?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ctdbd_conn: Use sys_poll_intr
[Samba.git] / ctdb / server / ctdb_recover.c
blob102854564bc9e8fe97c48aa6b07bd5f6689f6125
1 /*
2 ctdb recovery code
3
4 Copyright (C) Andrew Tridgell 2007
5 Copyright (C) Ronnie Sahlberg 2007
6
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
11
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
16
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, see <http://www.gnu.org/licenses/>.
19 */
20 #include "replace.h"
21 #include "system/time.h"
22 #include "system/network.h"
23 #include "system/filesys.h"
24 #include "system/wait.h"
25
26 #include <talloc.h>
27 #include <tevent.h>
28 #include <tdb.h>
29
30 #include "lib/tdb_wrap/tdb_wrap.h"
31 #include "lib/util/dlinklist.h"
32 #include "lib/util/debug.h"
33 #include "lib/util/time.h"
34 #include "lib/util/util_process.h"
35
36 #include "ctdb_private.h"
37 #include "ctdb_client.h"
38
39 #include "common/system.h"
40 #include "common/common.h"
41 #include "common/logging.h"
42
43 int
44 ctdb_control_getvnnmap(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
45 {
46 struct ctdb_vnn_map_wire *map;
47 size_t len;
48
49 CHECK_CONTROL_DATA_SIZE(0);
50
51 len = offsetof(struct ctdb_vnn_map_wire, map) + sizeof(uint32_t)*ctdb->vnn_map->size;
52 map = talloc_size(outdata, len);
53 CTDB_NO_MEMORY(ctdb, map);
54
55 map->generation = ctdb->vnn_map->generation;
56 map->size = ctdb->vnn_map->size;
57 memcpy(map->map, ctdb->vnn_map->map, sizeof(uint32_t)*map->size);
58
59 outdata->dsize = len;
60 outdata->dptr = (uint8_t *)map;
61
62 return 0;
63 }
64
65 int
66 ctdb_control_setvnnmap(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
67 {
68 struct ctdb_vnn_map_wire *map = (struct ctdb_vnn_map_wire *)indata.dptr;
69
70 if (ctdb->recovery_mode != CTDB_RECOVERY_ACTIVE) {
71 DEBUG(DEBUG_ERR, ("Attempt to set vnnmap when not in recovery\n"));
72 return -1;
73 }
74
75 talloc_free(ctdb->vnn_map);
76
77 ctdb->vnn_map = talloc(ctdb, struct ctdb_vnn_map);
78 CTDB_NO_MEMORY(ctdb, ctdb->vnn_map);
79
80 ctdb->vnn_map->generation = map->generation;
81 ctdb->vnn_map->size = map->size;
82 ctdb->vnn_map->map = talloc_array(ctdb->vnn_map, uint32_t, map->size);
83 CTDB_NO_MEMORY(ctdb, ctdb->vnn_map->map);
84
85 memcpy(ctdb->vnn_map->map, map->map, sizeof(uint32_t)*map->size);
86
87 return 0;
88 }
89
90 int
91 ctdb_control_getdbmap(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
92 {
93 uint32_t i, len;
94 struct ctdb_db_context *ctdb_db;
95 struct ctdb_dbid_map_old *dbid_map;
96
97 CHECK_CONTROL_DATA_SIZE(0);
98
99 len = 0;
100 for(ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next){
101 len++;
102 }
103
104
105 outdata->dsize = offsetof(struct ctdb_dbid_map_old, dbs) + sizeof(dbid_map->dbs[0])*len;
106 outdata->dptr = (unsigned char *)talloc_zero_size(outdata, outdata->dsize);
107 if (!outdata->dptr) {
108 DEBUG(DEBUG_ALERT, (__location__ " Failed to allocate dbmap array\n"));
109 exit(1);
110 }
111
112 dbid_map = (struct ctdb_dbid_map_old *)outdata->dptr;
113 dbid_map->num = len;
114 for (i=0,ctdb_db=ctdb->db_list;ctdb_db;i++,ctdb_db=ctdb_db->next){
115 dbid_map->dbs[i].db_id = ctdb_db->db_id;
116 if (ctdb_db->persistent != 0) {
117 dbid_map->dbs[i].flags |= CTDB_DB_FLAGS_PERSISTENT;
118 }
119 if (ctdb_db->readonly != 0) {
120 dbid_map->dbs[i].flags |= CTDB_DB_FLAGS_READONLY;
121 }
122 if (ctdb_db->sticky != 0) {
123 dbid_map->dbs[i].flags |= CTDB_DB_FLAGS_STICKY;
124 }
125 }
126
127 return 0;
128 }
129
130 int
131 ctdb_control_getnodemap(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
132 {
133 CHECK_CONTROL_DATA_SIZE(0);
134
135 outdata->dptr = (unsigned char *)ctdb_node_list_to_map(ctdb->nodes,
136 ctdb->num_nodes,
137 outdata);
138 if (outdata->dptr == NULL) {
139 return -1;
140 }
141
142 outdata->dsize = talloc_get_size(outdata->dptr);
143
144 return 0;
145 }
146
147 /*
148 reload the nodes file
149 */
150 int
151 ctdb_control_reload_nodes_file(struct ctdb_context *ctdb, uint32_t opcode)
152 {
153 int i, num_nodes;
154 TALLOC_CTX *tmp_ctx;
155 struct ctdb_node **nodes;
156
157 tmp_ctx = talloc_new(ctdb);
158
159 /* steal the old nodes file for a while */
160 talloc_steal(tmp_ctx, ctdb->nodes);
161 nodes = ctdb->nodes;
162 ctdb->nodes = NULL;
163 num_nodes = ctdb->num_nodes;
164 ctdb->num_nodes = 0;
165
166 /* load the new nodes file */
167 ctdb_load_nodes_file(ctdb);
168
169 for (i=0; i<ctdb->num_nodes; i++) {
170 /* keep any identical pre-existing nodes and connections */
171 if ((i < num_nodes) && ctdb_same_address(&ctdb->nodes[i]->address, &nodes[i]->address)) {
172 talloc_free(ctdb->nodes[i]);
173 ctdb->nodes[i] = talloc_steal(ctdb->nodes, nodes[i]);
174 continue;
175 }
176
177 if (ctdb->nodes[i]->flags & NODE_FLAGS_DELETED) {
178 continue;
179 }
180
181 /* any new or different nodes must be added */
182 if (ctdb->methods->add_node(ctdb->nodes[i]) != 0) {
183 DEBUG(DEBUG_CRIT, (__location__ " methods->add_node failed at %d\n", i));
184 ctdb_fatal(ctdb, "failed to add node. shutting down\n");
185 }
186 if (ctdb->methods->connect_node(ctdb->nodes[i]) != 0) {
187 DEBUG(DEBUG_CRIT, (__location__ " methods->add_connect failed at %d\n", i));
188 ctdb_fatal(ctdb, "failed to connect to node. shutting down\n");
189 }
190 }
191
192 /* tell the recovery daemon to reaload the nodes file too */
193 ctdb_daemon_send_message(ctdb, ctdb->pnn, CTDB_SRVID_RELOAD_NODES, tdb_null);
194
195 talloc_free(tmp_ctx);
196
197 return 0;
198 }
199
200 /*
201 a traverse function for pulling all relevent records from pulldb
202 */
203 struct pulldb_data {
204 struct ctdb_context *ctdb;
205 struct ctdb_db_context *ctdb_db;
206 struct ctdb_marshall_buffer *pulldata;
207 uint32_t len;
208 uint32_t allocated_len;
209 bool failed;
210 };
211
212 static int traverse_pulldb(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data, void *p)
213 {
214 struct pulldb_data *params = (struct pulldb_data *)p;
215 struct ctdb_rec_data_old *rec;
216 struct ctdb_context *ctdb = params->ctdb;
217 struct ctdb_db_context *ctdb_db = params->ctdb_db;
218
219 /* add the record to the blob */
220 rec = ctdb_marshall_record(params->pulldata, 0, key, NULL, data);
221 if (rec == NULL) {
222 params->failed = true;
223 return -1;
224 }
225 if (params->len + rec->length >= params->allocated_len) {
226 params->allocated_len = rec->length + params->len + ctdb->tunable.pulldb_preallocation_size;
227 params->pulldata = talloc_realloc_size(NULL, params->pulldata, params->allocated_len);
228 }
229 if (params->pulldata == NULL) {
230 DEBUG(DEBUG_CRIT,(__location__ " Failed to expand pulldb_data to %u\n", rec->length + params->len));
231 ctdb_fatal(params->ctdb, "failed to allocate memory for recovery. shutting down\n");
