4 Copyright (C) Andrew Tridgell 2007
5 Copyright (C) Ronnie Sahlberg 2007
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.
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.
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/>.
23 #include "lib/tdb/include/tdb.h"
24 #include "lib/util/dlinklist.h"
25 #include "lib/events/events.h"
26 #include "system/network.h"
27 #include "system/filesys.h"
28 #include "system/locale.h"
30 #include "../include/ctdb_private.h"
31 #include "lib/util/dlinklist.h"
36 allocate a packet for use in client<->daemon communication
38 struct ctdb_req_header *_ctdbd_allocate_pkt(struct ctdb_context *ctdb,
40 enum ctdb_operation operation,
41 size_t length, size_t slength,
45 struct ctdb_req_header *hdr;
47 length = MAX(length, slength);
48 size = (length+(CTDB_DS_ALIGNMENT-1)) & ~(CTDB_DS_ALIGNMENT-1);
50 hdr = (struct ctdb_req_header *)talloc_size(mem_ctx, size);
52 DEBUG(DEBUG_ERR,("Unable to allocate packet for operation %u of length %u\n",
53 operation, (unsigned)length));
56 talloc_set_name_const(hdr, type);
57 memset(hdr, 0, slength);
59 hdr->operation = operation;
60 hdr->ctdb_magic = CTDB_MAGIC;
61 hdr->ctdb_version = CTDB_VERSION;
62 hdr->srcnode = ctdb->pnn;
64 hdr->generation = ctdb->vnn_map->generation;
71 local version of ctdb_call
73 int ctdb_call_local(struct ctdb_db_context *ctdb_db, struct ctdb_call *call,
74 struct ctdb_ltdb_header *header, TALLOC_CTX *mem_ctx,
75 TDB_DATA *data, uint32_t caller)
77 struct ctdb_call_info *c;
78 struct ctdb_registered_call *fn;
79 struct ctdb_context *ctdb = ctdb_db->ctdb;
81 c = talloc(ctdb, struct ctdb_call_info);
82 CTDB_NO_MEMORY(ctdb, c);
85 c->call_data = &call->call_data;
86 c->record_data.dptr = talloc_memdup(c, data->dptr, data->dsize);
87 c->record_data.dsize = data->dsize;
88 CTDB_NO_MEMORY(ctdb, c->record_data.dptr);
93 for (fn=ctdb_db->calls;fn;fn=fn->next) {
94 if (fn->id == call->call_id) break;
97 ctdb_set_error(ctdb, "Unknown call id %u\n", call->call_id);
102 if (fn->fn(c) != 0) {
103 ctdb_set_error(ctdb, "ctdb_call %u failed\n", call->call_id);
108 if (header->laccessor != caller) {
111 header->laccessor = caller;
114 /* we need to force the record to be written out if this was a remote access,
115 so that the lacount is updated */
116 if (c->new_data == NULL && header->laccessor != ctdb->pnn) {
117 c->new_data = &c->record_data;
121 /* XXX check that we always have the lock here? */
122 if (ctdb_ltdb_store(ctdb_db, call->key, header, *c->new_data) != 0) {
123 ctdb_set_error(ctdb, "ctdb_call tdb_store failed\n");
130 call->reply_data = *c->reply_data;
132 talloc_steal(call, call->reply_data.dptr);
133 talloc_set_name_const(call->reply_data.dptr, __location__);
135 call->reply_data.dptr = NULL;
136 call->reply_data.dsize = 0;
138 call->status = c->status;
147 queue a packet for sending from client to daemon
149 static int ctdb_client_queue_pkt(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
151 return ctdb_queue_send(ctdb->daemon.queue, (uint8_t *)hdr, hdr->length);
156 called when a CTDB_REPLY_CALL packet comes in in the client
158 This packet comes in response to a CTDB_REQ_CALL request packet. It
159 contains any reply data from the call
161 static void ctdb_client_reply_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
163 struct ctdb_reply_call *c = (struct ctdb_reply_call *)hdr;
164 struct ctdb_client_call_state *state;
