@recaptime-dev's working patches + fork for Phorge, a community fork of Phabricator. (Upstream dev and stable branches are at upstream/main and upstream/stable respectively.)
hq.recaptime.dev/wiki/Phorge
phorge
phabricator
Allow "bin/repository thaw --demote" to demote an entire service, not just a single device
Summary: Ref T13222. See PHI992. If you lose an entire cluster, you may want to aggressively demote it out of existence. You currently need to `xargs` your way through this. Allow `--demote <service>`, which demotes all devices in a service.
Test Plan: Demoted with `--demote <device>` and `--demote <service>`. Hit the `--promote service` error.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19850
+160
-107
+148
-107
src/applications/repository/management/PhabricatorRepositoryManagementThawWorkflow.php
+148
-107
src/applications/repository/management/PhabricatorRepositoryManagementThawWorkflow.php
···
15
15
array(
16
16
array(
17
17
'name' => 'demote',
18
-
'param' => 'device',
18
+
'param' => 'device/service',
19
19
'help' => pht(
20
-
'Demote a device, discarding local changes. Clears stuck '.
21
-
'write locks and recovers from lost leaders.'),
20
+
'Demote a device (or all devices in a service) discarding '.
21
+
'local changes. Clears stuck write locks and recovers from '.
22
+
'lost leaders.'),
22
23
),
23
24
array(
24
25
'name' => 'promote',
···
61
62
pht('Specify either --promote or --demote, but not both.'));
62
63
}
63
64
64
-
$device_name = nonempty($promote, $demote);
65
+
$target_name = nonempty($promote, $demote);
65
66
66
-
$device = id(new AlmanacDeviceQuery())
67
+
$devices = id(new AlmanacDeviceQuery())
67
68
->setViewer($viewer)
68
-
->withNames(array($device_name))
69
-
->executeOne();
70
-
if (!$device) {
71
-
throw new PhutilArgumentUsageException(
72
-
pht('No device "%s" exists.', $device_name));
69
+
->withNames(array($target_name))
70
+
->execute();
71
+
if (!$devices) {
72
+
$service = id(new AlmanacServiceQuery())
73
+
->setViewer($viewer)
74
+
->withNames(array($target_name))
75
+
->executeOne();
76
+
77
+
if (!$service) {
78
+
throw new PhutilArgumentUsageException(
79
+
pht('No device or service named "%s" exists.', $target_name));
80
+
}
81
+
82
+
if ($promote) {
83
+
throw new PhutilArgumentUsageException(
84
+
pht(
85
+
'You can not "--promote" an entire service ("%s"). Only a single '.
86
+
'device may be promoted.',
87
+
$target_name));
88
+
}
89
+
90
+
$bindings = id(new AlmanacBindingQuery())
91
+
->setViewer($viewer)
92
+
->withServicePHIDs(array($service->getPHID()))
93
+
->execute();
94
+
if (!$bindings) {
95
+
throw new PhutilArgumentUsageException(
96
+
pht(
97
+
'Service "%s" is not bound to any devices.',
98
+
$target_name));
99
+
}
100
+
101
+
$interfaces = id(new AlmanacInterfaceQuery())
102
+
->setViewer($viewer)
103
+
->withPHIDs(mpull($bindings, 'getInterfacePHID'))
104
+
->execute();
105
+
106
+
$device_phids = mpull($interfaces, 'getDevicePHID');
107
+
108
+
$devices = id(new AlmanacDeviceQuery())
109
+
->setViewer($viewer)
110
+
->withPHIDs($device_phids)
111
+
->execute();
73
112
}
74
113
75
114
$repository_names = $args->getArg('repositories');
···
97
136
98
137
$services = id(new AlmanacServiceQuery())
99
138
->setViewer($viewer)
100
-
->withDevicePHIDs(array($device->getPHID()))
139
+
->withDevicePHIDs(mpull($devices, 'getPHID'))
101
140
->execute();
102
141
if ($services) {
103
142
$repositories = id(new PhabricatorRepositoryQuery())
···
108
147
109
148
if (!$repositories) {
110
149
throw new PhutilArgumentUsageException(
111
-
pht('There are no repositories on the selected device.'));
150
+
pht('There are no repositories on the selected device or service.'));
112
151
}
113
152
}
114
153
···
150
189
pht('User aborted the workflow.'));
151
190
}
152
191
153
-
foreach ($repositories as $repository) {
154
-
$repository_phid = $repository->getPHID();
155
-
156
-
$write_lock = PhabricatorRepositoryWorkingCopyVersion::getWriteLock(
157
-
$repository_phid);
192
+
foreach ($devices as $device) {
193
+
foreach ($repositories as $repository) {
194
+
$repository_phid = $repository->getPHID();
158
195
159
-
echo tsprintf(
160
-
"%s\n",
161
-
pht(
162
-
'Waiting to acquire write lock for "%s"...',
163
-
$repository->getDisplayName()));
196
+
$write_lock = PhabricatorRepositoryWorkingCopyVersion::getWriteLock(
197
+
$repository_phid);
164
198
165
-
$write_lock->lock(phutil_units('5 minutes in seconds'));
166
-
try {
199
+
echo tsprintf(
200
+
"%s\n",
201
+
pht(
202
+
'Waiting to acquire write lock for "%s"...',
203
+
$repository->getDisplayName()));
167
204
168
-
$service = $repository->loadAlmanacService();
169
-
if (!$service) {
170
-
throw new PhutilArgumentUsageException(
171
-
pht(
172
-
'Repository "%s" is not a cluster repository: it is not '.