232 }
233 params->pulldata->count++;
234 memcpy(params->len+(uint8_t *)params->pulldata, rec, rec->length);
235 params->len += rec->length;
236
237 if (ctdb->tunable.db_record_size_warn != 0 && rec->length > ctdb->tunable.db_record_size_warn) {
238 DEBUG(DEBUG_ERR,("Data record in %s is big. Record size is %d bytes\n", ctdb_db->db_name, (int)rec->length));
239 }
240
241 talloc_free(rec);
242
243 return 0;
244 }
245
246 /*
247 pull a bunch of records from a ltdb, filtering by lmaster
248 */
249 int32_t ctdb_control_pull_db(struct ctdb_context *ctdb, TDB_DATA indata, TDB_DATA *outdata)
250 {
251 struct ctdb_pulldb *pull;
252 struct ctdb_db_context *ctdb_db;
253 struct pulldb_data params;
254 struct ctdb_marshall_buffer *reply;
255
256 pull = (struct ctdb_pulldb *)indata.dptr;
257
258 ctdb_db = find_ctdb_db(ctdb, pull->db_id);
259 if (!ctdb_db) {
260 DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n", pull->db_id));
261 return -1;
262 }
263
264 if (!ctdb_db_frozen(ctdb_db)) {
265 DEBUG(DEBUG_ERR,
266 ("rejecting ctdb_control_pull_db when not frozen\n"));
267 return -1;
268 }
269
270 reply = talloc_zero(outdata, struct ctdb_marshall_buffer);
271 CTDB_NO_MEMORY(ctdb, reply);
272
273 reply->db_id = pull->db_id;
274
275 params.ctdb = ctdb;
276 params.ctdb_db = ctdb_db;
277 params.pulldata = reply;
278 params.len = offsetof(struct ctdb_marshall_buffer, data);
279 params.allocated_len = params.len;
280 params.failed = false;
281
282 if (ctdb_db->unhealthy_reason) {
283 /* this is just a warning, as the tdb should be empty anyway */
284 DEBUG(DEBUG_WARNING,("db(%s) unhealty in ctdb_control_pull_db: %s\n",
285 ctdb_db->db_name, ctdb_db->unhealthy_reason));
286 }
287
288 if (ctdb_lockdb_mark(ctdb_db) != 0) {
289 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock on entire db - failing\n"));
290 return -1;
291 }
292
293 if (tdb_traverse_read(ctdb_db->ltdb->tdb, traverse_pulldb, &params) == -1) {
294 DEBUG(DEBUG_ERR,(__location__ " Failed to get traverse db '%s'\n", ctdb_db->db_name));
295 ctdb_lockdb_unmark(ctdb_db);
296 talloc_free(params.pulldata);
297 return -1;
298 }
299
300 ctdb_lockdb_unmark(ctdb_db);
301
302 outdata->dptr = (uint8_t *)params.pulldata;
303 outdata->dsize = params.len;
304
305 if (ctdb->tunable.db_record_count_warn != 0 && params.pulldata->count > ctdb->tunable.db_record_count_warn) {
306 DEBUG(DEBUG_ERR,("Database %s is big. Contains %d records\n", ctdb_db->db_name, params.pulldata->count));
307 }
308 if (ctdb->tunable.db_size_warn != 0 && outdata->dsize > ctdb->tunable.db_size_warn) {
309 DEBUG(DEBUG_ERR,("Database %s is big. Contains %d bytes\n", ctdb_db->db_name, (int)outdata->dsize));
310 }
311
312
313 return 0;
314 }
315
316 struct db_pull_state {
317 struct ctdb_context *ctdb;
318 struct ctdb_db_context *ctdb_db;
319 struct ctdb_marshall_buffer *recs;
320 uint32_t pnn;
321 uint64_t srvid;
322 uint32_t num_records;
323 };
324
325 static int traverse_db_pull(struct tdb_context *tdb, TDB_DATA key,
326 TDB_DATA data, void *private_data)
327 {
328 struct db_pull_state *state = (struct db_pull_state *)private_data;
329 struct ctdb_marshall_buffer *recs;
330
331 recs = ctdb_marshall_add(state->ctdb, state->recs,
332 state->ctdb_db->db_id, 0, key, NULL, data);
333 if (recs == NULL) {
334 TALLOC_FREE(state->recs);
335 return -1;
336 }
337 state->recs = recs;
338
339 if (talloc_get_size(state->recs) >=
340 state->ctdb->tunable.rec_buffer_size_limit) {
341 TDB_DATA buffer;
342 int ret;
343
344 buffer = ctdb_marshall_finish(state->recs);
345 ret = ctdb_daemon_send_message(state->ctdb, state->pnn,
346 state->srvid, buffer);
347 if (ret != 0) {
348 TALLOC_FREE(state->recs);
349 return -1;
350 }
351
352 state->num_records += state->recs->count;
353 TALLOC_FREE(state->recs);
354 }
355
356 return 0;
357 }
358
359 int32_t ctdb_control_db_pull(struct ctdb_context *ctdb,
360 struct ctdb_req_control_old *c,
361 TDB_DATA indata, TDB_DATA *outdata)
362 {
363 struct ctdb_pulldb_ext *pulldb_ext;
364 struct ctdb_db_context *ctdb_db;
365 struct db_pull_state state;
366 int ret;
367
368 pulldb_ext = (struct ctdb_pulldb_ext *)indata.dptr;
369
370 ctdb_db = find_ctdb_db(ctdb, pulldb_ext->db_id);
371 if (ctdb_db == NULL) {
372 DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n",
373 pulldb_ext->db_id));
374 return -1;
375 }
376
377 if (!ctdb_db_frozen(ctdb_db)) {
378 DEBUG(DEBUG_ERR,
379 ("rejecting ctdb_control_pull_db when not frozen\n"));
380 return -1;
381 }
382
383 if (ctdb_db->unhealthy_reason) {
384 /* this is just a warning, as the tdb should be empty anyway */
385 DEBUG(DEBUG_WARNING,
386 ("db(%s) unhealty in ctdb_control_db_pull: %s\n",
387 ctdb_db->db_name, ctdb_db->unhealthy_reason));
388 }
389
390 state.ctdb = ctdb;
391 state.ctdb_db = ctdb_db;
392 state.recs = NULL;
393 state.pnn = c->hdr.srcnode;
394 state.srvid = pulldb_ext->srvid;
395 state.num_records = 0;
396
397 if (ctdb_lockdb_mark(ctdb_db) != 0) {
398 DEBUG(DEBUG_ERR,
399 (__location__ " Failed to get lock on entire db - failing\n"));
400 return -1;
401 }
402
403 ret = tdb_traverse_read(ctdb_db->ltdb->tdb, traverse_db_pull, &state);
404 if (ret == -1) {
405 DEBUG(DEBUG_ERR,
406 (__location__ " Failed to get traverse db '%s'\n",
407 ctdb_db->db_name));
408 ctdb_lockdb_unmark(ctdb_db);
409 return -1;
410 }
411
412 /* Last few records */
413 if (state.recs != NULL) {
414 TDB_DATA buffer;
415
416 buffer = ctdb_marshall_finish(state.recs);
417 ret = ctdb_daemon_send_message(state.ctdb, state.pnn,
418 state.srvid, buffer);
419 if (ret != 0) {
420 TALLOC_FREE(state.recs);
421 ctdb_lockdb_unmark(ctdb_db);
422 return -1;
423 }
424
425 state.num_records += state.recs->count;
426 TALLOC_FREE(state.recs);
427 }
428
429 ctdb_lockdb_unmark(ctdb_db);
430
431 outdata->dptr = talloc_size(outdata, sizeof(uint32_t));
432 if (outdata->dptr == NULL) {
433 DEBUG(DEBUG_ERR, (__location__ " Memory allocation error\n"));
434 return -1;
435 }
436
437 memcpy(outdata->dptr, (uint8_t *)&state.num_records, sizeof(uint32_t));
438 outdata->dsize = sizeof(uint32_t);
439
440 return 0;
441 }
442
443 /*
444 push a bunch of records into a ltdb, filtering by rsn
445 */
446 int32_t ctdb_control_push_db(struct ctdb_context *ctdb, TDB_DATA indata)
447 {
448 struct ctdb_marshall_buffer *reply = (struct ctdb_marshall_buffer *)indata.dptr;
449 struct ctdb_db_context *ctdb_db;
450 int i, ret;
451 struct ctdb_rec_data_old *rec;
452
453 if (indata.dsize < offsetof(struct ctdb_marshall_buffer, data)) {
454 DEBUG(DEBUG_ERR,(__location__ " invalid data in pulldb reply\n"));
455 return -1;
456 }
457
458 ctdb_db = find_ctdb_db(ctdb, reply->db_id);
459 if (!ctdb_db) {
460 DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n", reply->db_id));
461 return -1;
462 }
463