166 state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_client_call_state);
168 DEBUG(DEBUG_ERR,(__location__ " reqid %u not found\n", hdr->reqid));
172 if (hdr->reqid != state->reqid) {
173 /* we found a record but it was the wrong one */
174 DEBUG(DEBUG_ERR, ("Dropped client call reply with reqid:%u\n",hdr->reqid));
178 state->call->reply_data.dptr = c->data;
179 state->call->reply_data.dsize = c->datalen;
180 state->call->status = c->status;
182 talloc_steal(state, c);
184 state->state = CTDB_CALL_DONE;
186 if (state->async.fn) {
187 state->async.fn(state);
191 static void ctdb_client_reply_control(struct ctdb_context *ctdb, struct ctdb_req_header *hdr);
194 this is called in the client, when data comes in from the daemon
196 static void ctdb_client_read_cb(uint8_t *data, size_t cnt, void *args)
198 struct ctdb_context *ctdb = talloc_get_type(args, struct ctdb_context);
199 struct ctdb_req_header *hdr = (struct ctdb_req_header *)data;
202 /* place the packet as a child of a tmp_ctx. We then use
203 talloc_free() below to free it. If any of the calls want
204 to keep it, then they will steal it somewhere else, and the
205 talloc_free() will be a no-op */
206 tmp_ctx = talloc_new(ctdb);
207 talloc_steal(tmp_ctx, hdr);
210 DEBUG(DEBUG_INFO,("Daemon has exited - shutting down client\n"));
214 if (cnt < sizeof(*hdr)) {
215 DEBUG(DEBUG_CRIT,("Bad packet length %u in client\n", (unsigned)cnt));
218 if (cnt != hdr->length) {
219 ctdb_set_error(ctdb, "Bad header length %u expected %u in client\n",
220 (unsigned)hdr->length, (unsigned)cnt);
224 if (hdr->ctdb_magic != CTDB_MAGIC) {
225 ctdb_set_error(ctdb, "Non CTDB packet rejected in client\n");
229 if (hdr->ctdb_version != CTDB_VERSION) {
230 ctdb_set_error(ctdb, "Bad CTDB version 0x%x rejected in client\n", hdr->ctdb_version);
234 switch (hdr->operation) {
235 case CTDB_REPLY_CALL:
236 ctdb_client_reply_call(ctdb, hdr);
239 case CTDB_REQ_MESSAGE:
240 ctdb_request_message(ctdb, hdr);
243 case CTDB_REPLY_CONTROL:
244 ctdb_client_reply_control(ctdb, hdr);
248 DEBUG(DEBUG_CRIT,("bogus operation code:%u\n",hdr->operation));
252 talloc_free(tmp_ctx);
256 connect to a unix domain socket
258 int ctdb_socket_connect(struct ctdb_context *ctdb)
260 struct sockaddr_un addr;
262 memset(&addr, 0, sizeof(addr));
263 addr.sun_family = AF_UNIX;
264 strncpy(addr.sun_path, ctdb->daemon.name, sizeof(addr.sun_path));
266 ctdb->daemon.sd = socket(AF_UNIX, SOCK_STREAM, 0);
267 if (ctdb->daemon.sd == -1) {
268 DEBUG(DEBUG_ERR,(__location__ " Failed to open client socket. Errno:%s(%d)\n", strerror(errno), errno));
272 set_nonblocking(ctdb->daemon.sd);
273 set_close_on_exec(ctdb->daemon.sd);
275 if (connect(ctdb->daemon.sd, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
276 close(ctdb->daemon.sd);
277 ctdb->daemon.sd = -1;
278 DEBUG(DEBUG_ERR,(__location__ " Failed to connect client socket to daemon. Errno:%s(%d)\n", strerror(errno), errno));
282 ctdb->daemon.queue = ctdb_queue_setup(ctdb, ctdb, ctdb->daemon.sd,
284 ctdb_client_read_cb, ctdb);
289 struct ctdb_record_handle {
290 struct ctdb_db_context *ctdb_db;
293 struct ctdb_ltdb_header header;
298 make a recv call to the local ctdb daemon - called from client context
300 This is called when the program wants to wait for a ctdb_call to complete and get the
301 results. This call will block unless the call has already completed.