173
-
'bound to an Almanac service.',
174
-
$repository->getDisplayName()));
175
-
}
205
+
$write_lock->lock(phutil_units('5 minutes in seconds'));
206
+
try {
176
207
177
-
if ($promote) {
178
-
// You can only promote active devices. (You may demote active or
179
-
// inactive devices.)
180
-
$bindings = $service->getActiveBindings();
181
-
$bindings = mpull($bindings, null, 'getDevicePHID');
182
-
if (empty($bindings[$device->getPHID()])) {
208
+
$service = $repository->loadAlmanacService();
209
+
if (!$service) {
183
210
throw new PhutilArgumentUsageException(
184
211
pht(
185
-
'Repository "%s" has no active binding to device "%s". Only '.
186
-
'actively bound devices can be promoted.',
187
-
$repository->getDisplayName(),
188
-
$device->getName()));
212
+
'Repository "%s" is not a cluster repository: it is not '.
213
+
'bound to an Almanac service.',
214
+
$repository->getDisplayName()));
189
215
}
190
216
191
-
$versions = PhabricatorRepositoryWorkingCopyVersion::loadVersions(
192
-
$repository->getPHID());
193
-
$versions = mpull($versions, null, 'getDevicePHID');
217
+
if ($promote) {
218
+
// You can only promote active devices. (You may demote active or
219
+
// inactive devices.)
220
+
$bindings = $service->getActiveBindings();
221
+
$bindings = mpull($bindings, null, 'getDevicePHID');
222
+
if (empty($bindings[$device->getPHID()])) {
223
+
throw new PhutilArgumentUsageException(
224
+
pht(
225
+
'Repository "%s" has no active binding to device "%s". '.
226
+
'Only actively bound devices can be promoted.',
227
+
$repository->getDisplayName(),
228
+
$device->getName()));
229
+
}
230
+
231
+
$versions = PhabricatorRepositoryWorkingCopyVersion::loadVersions(
232
+
$repository->getPHID());
233
+
$versions = mpull($versions, null, 'getDevicePHID');
194
234
195
-
// Before we promote, make sure there are no outstanding versions on
196
-
// devices with inactive bindings. If there are, you need to demote
197
-
// these first.
198
-
$inactive = array();
199
-
foreach ($versions as $device_phid => $version) {
200
-
if (isset($bindings[$device_phid])) {
201
-
continue;
235
+
// Before we promote, make sure there are no outstanding versions
236
+
// on devices with inactive bindings. If there are, you need to
237
+
// demote these first.
238
+
$inactive = array();
239
+
foreach ($versions as $device_phid => $version) {
240
+
if (isset($bindings[$device_phid])) {
241
+
continue;
242
+
}
243
+
$inactive[$device_phid] = $version;
202
244
}
203
-
$inactive[$device_phid] = $version;
204
-
}
205
245
206
-
if ($inactive) {
207
-
$handles = $viewer->loadHandles(array_keys($inactive));
246
+
if ($inactive) {
247
+
$handles = $viewer->loadHandles(array_keys($inactive));
208
248
209
-
$handle_list = iterator_to_array($handles);
210
-
$handle_list = mpull($handle_list, 'getName');
211
-
$handle_list = implode(', ', $handle_list);
249
+
$handle_list = iterator_to_array($handles);
250
+
$handle_list = mpull($handle_list, 'getName');
251
+
$handle_list = implode(', ', $handle_list);
212
252
213
-
throw new PhutilArgumentUsageException(
214
-
pht(
215
-
'Repository "%s" has versions on inactive devices. Demote '.
216
-
'(or reactivate) these devices before promoting a new '.
217
-
'leader: %s.',
218
-
$repository->getDisplayName(),
219
-
$handle_list));
220
-
}
253
+
throw new PhutilArgumentUsageException(
254
+
pht(
255
+
'Repository "%s" has versions on inactive devices. Demote '.