464 if (!ctdb_db_frozen(ctdb_db)) {
465 DEBUG(DEBUG_ERR,
466 ("rejecting ctdb_control_push_db when not frozen\n"));
467 return -1;
468 }
469
470 if (ctdb_lockdb_mark(ctdb_db) != 0) {
471 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock on entire db - failing\n"));
472 return -1;
473 }
474
475 rec = (struct ctdb_rec_data_old *)&reply->data[0];
476
477 DEBUG(DEBUG_INFO,("starting push of %u records for dbid 0x%x\n",
478 reply->count, reply->db_id));
479
480 for (i=0;i<reply->count;i++) {
481 TDB_DATA key, data;
482 struct ctdb_ltdb_header *hdr;
483
484 key.dptr = &rec->data[0];
485 key.dsize = rec->keylen;
486 data.dptr = &rec->data[key.dsize];
487 data.dsize = rec->datalen;
488
489 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
490 DEBUG(DEBUG_CRIT,(__location__ " bad ltdb record\n"));
491 goto failed;
492 }
493 hdr = (struct ctdb_ltdb_header *)data.dptr;
494 /* strip off any read only record flags. All readonly records
495 are revoked implicitely by a recovery
496 */
497 hdr->flags &= ~CTDB_REC_RO_FLAGS;
498
499 data.dptr += sizeof(*hdr);
500 data.dsize -= sizeof(*hdr);
501
502 ret = ctdb_ltdb_store(ctdb_db, key, hdr, data);
503 if (ret != 0) {
504 DEBUG(DEBUG_CRIT, (__location__ " Unable to store record\n"));
505 goto failed;
506 }
507
508 rec = (struct ctdb_rec_data_old *)(rec->length + (uint8_t *)rec);
509 }
510
511 DEBUG(DEBUG_DEBUG,("finished push of %u records for dbid 0x%x\n",
512 reply->count, reply->db_id));
513
514 if (ctdb_db->readonly) {
515 DEBUG(DEBUG_CRIT,("Clearing the tracking database for dbid 0x%x\n",
516 ctdb_db->db_id));
517 if (tdb_wipe_all(ctdb_db->rottdb) != 0) {
518 DEBUG(DEBUG_ERR,("Failed to wipe tracking database for 0x%x. Dropping read-only delegation support\n", ctdb_db->db_id));
519 ctdb_db->readonly = false;
520 tdb_close(ctdb_db->rottdb);
521 ctdb_db->rottdb = NULL;
522 ctdb_db->readonly = false;
523 }
524 while (ctdb_db->revokechild_active != NULL) {
525 talloc_free(ctdb_db->revokechild_active);
526 }
527 }
528
529 ctdb_lockdb_unmark(ctdb_db);
530 return 0;
531
532 failed:
533 ctdb_lockdb_unmark(ctdb_db);
534 return -1;
535 }
536
537 struct db_push_state {
538 struct ctdb_context *ctdb;
539 struct ctdb_db_context *ctdb_db;
540 uint64_t srvid;
541 uint32_t num_records;
542 bool failed;
543 };
544
545 static void db_push_msg_handler(uint64_t srvid, TDB_DATA indata,
546 void *private_data)
547 {
548 struct db_push_state *state = talloc_get_type(
549 private_data, struct db_push_state);
550 struct ctdb_marshall_buffer *recs;
551 struct ctdb_rec_data_old *rec;
552 int i, ret;
553
554 if (state->failed) {
555 return;
556 }
557
558 recs = (struct ctdb_marshall_buffer *)indata.dptr;
559 rec = (struct ctdb_rec_data_old *)&recs->data[0];
560
561 DEBUG(DEBUG_INFO, ("starting push of %u records for dbid 0x%x\n",
562 recs->count, recs->db_id));
563
564 for (i=0; i<recs->count; i++) {
565 TDB_DATA key, data;
566 struct ctdb_ltdb_header *hdr;
567
568 key.dptr = &rec->data[0];
569 key.dsize = rec->keylen;
570 data.dptr = &rec->data[key.dsize];
571 data.dsize = rec->datalen;
572
573 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
574 DEBUG(DEBUG_CRIT,(__location__ " bad ltdb record\n"));
575 goto failed;
576 }
577
578 hdr = (struct ctdb_ltdb_header *)data.dptr;
579 /* Strip off any read only record flags.
580 * All readonly records are revoked implicitely by a recovery.
581 */
582 hdr->flags &= ~CTDB_REC_RO_FLAGS;
583
584 data.dptr += sizeof(*hdr);
585 data.dsize -= sizeof(*hdr);
586
587 ret = ctdb_ltdb_store(state->ctdb_db, key, hdr, data);
588 if (ret != 0) {
589 DEBUG(DEBUG_ERR,
590 (__location__ " Unable to store record\n"));
591 goto failed;
592 }
593
594 rec = (struct ctdb_rec_data_old *)(rec->length + (uint8_t *)rec);
595 }
596
597 DEBUG(DEBUG_DEBUG, ("finished push of %u records for dbid 0x%x\n",
598 recs->count, recs->db_id));
599
600 state->num_records += recs->count;
601 return;
602
603 failed:
604 state->failed = true;
605 }
606
607 int32_t ctdb_control_db_push_start(struct ctdb_context *ctdb, TDB_DATA indata)
608 {
609 struct ctdb_pulldb_ext *pulldb_ext;
610 struct ctdb_db_context *ctdb_db;
611 struct db_push_state *state;
612 int ret;
613
614 pulldb_ext = (struct ctdb_pulldb_ext *)indata.dptr;
615
616 ctdb_db = find_ctdb_db(ctdb, pulldb_ext->db_id);
617 if (ctdb_db == NULL) {
618 DEBUG(DEBUG_ERR,
619 (__location__ " Unknown db 0x%08x\n", pulldb_ext->db_id));
620 return -1;
621 }
622
623 if (!ctdb_db_frozen(ctdb_db)) {
624 DEBUG(DEBUG_ERR,
625 ("rejecting ctdb_control_db_push_start when not frozen\n"));
626 return -1;
627 }
628
629 if (ctdb_db->push_started) {
630 DEBUG(DEBUG_WARNING,
631 (__location__ " DB push already started for %s\n",
632 ctdb_db->db_name));
633
634 /* De-register old state */
635 state = (struct db_push_state *)ctdb_db->push_state;
636 if (state != NULL) {
637 srvid_deregister(ctdb->srv, state->srvid, state);
638 talloc_free(state);
639 ctdb_db->push_state = NULL;
640 }
641 }
642
643 state = talloc_zero(ctdb_db, struct db_push_state);
644 if (state == NULL) {
645 DEBUG(DEBUG_ERR, (__location__ " Memory allocation error\n"));
646 return -1;
647 }
648
649 state->ctdb = ctdb;
650 state->ctdb_db = ctdb_db;
651 state->srvid = pulldb_ext->srvid;
652 state->failed = false;
653
654 ret = srvid_register(ctdb->srv, state, state->srvid,
655 db_push_msg_handler, state);
656 if (ret != 0) {
657 DEBUG(DEBUG_ERR,
658 (__location__ " Failed to register srvid for db push\n"));
659 talloc_free(state);
660 return -1;
661 }
662
663 if (ctdb_lockdb_mark(ctdb_db) != 0) {
664 DEBUG(DEBUG_ERR,
665 (__location__ " Failed to get lock on entire db - failing\n"));
666 srvid_deregister(ctdb->srv, state->srvid, state);
667 talloc_free(state);
668 return -1;
669 }
670
671 ctdb_db->push_started = true;
672 ctdb_db->push_state = state;
673
674 return 0;
675 }
676
677 int32_t ctdb_control_db_push_confirm(struct ctdb_context *ctdb,
678 TDB_DATA indata, TDB_DATA *outdata)
679 {
680 uint32_t db_id;
681 struct ctdb_db_context *ctdb_db;
682 struct db_push_state *state;
683
684 db_id = *(uint32_t *)indata.dptr;
685
686 ctdb_db = find_ctdb_db(ctdb, db_id);
687 if (ctdb_db == NULL) {
688 DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n", db_id));
689 return -1;
690 }
691
692 if (!ctdb_db_frozen(ctdb_db)) {
693 DEBUG(DEBUG_ERR,
694 ("rejecting ctdb_control_db_push_confirm when not frozen\n"));
695 return -1;
696 }
697
698 if (!ctdb_db->push_started) {
699 DEBUG(DEBUG_ERR, (__location__ " DB push not started\n"));
700 return -1;
701 }
702
703 if (ctdb_db->readonly) {
704 DEBUG(DEBUG_ERR,
705 ("Clearing the tracking database for dbid 0x%x\n",
706 ctdb_db->db_id));
707 if (tdb_wipe_all(ctdb_db->rottdb) != 0) {
708 DEBUG(DEBUG_ERR,
709 ("Failed to wipe tracking database for 0x%x."
