.now() |
217 |
backend.process_op_status(vm=vm, etime=event_time, jobid=-0,
|
218 |
opcode=opcode, status='success',
|
219 |
logmsg='Reconciliation: simulated Ganeti event')
|
220 |
print >> sys.stderr, " ...done" |
221 |
|
222 |
if options['fix_build_errors'] and len(build_errors) > 0: |
223 |
print >> sys.stderr, "Setting the state of %d build-errors VMs:" % \ |
224 |
len(build_errors)
|
225 |
for id in build_errors: |
226 |
vm = VirtualMachine.objects.get(pk=id)
|
227 |
event_time = datetime.datetime.now() |
228 |
backend.process_op_status(vm=vm, etime=event_time, jobid=-0,
|
229 |
opcode="OP_INSTANCE_CREATE", status='error', |
230 |
logmsg='Reconciliation: simulated Ganeti event')
|
231 |
print >> sys.stderr, " ...done" |
232 |
|
233 |
if options['fix_unsynced_nics'] and len(unsynced_nics) > 0: |
234 |
print >> sys.stderr, "Setting the nics of %d out-of-sync VMs:" % \ |
235 |
len(unsynced_nics)
|
236 |
for id, nics in unsynced_nics.items(): |
237 |
vm = VirtualMachine.objects.get(pk=id)
|
238 |
nics = nics[1] # Ganeti nics |
239 |
if nics == {}: # No nics |
240 |
vm.nics.all.delete() |
241 |
continue
|
242 |
for index, nic in nics.items(): |
243 |
# Produce ipv6
|
244 |
ipv6 = mac2eui64(nic['mac'], settings.PUBLIC_IPV6_PREFIX)
|
245 |
nic['ipv6'] = ipv6
|
246 |
# Rename ipv4 to ip
|
247 |
nic['ip'] = nic['ipv4'] |
248 |
# Dict to sorted list
|
249 |
final_nics = [] |
250 |
nics_keys = nics.keys() |
251 |
nics_keys.sort() |
252 |
for i in nics_keys: |
253 |
if nics[i]['network']: |
254 |
final_nics.append(nics[i]) |
255 |
else:
|
256 |
print 'Network of nic %d of vm %s is None. ' \ |
257 |
'Can not reconcile' % (i, vm.backend_vm_id)
|
258 |
event_time = datetime.datetime.now() |
259 |
backend.process_net_status(vm=vm, etime=event_time, nics=final_nics) |
260 |
print >> sys.stderr, " ...done" |
261 |
|
262 |
|
263 |
def mac2eui64(mac, prefixstr): |
264 |
process = subprocess.Popen(["mac2eui64", mac, prefixstr],
|
265 |
stdout=subprocess.PIPE) |
266 |
return process.stdout.read().rstrip()
|