256
+
'(or reactivate) these devices before promoting a new '.
257
+
'leader: %s.',
258
+
$repository->getDisplayName(),
259
+
$handle_list));
260
+
}
221
261
222
-
// Now, make sure there are no outstanding versions on devices with
223
-
// active bindings. These also need to be demoted (or promoting is a
224
-
// mistake or already happened).
225
-
$active = array_select_keys($versions, array_keys($bindings));
226
-
if ($active) {
227
-
$handles = $viewer->loadHandles(array_keys($active));
262
+
// Now, make sure there are no outstanding versions on devices with
263
+
// active bindings. These also need to be demoted (or promoting is
264
+
// a mistake or already happened).
265
+
$active = array_select_keys($versions, array_keys($bindings));
266
+
if ($active) {
267
+
$handles = $viewer->loadHandles(array_keys($active));
268
+
269
+
$handle_list = iterator_to_array($handles);
270
+
$handle_list = mpull($handle_list, 'getName');
271
+
$handle_list = implode(', ', $handle_list);
272
+
273
+
throw new PhutilArgumentUsageException(
274
+
pht(
275
+
'Unable to promote "%s" for repository "%s" because this '.
276
+
'cluster already has one or more unambiguous leaders: %s.',
277
+
$device->getName(),
278
+
$repository->getDisplayName(),
279
+
$handle_list));
280
+
}
228
281
229
-
$handle_list = iterator_to_array($handles);
230
-
$handle_list = mpull($handle_list, 'getName');
231
-
$handle_list = implode(', ', $handle_list);
282
+
PhabricatorRepositoryWorkingCopyVersion::updateVersion(
283
+
$repository->getPHID(),
284
+
$device->getPHID(),
285
+
0);
232
286
233
-
throw new PhutilArgumentUsageException(
287
+
echo tsprintf(
288
+
"%s\n",
234
289
pht(
235
-
'Unable to promote "%s" for repository "%s" because this '.
236
-
'cluster already has one or more unambiguous leaders: %s.',
290
+
'Promoted "%s" to become a leader for "%s".',
237
291
$device->getName(),
238
-
$repository->getDisplayName(),
239
-
$handle_list));
292
+
$repository->getDisplayName()));
240
293
}
241
294
242
-
PhabricatorRepositoryWorkingCopyVersion::updateVersion(
243
-
$repository->getPHID(),
244
-
$device->getPHID(),
245
-
0);
295
+
if ($demote) {
296
+
PhabricatorRepositoryWorkingCopyVersion::demoteDevice(
297
+
$repository->getPHID(),
298
+
$device->getPHID());
246
299
247
-
echo tsprintf(
248
-
"%s\n",
249
-
pht(
250
-
'Promoted "%s" to become a leader for "%s".',
251
-
$device->getName(),
252
-
$repository->getDisplayName()));
300
+
echo tsprintf(
301
+
"%s\n",
302
+
pht(
303
+
'Demoted "%s" from leadership of repository "%s".',
304
+
$device->getName(),
305
+
$repository->getDisplayName()));
306
+
}
307
+
} catch (Exception $ex) {
308
+
$write_lock->unlock();
309
+
throw $ex;
253
310
}
254
311
255
-
if ($demote) {
256
-
PhabricatorRepositoryWorkingCopyVersion::demoteDevice(
257
-
$repository->getPHID(),
258
-
$device->getPHID());
259
-
260
-
echo tsprintf(
261
-
"%s\n",
262
-
pht(
263
-
'Demoted "%s" from leadership of repository "%s".',
264
-
$device->getName(),
265
-
$repository->getDisplayName()));
266
-
}
267
-
} catch (Exception $ex) {
268
312
$write_lock->unlock();
269
-
throw $ex;
270
313
}
271
-
272
-
$write_lock->unlock();
273
314
}
274
315
275
316
return 0;
+12
src/docs/user/cluster/cluster_repositories.diviner
+12
src/docs/user/cluster/cluster_repositories.diviner
···
433
433
changes on the affected leaders which have not replicated to other devices
434
434
in the cluster.
435
435
436
+
If you have lost an entire cluster and replaced it with new devices that you
437
+
have restored from backups, you can aggressively wipe all memory of the old
438
+
devices by using `--demote <service>` and `--all-repositories`. **This is
439
+
dangerous and discards all unreplicated data in any repository on any device.**
440
+
441
+
```
442
+
phabricator/ $ ./bin/repository thaw --demote repo.corp.net --all-repositories
443
+
```
444
+
445
+
After you do this, continue below to promote a leader and restore the cluster
446
+
to service.
447
+
436
448
437
449
Ambiguous Leaders
438
450
=================