710 " Dropping read-only delegation support\n",
711 ctdb_db->db_id));
712 ctdb_db->readonly = false;
713 tdb_close(ctdb_db->rottdb);
714 ctdb_db->rottdb = NULL;
715 ctdb_db->readonly = false;
716 }
717
718 while (ctdb_db->revokechild_active != NULL) {
719 talloc_free(ctdb_db->revokechild_active);
720 }
721 }
722
723 ctdb_lockdb_unmark(ctdb_db);
724
725 state = (struct db_push_state *)ctdb_db->push_state;
726 if (state == NULL) {
727 DEBUG(DEBUG_ERR, (__location__ " Missing push db state\n"));
728 return -1;
729 }
730
731 srvid_deregister(ctdb->srv, state->srvid, state);
732
733 outdata->dptr = talloc_size(outdata, sizeof(uint32_t));
734 if (outdata->dptr == NULL) {
735 DEBUG(DEBUG_ERR, (__location__ " Memory allocation error\n"));
736 talloc_free(state);
737 ctdb_db->push_state = NULL;
738 return -1;
739 }
740
741 memcpy(outdata->dptr, (uint8_t *)&state->num_records, sizeof(uint32_t));
742 outdata->dsize = sizeof(uint32_t);
743
744 talloc_free(state);
745 ctdb_db->push_state = NULL;
746
747 return 0;
748 }
749
750 struct ctdb_set_recmode_state {
751 struct ctdb_context *ctdb;
752 struct ctdb_req_control_old *c;
753 int fd[2];
754 struct tevent_timer *te;
755 struct tevent_fd *fde;
756 pid_t child;
757 struct timeval start_time;
758 };
759
760 /*
761 called if our set_recmode child times out. this would happen if
762 ctdb_recovery_lock() would block.
763 */
764 static void ctdb_set_recmode_timeout(struct tevent_context *ev,
765 struct tevent_timer *te,
766 struct timeval t, void *private_data)
767 {
768 struct ctdb_set_recmode_state *state = talloc_get_type(private_data,
769 struct ctdb_set_recmode_state);
770
771 /* we consider this a success, not a failure, as we failed to
772 set the recovery lock which is what we wanted. This can be
773 caused by the cluster filesystem being very slow to
774 arbitrate locks immediately after a node failure.
775 */
776 DEBUG(DEBUG_ERR,(__location__ " set_recmode child process hung/timedout CFS slow to grant locks? (allowing recmode set anyway)\n"));
777 state->ctdb->recovery_mode = CTDB_RECOVERY_NORMAL;
778 ctdb_request_control_reply(state->ctdb, state->c, NULL, 0, NULL);
779 talloc_free(state);
780 }
781
782
783 /* when we free the recmode state we must kill any child process.
784 */
785 static int set_recmode_destructor(struct ctdb_set_recmode_state *state)
786 {
787 if (state->fd[0] != -1) {
788 state->fd[0] = -1;
789 }
790 ctdb_kill(state->ctdb, state->child, SIGKILL);
791 return 0;
792 }
793
794 /* this is called when the client process has completed ctdb_recovery_lock()
795 and has written data back to us through the pipe.
796 */
797 static void set_recmode_handler(struct tevent_context *ev,
798 struct tevent_fd *fde,
799 uint16_t flags, void *private_data)
800 {
801 struct ctdb_set_recmode_state *state= talloc_get_type(private_data,
802 struct ctdb_set_recmode_state);
803 char c = 0;
804 int ret;
805 int status = 0;
806 const char *err = NULL;
807
808 /* we got a response from our child process so we can abort the
809 timeout.
810 */
811 talloc_free(state->te);
812 state->te = NULL;
813
814 ret = sys_read(state->fd[0], &c, 1);
815 if (ret == 1) {
816 /* Child wrote status. EACCES indicates that it was unable
817 * to take the lock, which is the expected outcome.
818 * 0 indicates that it was able to take the
819 * lock, which is an error because the recovery daemon
820 * should be holding the lock. */
821 double l = timeval_elapsed(&state->start_time);
822
823 if (c == EACCES) {
824 status = 0;
825 err = NULL;
826
827 state->ctdb->recovery_mode = CTDB_RECOVERY_NORMAL;
828
829 /* release any deferred attach calls from clients */
830 ctdb_process_deferred_attach(state->ctdb);
831
832 CTDB_UPDATE_RECLOCK_LATENCY(state->ctdb, "daemon reclock", reclock.ctdbd, l);
833 } else {
834 status = -1;
835 err = "Took recovery lock from daemon during recovery - probably a cluster filesystem lock coherence problem";
836 }
837 } else {
838 /* Child did not write status. Unexpected error.
839 * Child may have received a signal. */
840 status = -1;
841 err = "Unexpected error when testing recovery lock";
842 }
843
844 ctdb_request_control_reply(state->ctdb, state->c, NULL, status, err);
845 talloc_free(state);
846 }
847
848 static void
849 ctdb_drop_all_ips_event(struct tevent_context *ev, struct tevent_timer *te,
850 struct timeval t, void *private_data)
851 {
852 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
853
854 DEBUG(DEBUG_ERR,(__location__ " Been in recovery mode for too long. Dropping all IPS\n"));
855 talloc_free(ctdb->release_ips_ctx);
856 ctdb->release_ips_ctx = NULL;
857
858 ctdb_release_all_ips(ctdb);
859 }
860
861 /*
862 * Set up an event to drop all public ips if we remain in recovery for too
863 * long
864 */
865 int ctdb_deferred_drop_all_ips(struct ctdb_context *ctdb)
866 {
867 if (ctdb->release_ips_ctx != NULL) {
868 talloc_free(ctdb->release_ips_ctx);
869 }
870 ctdb->release_ips_ctx = talloc_new(ctdb);
871 CTDB_NO_MEMORY(ctdb, ctdb->release_ips_ctx);
872
873 tevent_add_timer(ctdb->ev, ctdb->release_ips_ctx,
874 timeval_current_ofs(ctdb->tunable.recovery_drop_all_ips, 0),
875 ctdb_drop_all_ips_event, ctdb);
876 return 0;
877 }
878
879 /*
880 set the recovery mode
881 */
882 int32_t ctdb_control_set_recmode(struct ctdb_context *ctdb,
883 struct ctdb_req_control_old *c,
884 TDB_DATA indata, bool *async_reply,
885 const char **errormsg)
886 {
887 uint32_t recmode = *(uint32_t *)indata.dptr;
888 int i, ret;
889 struct ctdb_set_recmode_state *state;
890 pid_t parent = getpid();
891 struct ctdb_db_context *ctdb_db;
892
893 /* if we enter recovery but stay in recovery for too long
894 we will eventually drop all our ip addresses
895 */
896 if (recmode == CTDB_RECOVERY_NORMAL) {
897 talloc_free(ctdb->release_ips_ctx);
898 ctdb->release_ips_ctx = NULL;
899 } else {
900 if (ctdb_deferred_drop_all_ips(ctdb) != 0) {
901 DEBUG(DEBUG_ERR,("Failed to set up deferred drop all ips\n"));
902 }
903 }
904
905 if (recmode != ctdb->recovery_mode) {
906 DEBUG(DEBUG_NOTICE,(__location__ " Recovery mode set to %s\n",
907 recmode==CTDB_RECOVERY_NORMAL?"NORMAL":"ACTIVE"));
908 }
909
910 if (recmode != CTDB_RECOVERY_NORMAL ||
911 ctdb->recovery_mode != CTDB_RECOVERY_ACTIVE) {
912 ctdb->recovery_mode = recmode;
913 return 0;
914 }
915
916 /* From this point: recmode == CTDB_RECOVERY_NORMAL
917 *
918 * Therefore, what follows is special handling when setting
919 * recovery mode back to normal */
920
921 for (ctdb_db = ctdb->db_list; ctdb_db != NULL; ctdb_db = ctdb_db->next) {
922 if (ctdb_db->generation != ctdb->vnn_map->generation) {
923 DEBUG(DEBUG_ERR,
924 ("Inconsistent DB generation %u for %s\n",
925 ctdb_db->generation, ctdb_db->db_name));
926 DEBUG(DEBUG_ERR, ("Recovery mode set to ACTIVE\n"));
927 return -1;
928 }
929 }
930
931 /* force the databases to thaw */
932 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
933 if (ctdb_db_prio_frozen(ctdb, i)) {
934 ctdb_control_thaw(ctdb, i, false);
935 }
936 }
937
938 /* release any deferred attach calls from clients */
939 if (recmode == CTDB_RECOVERY_NORMAL) {
940 ctdb_process_deferred_attach(ctdb);
941 }
942
943 if (ctdb->recovery_lock_file == NULL) {
944 /* Not using recovery lock file */
945 ctdb->recovery_mode = recmode;
946 return 0;
947 }
948
949 state = talloc(ctdb, struct ctdb_set_recmode_state);
950 CTDB_NO_MEMORY(ctdb, state);
951
952 state->start_time = timeval_current();
953 state->fd[0] = -1;
954 state->fd[1] = -1;
955
956 /* For the rest of what needs to be done, we need to do this in
957 a child process since
958 1, the call to ctdb_recovery_lock() can block if the cluster
959 filesystem is in the process of recovery.
960 */
961 ret = pipe(state->fd);
962 if (ret != 0) {
963 talloc_free(state);
964 DEBUG(DEBUG_CRIT,(__location__ " Failed to open pipe for set_recmode child\n"));
965 return -1;
966 }
967
968 state->child = ctdb_fork(ctdb);
969 if (state->child == (pid_t)-1) {
970 close(state->fd[0]);
971 close(state->fd[1]);
972 talloc_free(state);
973 return -1;
974 }
975
976 if (state->child == 0) {
977 char cc = EACCES;
978 close(state->fd[0]);
979
980 prctl_set_comment("ctdb_recmode");
981 debug_extra = talloc_asprintf(NULL, "set_recmode:");
982 /* Daemon should not be able to get the recover lock,
983 * as it should be held by the recovery master */
984 if (ctdb_recovery_lock(ctdb)) {
985 DEBUG(DEBUG_ERR,
986 ("ERROR: Daemon able to take recovery lock on \"%s\" during recovery\n",
987 ctdb->recovery_lock_file));
988 ctdb_recovery_unlock(ctdb);
989 cc = 0;
990 }
991
992 sys_write(state->fd[1], &cc, 1);
993 ctdb_wait_for_process_to_exit(parent);
994 _exit(0);
995 }
996 close(state->fd[1]);
997 set_close_on_exec(state->fd[0]);
998
999 state->fd[1] = -1;
1000
1001 talloc_set_destructor(state, set_recmode_destructor);
1002
1003 DEBUG(DEBUG_DEBUG, (__location__ " Created PIPE FD:%d for setrecmode\n", state->fd[0]));
1004
1005 state->te = tevent_add_timer(ctdb->ev, state, timeval_current_ofs(5, 0),
1006 ctdb_set_recmode_timeout, state);
1007
1008 state->fde = tevent_add_fd(ctdb->ev, state, state->fd[0], TEVENT_FD_READ,
1009 set_recmode_handler, (void *)state);
1010
1011 if (state->fde == NULL) {
1012 talloc_free(state);
1013 return -1;
1014 }
1015 tevent_fd_set_auto_close(state->fde);
1016
1017 state->ctdb = ctdb;
1018 state->c = talloc_steal(state, c);
1019
1020 *async_reply = true;
1021
1022 return 0;
1023 }
1024
1025
1026 bool ctdb_recovery_have_lock(struct ctdb_context *ctdb)
1027 {
1028 return ctdb->recovery_lock_fd != -1;
1029 }
1030
1031 /*
1032 try and get the recovery lock in shared storage - should only work
1033 on the recovery master recovery daemon. Anywhere else is a bug
1034 */
1035 bool ctdb_recovery_lock(struct ctdb_context *ctdb)
1036 {
1037 struct flock lock;
1038
1039 ctdb->recovery_lock_fd = open(ctdb->recovery_lock_file,
1040 O_RDWR|O_CREAT, 0600);
1041 if (ctdb->recovery_lock_fd == -1) {
1042 DEBUG(DEBUG_ERR,
1043 ("ctdb_recovery_lock: Unable to open %s - (%s)\n",
1044 ctdb->recovery_lock_file, strerror(errno)));
1045 return false;
1046 }
1047
1048 set_close_on_exec(ctdb->recovery_lock_fd);
1049
1050 lock.l_type = F_WRLCK;
1051 lock.l_whence = SEEK_SET;
1052 lock.l_start = 0;
1053 lock.l_len = 1;
1054 lock.l_pid = 0;
1055
1056 if (fcntl(ctdb->recovery_lock_fd, F_SETLK, &lock) != 0) {
1057 int saved_errno = errno;
1058 close(ctdb->recovery_lock_fd);
1059 ctdb->recovery_lock_fd = -1;
1060 /* Fail silently on these errors, since they indicate
1061 * lock contention, but log an error for any other
1062 * failure. */
1063 if (saved_errno != EACCES &&
1064 saved_errno != EAGAIN) {
1065 DEBUG(DEBUG_ERR,("ctdb_recovery_lock: Failed to get "
1066 "recovery lock on '%s' - (%s)\n",
1067 ctdb->recovery_lock_file,
1068 strerror(saved_errno)));
1069 }
1070 return false;
1071 }
1072
1073 return true;
1074 }
1075
1076 void ctdb_recovery_unlock(struct ctdb_context *ctdb)
1077 {
1078 if (ctdb->recovery_lock_fd != -1) {
1079 DEBUG(DEBUG_NOTICE, ("Releasing recovery lock\n"));
1080 close(ctdb->recovery_lock_fd);
1081 ctdb->recovery_lock_fd = -1;
1082 }
1083 }
1084
1085 /*
1086 delete a record as part of the vacuum process
1087 only delete if we are not lmaster or dmaster, and our rsn is <= the provided rsn
1088 use non-blocking locks
1089
1090 return 0 if the record was successfully deleted (i.e. it does not exist
1091 when the function returns)
1092 or !0 is the record still exists in the tdb after returning.
1093 */
1094 static int delete_tdb_record(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, struct ctdb_rec_data_old *rec)
1095 {
1096 TDB_DATA key, data, data2;
1097 struct ctdb_ltdb_header *hdr, *hdr2;
1098
1099 /* these are really internal tdb functions - but we need them here for
1100 non-blocking lock of the freelist */
1101 int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype);
1102 int tdb_unlock(struct tdb_context *tdb, int list, int ltype);
1103
1104
1105 key.dsize = rec->keylen;
1106 key.dptr = &rec->data[0];
1107 data.dsize = rec->datalen;
1108 data.dptr = &rec->data[rec->keylen];
1109
1110 if (ctdb_lmaster(ctdb, &key) == ctdb->pnn) {
1111 DEBUG(DEBUG_INFO,(__location__ " Called delete on record where we are lmaster\n"));
1112 return -1;
1113 }
1114
1115 if (data.dsize != sizeof(struct ctdb_ltdb_header)) {
1116 DEBUG(DEBUG_ERR,(__location__ " Bad record size\n"));
1117 return -1;
1118 }
1119
1120 hdr = (struct ctdb_ltdb_header *)data.dptr;
1121
1122 /* use a non-blocking lock */
1123 if (tdb_chainlock_nonblock(ctdb_db->ltdb->tdb, key) != 0) {
1124 return -1;
1125 }
1126
1127 data2 = tdb_fetch(ctdb_db->ltdb->tdb, key);
1128 if (data2.dptr == NULL) {
1129 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1130 return 0;
1131 }
1132
1133 if (data2.dsize < sizeof(struct ctdb_ltdb_header)) {
1134 if (tdb_lock_nonblock(ctdb_db->ltdb->tdb, -1, F_WRLCK) == 0) {
1135 if (tdb_delete(ctdb_db->ltdb->tdb, key) != 0) {
1136 DEBUG(DEBUG_CRIT,(__location__ " Failed to delete corrupt record\n"));
1137 }
1138 tdb_unlock(ctdb_db->ltdb->tdb, -1, F_WRLCK);
1139 DEBUG(DEBUG_CRIT,(__location__ " Deleted corrupt record\n"));
1140 }
1141 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1142 free(data2.dptr);
1143 return 0;
1144 }
1145
1146 hdr2 = (struct ctdb_ltdb_header *)data2.dptr;
1147
1148 if (hdr2->rsn > hdr->rsn) {
1149 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1150 DEBUG(DEBUG_INFO,(__location__ " Skipping record with rsn=%llu - called with rsn=%llu\n",
1151 (unsigned long long)hdr2->rsn, (unsigned long long)hdr->rsn));
1152 free(data2.dptr);
1153 return -1;
1154 }
1155
1156 /* do not allow deleting record that have readonly flags set. */
1157 if (hdr->flags & CTDB_REC_RO_FLAGS) {
1158 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1159 DEBUG(DEBUG_INFO,(__location__ " Skipping record with readonly flags set\n"));
1160 free(data2.dptr);
1161 return -1;
1162 }
1163 if (hdr2->flags & CTDB_REC_RO_FLAGS) {
1164 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1165 DEBUG(DEBUG_INFO,(__location__ " Skipping record with readonly flags set\n"));
1166 free(data2.dptr);
1167 return -1;
1168 }
1169
1170 if (hdr2->dmaster == ctdb->pnn) {
1171 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1172 DEBUG(DEBUG_INFO,(__location__ " Attempted delete record where we are the dmaster\n"));
1173 free(data2.dptr);
1174 return -1;
1175 }
1176
1177 if (tdb_lock_nonblock(ctdb_db->ltdb->tdb, -1, F_WRLCK) != 0) {
1178 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1179 free(data2.dptr);
1180 return -1;
1181 }
1182
1183 if (tdb_delete(ctdb_db->ltdb->tdb, key) != 0) {
1184 tdb_unlock(ctdb_db->ltdb->tdb, -1, F_WRLCK);
1185 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1186 DEBUG(DEBUG_INFO,(__location__ " Failed to delete record\n"));
1187 free(data2.dptr);
1188 return -1;
1189 }
1190
1191 tdb_unlock(ctdb_db->ltdb->tdb, -1, F_WRLCK);
1192 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1193 free(data2.dptr);
1194 return 0;
1195 }
1196
1197
1198
1199 struct recovery_callback_state {
1200 struct ctdb_req_control_old *c;
1201 };
1202
1203
1204 /*
1205 called when the 'recovered' event script has finished
1206 */
1207 static void ctdb_end_recovery_callback(struct ctdb_context *ctdb, int status, void *p)
1208 {
1209 struct recovery_callback_state *state = talloc_get_type(p, struct recovery_callback_state);
1210
1211 ctdb_enable_monitoring(ctdb);
1212 CTDB_INCREMENT_STAT(ctdb, num_recoveries);
1213
1214 if (status != 0) {
1215 DEBUG(DEBUG_ERR,(__location__ " recovered event script failed (status %d)\n", status));
1216 if (status == -ETIME) {
1217 ctdb_ban_self(ctdb);
1218 }
1219 }
1220
1221 ctdb_request_control_reply(ctdb, state->c, NULL, status, NULL);
1222 talloc_free(state);
1223
1224 gettimeofday(&ctdb->last_recovery_finished, NULL);
1225
1226 if (ctdb->runstate == CTDB_RUNSTATE_FIRST_RECOVERY) {
1227 ctdb_set_runstate(ctdb, CTDB_RUNSTATE_STARTUP);
1228 }
1229 }
1230
1231 /*
1232 recovery has finished
1233 */
1234 int32_t ctdb_control_end_recovery(struct ctdb_context *ctdb,
1235 struct ctdb_req_control_old *c,
1236 bool *async_reply)
1237 {
1238 int ret;
1239 struct recovery_callback_state *state;
1240
1241 DEBUG(DEBUG_NOTICE,("Recovery has finished\n"));
1242
1243 ctdb_persistent_finish_trans3_commits(ctdb);
1244
1245 state = talloc(ctdb, struct recovery_callback_state);
1246 CTDB_NO_MEMORY(ctdb, state);
1247
1248 state->c = c;
1249
1250 ctdb_disable_monitoring(ctdb);
1251
1252 ret = ctdb_event_script_callback(ctdb, state,
1253 ctdb_end_recovery_callback,
1254 state,
1255 CTDB_EVENT_RECOVERED, "%s", "");
1256
1257 if (ret != 0) {
1258 ctdb_enable_monitoring(ctdb);
1259
1260 DEBUG(DEBUG_ERR,(__location__ " Failed to end recovery\n"));
1261 talloc_free(state);
1262 return -1;
1263 }
1264
1265 /* tell the control that we will be reply asynchronously */
1266 state->c = talloc_steal(state, c);
1267 *async_reply = true;
1268 return 0;
1269 }
1270
1271 /*
1272 called when the 'startrecovery' event script has finished
1273 */
1274 static void ctdb_start_recovery_callback(struct ctdb_context *ctdb, int status, void *p)
1275 {
1276 struct recovery_callback_state *state = talloc_get_type(p, struct recovery_callback_state);
1277
1278 if (status != 0) {
1279 DEBUG(DEBUG_ERR,(__location__ " startrecovery event script failed (status %d)\n", status));
1280 }
1281
1282 ctdb_request_control_reply(ctdb, state->c, NULL, status, NULL);
1283 talloc_free(state);
1284 }
1285
1286 /*
1287 run the startrecovery eventscript
1288 */
1289 int32_t ctdb_control_start_recovery(struct ctdb_context *ctdb,
1290 struct ctdb_req_control_old *c,
1291 bool *async_reply)
1292 {
1293 int ret;
1294 struct recovery_callback_state *state;
1295
1296 DEBUG(DEBUG_NOTICE,(__location__ " startrecovery eventscript has been invoked\n"));
1297 gettimeofday(&ctdb->last_recovery_started, NULL);
1298
1299 state = talloc(ctdb, struct recovery_callback_state);
1300 CTDB_NO_MEMORY(ctdb, state);
1301
1302 state->c = talloc_steal(state, c);
1303
1304 ctdb_disable_monitoring(ctdb);
1305
1306 ret = ctdb_event_script_callback(ctdb, state,
1307 ctdb_start_recovery_callback,
1308 state,
1309 CTDB_EVENT_START_RECOVERY,
1310 "%s", "");
1311
1312 if (ret != 0) {
1313 DEBUG(DEBUG_ERR,(__location__ " Failed to start recovery\n"));
1314 talloc_free(state);
1315 return -1;
1316 }
1317
1318 /* tell the control that we will be reply asynchronously */
1319 *async_reply = true;
1320 return 0;
1321 }
1322
1323 /*
1324 try to delete all these records as part of the vacuuming process
1325 and return the records we failed to delete
1326 */
1327 int32_t ctdb_control_try_delete_records(struct ctdb_context *ctdb, TDB_DATA indata, TDB_DATA *outdata)
1328 {
1329 struct ctdb_marshall_buffer *reply = (struct ctdb_marshall_buffer *)indata.dptr;
1330 struct ctdb_db_context *ctdb_db;
1331 int i;
1332 struct ctdb_rec_data_old *rec;
1333 struct ctdb_marshall_buffer *records;
1334
1335 if (indata.dsize < offsetof(struct ctdb_marshall_buffer, data)) {
1336 DEBUG(DEBUG_ERR,(__location__ " invalid data in try_delete_records\n"));
1337 return -1;
1338 }
1339
1340 ctdb_db = find_ctdb_db(ctdb, reply->db_id);
1341 if (!ctdb_db) {
1342 DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n", reply->db_id));
1343 return -1;
1344 }
1345
1346
1347 DEBUG(DEBUG_DEBUG,("starting try_delete_records of %u records for dbid 0x%x\n",
1348 reply->count, reply->db_id));
1349
1350
1351 /* create a blob to send back the records we couldnt delete */
1352 records = (struct ctdb_marshall_buffer *)
1353 talloc_zero_size(outdata,
1354 offsetof(struct ctdb_marshall_buffer, data));
1355 if (records == NULL) {
1356 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
1357 return -1;
1358 }
1359 records->db_id = ctdb_db->db_id;
1360
1361
1362 rec = (struct ctdb_rec_data_old *)&reply->data[0];
1363 for (i=0;i<reply->count;i++) {
1364 TDB_DATA key, data;
1365
1366 key.dptr = &rec->data[0];
1367 key.dsize = rec->keylen;
1368 data.dptr = &rec->data[key.dsize];
1369 data.dsize = rec->datalen;
1370
1371 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
1372 DEBUG(DEBUG_CRIT,(__location__ " bad ltdb record in indata\n"));
1373 return -1;
1374 }
1375
1376 /* If we cant delete the record we must add it to the reply
1377 so the lmaster knows it may not purge this record
1378 */
1379 if (delete_tdb_record(ctdb, ctdb_db, rec) != 0) {
1380 size_t old_size;
1381 struct ctdb_ltdb_header *hdr;
1382
1383 hdr = (struct ctdb_ltdb_header *)data.dptr;
1384 data.dptr += sizeof(*hdr);
1385 data.dsize -= sizeof(*hdr);
1386
1387 DEBUG(DEBUG_INFO, (__location__ " Failed to vacuum delete record with hash 0x%08x\n", ctdb_hash(&key)));
1388
1389 old_size = talloc_get_size(records);
1390 records = talloc_realloc_size(outdata, records, old_size + rec->length);
1391 if (records == NULL) {
1392 DEBUG(DEBUG_ERR,(__location__ " Failed to expand\n"));
1393 return -1;
1394 }
1395 records->count++;
1396 memcpy(old_size+(uint8_t *)records, rec, rec->length);
1397 }
1398
1399 rec = (struct ctdb_rec_data_old *)(rec->length + (uint8_t *)rec);
1400 }
1401
1402
1403 *outdata = ctdb_marshall_finish(records);
1404
1405 return 0;
1406 }
1407
1408 /**
1409 * Store a record as part of the vacuum process:
1410 * This is called from the RECEIVE_RECORD control which
1411 * the lmaster uses to send the current empty copy
1412 * to all nodes for storing, before it lets the other
1413 * nodes delete the records in the second phase with
1414 * the TRY_DELETE_RECORDS control.
1415 *
1416 * Only store if we are not lmaster or dmaster, and our
1417 * rsn is <= the provided rsn. Use non-blocking locks.
1418 *
1419 * return 0 if the record was successfully stored.
1420 * return !0 if the record still exists in the tdb after returning.
1421 */
1422 static int store_tdb_record(struct ctdb_context *ctdb,
1423 struct ctdb_db_context *ctdb_db,
1424 struct ctdb_rec_data_old *rec)
1425 {
1426 TDB_DATA key, data, data2;
1427 struct ctdb_ltdb_header *hdr, *hdr2;
1428 int ret;
1429
1430 key.dsize = rec->keylen;
1431 key.dptr = &rec->data[0];
1432 data.dsize = rec->datalen;
1433 data.dptr = &rec->data[rec->keylen];
1434
1435 if (ctdb_lmaster(ctdb, &key) == ctdb->pnn) {
1436 DEBUG(DEBUG_INFO, (__location__ " Called store_tdb_record "
1437 "where we are lmaster\n"));
1438 return -1;
1439 }
1440
1441 if (data.dsize != sizeof(struct ctdb_ltdb_header)) {
1442 DEBUG(DEBUG_ERR, (__location__ " Bad record size\n"));
1443 return -1;
1444 }
1445
1446 hdr = (struct ctdb_ltdb_header *)data.dptr;
1447
1448 /* use a non-blocking lock */
1449 if (tdb_chainlock_nonblock(ctdb_db->ltdb->tdb, key) != 0) {
1450 DEBUG(DEBUG_INFO, (__location__ " Failed to lock chain in non-blocking mode\n"));
1451 return -1;
1452 }
1453
1454 data2 = tdb_fetch(ctdb_db->ltdb->tdb, key);
1455 if (data2.dptr == NULL || data2.dsize < sizeof(struct ctdb_ltdb_header)) {
1456 if (tdb_store(ctdb_db->ltdb->tdb, key, data, 0) == -1) {
1457 DEBUG(DEBUG_ERR, (__location__ "Failed to store record\n"));
1458 ret = -1;
1459 goto done;
1460 }
1461 DEBUG(DEBUG_INFO, (__location__ " Stored record\n"));
1462 ret = 0;
1463 goto done;
1464 }
1465
1466 hdr2 = (struct ctdb_ltdb_header *)data2.dptr;
1467
1468 if (hdr2->rsn > hdr->rsn) {
1469 DEBUG(DEBUG_INFO, (__location__ " Skipping record with "
1470 "rsn=%llu - called with rsn=%llu\n",
1471 (unsigned long long)hdr2->rsn,
1472 (unsigned long long)hdr->rsn));
1473 ret = -1;
1474 goto done;
1475 }
1476
1477 /* do not allow vacuuming of records that have readonly flags set. */
1478 if (hdr->flags & CTDB_REC_RO_FLAGS) {
1479 DEBUG(DEBUG_INFO,(__location__ " Skipping record with readonly "
1480 "flags set\n"));
1481 ret = -1;
1482 goto done;
1483 }
1484 if (hdr2->flags & CTDB_REC_RO_FLAGS) {
1485 DEBUG(DEBUG_INFO,(__location__ " Skipping record with readonly "
1486 "flags set\n"));
1487 ret = -1;
1488 goto done;
1489 }
1490
1491 if (hdr2->dmaster == ctdb->pnn) {
1492 DEBUG(DEBUG_INFO, (__location__ " Attempted to store record "
1493 "where we are the dmaster\n"));
1494 ret = -1;
1495 goto done;
1496 }
1497
1498 if (tdb_store(ctdb_db->ltdb->tdb, key, data, 0) != 0) {
1499 DEBUG(DEBUG_INFO,(__location__ " Failed to store record\n"));
1500 ret = -1;
1501 goto done;
1502 }
1503
1504 ret = 0;
1505
1506 done:
1507 tdb_chainunlock(ctdb_db->ltdb->tdb, key);
1508 free(data2.dptr);
1509 return ret;
1510 }
1511
1512
1513
1514 /**
1515 * Try to store all these records as part of the vacuuming process
1516 * and return the records we failed to store.
1517 */
1518 int32_t ctdb_control_receive_records(struct ctdb_context *ctdb,
1519 TDB_DATA indata, TDB_DATA *outdata)
1520 {
1521 struct ctdb_marshall_buffer *reply = (struct ctdb_marshall_buffer *)indata.dptr;
1522 struct ctdb_db_context *ctdb_db;
1523 int i;
1524 struct ctdb_rec_data_old *rec;
1525 struct ctdb_marshall_buffer *records;
1526
1527 if (indata.dsize < offsetof(struct ctdb_marshall_buffer, data)) {
1528 DEBUG(DEBUG_ERR,
1529 (__location__ " invalid data in receive_records\n"));
1530 return -1;
1531 }
1532
1533 ctdb_db = find_ctdb_db(ctdb, reply->db_id);
1534 if (!ctdb_db) {
1535 DEBUG(DEBUG_ERR, (__location__ " Unknown db 0x%08x\n",
1536 reply->db_id));
1537 return -1;
1538 }
1539
1540 DEBUG(DEBUG_DEBUG, ("starting receive_records of %u records for "
1541 "dbid 0x%x\n", reply->count, reply->db_id));
1542
1543 /* create a blob to send back the records we could not store */
1544 records = (struct ctdb_marshall_buffer *)
1545 talloc_zero_size(outdata,
1546 offsetof(struct ctdb_marshall_buffer, data));
1547 if (records == NULL) {
1548 DEBUG(DEBUG_ERR, (__location__ " Out of memory\n"));
1549 return -1;
1550 }
1551 records->db_id = ctdb_db->db_id;
1552
1553 rec = (struct ctdb_rec_data_old *)&reply->data[0];
1554 for (i=0; i<reply->count; i++) {
1555 TDB_DATA key, data;
1556
1557 key.dptr = &rec->data[0];
1558 key.dsize = rec->keylen;
1559 data.dptr = &rec->data[key.dsize];
1560 data.dsize = rec->datalen;
1561
1562 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
1563 DEBUG(DEBUG_CRIT, (__location__ " bad ltdb record "
1564 "in indata\n"));
1565 return -1;
1566 }
1567
1568 /*
1569 * If we can not store the record we must add it to the reply
1570 * so the lmaster knows it may not purge this record.
1571 */
1572 if (store_tdb_record(ctdb, ctdb_db, rec) != 0) {
1573 size_t old_size;
1574 struct ctdb_ltdb_header *hdr;
1575
1576 hdr = (struct ctdb_ltdb_header *)data.dptr;
1577 data.dptr += sizeof(*hdr);
1578 data.dsize -= sizeof(*hdr);
1579
1580 DEBUG(DEBUG_INFO, (__location__ " Failed to store "
1581 "record with hash 0x%08x in vacuum "
1582 "via RECEIVE_RECORDS\n",
1583 ctdb_hash(&key)));
1584
1585 old_size = talloc_get_size(records);
1586 records = talloc_realloc_size(outdata, records,
1587 old_size + rec->length);
1588 if (records == NULL) {
1589 DEBUG(DEBUG_ERR, (__location__ " Failed to "
1590 "expand\n"));
1591 return -1;
1592 }
1593 records->count++;
1594 memcpy(old_size+(uint8_t *)records, rec, rec->length);
1595 }
1596
1597 rec = (struct ctdb_rec_data_old *)(rec->length + (uint8_t *)rec);
1598 }
1599
1600 *outdata = ctdb_marshall_finish(records);
1601
1602 return 0;
1603 }
1604
1605
1606 /*
1607 report capabilities
1608 */
1609 int32_t ctdb_control_get_capabilities(struct ctdb_context *ctdb, TDB_DATA *outdata)
1610 {
1611 uint32_t *capabilities = NULL;
1612
1613 capabilities = talloc(outdata, uint32_t);
1614 CTDB_NO_MEMORY(ctdb, capabilities);
1615 *capabilities = ctdb->capabilities;
1616
1617 outdata->dsize = sizeof(uint32_t);
1618 outdata->dptr = (uint8_t *)capabilities;
1619
1620 return 0;
1621 }
1622
1623 /* The recovery daemon will ping us at regular intervals.
1624 If we havent been pinged for a while we assume the recovery
1625 daemon is inoperable and we restart.
1626 */
1627 static void ctdb_recd_ping_timeout(struct tevent_context *ev,
1628 struct tevent_timer *te,
1629 struct timeval t, void *p)
1630 {
1631 struct ctdb_context *ctdb = talloc_get_type(p, struct ctdb_context);
1632 uint32_t *count = talloc_get_type(ctdb->recd_ping_count, uint32_t);
1633
1634 DEBUG(DEBUG_ERR, ("Recovery daemon ping timeout. Count : %u\n", *count));
1635
1636 if (*count < ctdb->tunable.recd_ping_failcount) {
1637 (*count)++;
1638 tevent_add_timer(ctdb->ev, ctdb->recd_ping_count,
1639 timeval_current_ofs(ctdb->tunable.recd_ping_timeout, 0),
1640 ctdb_recd_ping_timeout, ctdb);
1641 return;
1642 }
1643
1644 DEBUG(DEBUG_ERR, ("Final timeout for recovery daemon ping. Restarting recovery daemon. (This can be caused if the cluster filesystem has hung)\n"));
1645
1646 ctdb_stop_recoverd(ctdb);
1647 ctdb_start_recoverd(ctdb);
1648 }
1649
1650 int32_t ctdb_control_recd_ping(struct ctdb_context *ctdb)
1651 {
1652 talloc_free(ctdb->recd_ping_count);
1653
1654 ctdb->recd_ping_count = talloc_zero(ctdb, uint32_t);
1655 CTDB_NO_MEMORY(ctdb, ctdb->recd_ping_count);
1656
1657 if (ctdb->tunable.recd_ping_timeout != 0) {
1658 tevent_add_timer(ctdb->ev, ctdb->recd_ping_count,
1659 timeval_current_ofs(ctdb->tunable.recd_ping_timeout, 0),
1660 ctdb_recd_ping_timeout, ctdb);
1661 }
1662
1663 return 0;
1664 }
1665
1666
1667
1668 int32_t ctdb_control_set_recmaster(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata)
1669 {
1670 uint32_t new_recmaster;
1671
1672 CHECK_CONTROL_DATA_SIZE(sizeof(uint32_t));
1673 new_recmaster = ((uint32_t *)(&indata.dptr[0]))[0];
1674
1675 if (ctdb->pnn != new_recmaster && ctdb->recovery_master == ctdb->pnn) {
1676 DEBUG(DEBUG_NOTICE,
1677 ("This node (%u) is no longer the recovery master\n", ctdb->pnn));
1678 }
1679
1680 if (ctdb->pnn == new_recmaster && ctdb->recovery_master != new_recmaster) {
1681 DEBUG(DEBUG_NOTICE,
1682 ("This node (%u) is now the recovery master\n", ctdb->pnn));
1683 }
1684
1685 ctdb->recovery_master = new_recmaster;
1686 return 0;
1687 }
1688
1689
1690 int32_t ctdb_control_stop_node(struct ctdb_context *ctdb)
1691 {
1692 DEBUG(DEBUG_NOTICE, ("Stopping node\n"));
1693 ctdb_disable_monitoring(ctdb);
1694 ctdb->nodes[ctdb->pnn]->flags |= NODE_FLAGS_STOPPED;
1695
1696 return 0;
1697 }
1698
1699 int32_t ctdb_control_continue_node(struct ctdb_context *ctdb)
1700 {
1701 DEBUG(DEBUG_NOTICE, ("Continue node\n"));
1702 ctdb->nodes[ctdb->pnn]->flags &= ~NODE_FLAGS_STOPPED;
1703
1704 return 0;
1705 }
1706
Samba GIT mirror