303 int ctdb_call_recv(struct ctdb_client_call_state *state, struct ctdb_call *call)
309 while (state->state < CTDB_CALL_DONE) {
310 event_loop_once(state->ctdb_db->ctdb->ev);
312 if (state->state != CTDB_CALL_DONE) {
313 DEBUG(DEBUG_ERR,(__location__ " ctdb_call_recv failed\n"));
318 if (state->call->reply_data.dsize) {
319 call->reply_data.dptr = talloc_memdup(state->ctdb_db,
320 state->call->reply_data.dptr,
321 state->call->reply_data.dsize);
322 call->reply_data.dsize = state->call->reply_data.dsize;
324 call->reply_data.dptr = NULL;
325 call->reply_data.dsize = 0;
327 call->status = state->call->status;
337 destroy a ctdb_call in client
339 static int ctdb_client_call_destructor(struct ctdb_client_call_state *state)
341 ctdb_reqid_remove(state->ctdb_db->ctdb, state->reqid);
346 construct an event driven local ctdb_call
348 this is used so that locally processed ctdb_call requests are processed
349 in an event driven manner
351 static struct ctdb_client_call_state *ctdb_client_call_local_send(struct ctdb_db_context *ctdb_db,
352 struct ctdb_call *call,
353 struct ctdb_ltdb_header *header,
356 struct ctdb_client_call_state *state;
357 struct ctdb_context *ctdb = ctdb_db->ctdb;
360 state = talloc_zero(ctdb_db, struct ctdb_client_call_state);
361 CTDB_NO_MEMORY_NULL(ctdb, state);
362 state->call = talloc_zero(state, struct ctdb_call);
363 CTDB_NO_MEMORY_NULL(ctdb, state->call);
365 talloc_steal(state, data->dptr);
367 state->state = CTDB_CALL_DONE;
368 *(state->call) = *call;
369 state->ctdb_db = ctdb_db;
371 ret = ctdb_call_local(ctdb_db, state->call, header, state, data, ctdb->pnn);
377 make a ctdb call to the local daemon - async send. Called from client context.
379 This constructs a ctdb_call request and queues it for processing.
380 This call never blocks.
382 struct ctdb_client_call_state *ctdb_call_send(struct ctdb_db_context *ctdb_db,
383 struct ctdb_call *call)
385 struct ctdb_client_call_state *state;
386 struct ctdb_context *ctdb = ctdb_db->ctdb;
387 struct ctdb_ltdb_header header;
391 struct ctdb_req_call *c;
393 /* if the domain socket is not yet open, open it */
394 if (ctdb->daemon.sd==-1) {
395 ctdb_socket_connect(ctdb);
398 ret = ctdb_ltdb_lock(ctdb_db, call->key);
400 DEBUG(DEBUG_ERR,(__location__ " Failed to get chainlock\n"));
404 ret = ctdb_ltdb_fetch(ctdb_db, call->key, &header, ctdb_db, &data);
406 if (ret == 0 && header.dmaster == ctdb->pnn) {
407 state = ctdb_client_call_local_send(ctdb_db, call, &header, &data);
408 talloc_free(data.dptr);
409 ctdb_ltdb_unlock(ctdb_db, call->key);
413 ctdb_ltdb_unlock(ctdb_db, call->key);
414 talloc_free(data.dptr);
416 state = talloc_zero(ctdb_db, struct ctdb_client_call_state);
418 DEBUG(DEBUG_ERR, (__location__ " failed to allocate state\n"));
421 state->call = talloc_zero(state, struct ctdb_call);
422 if (state->call == NULL) {
423 DEBUG(DEBUG_ERR, (__location__ " failed to allocate state->call\n"));
427 len = offsetof(struct ctdb_req_call, data) + call->key.dsize + call->call_data.dsize;
428 c = ctdbd_allocate_pkt(ctdb, state, CTDB_REQ_CALL, len, struct ctdb_req_call);
430 DEBUG(DEBUG_ERR, (__location__ " failed to allocate packet\n"));
434 state->reqid = ctdb_reqid_new(ctdb, state);
435 state->ctdb_db = ctdb_db;
436 talloc_set_destructor(state, ctdb_client_call_destructor);
438 c->hdr.reqid = state->reqid;
439 c->flags = call->flags;
440 c->db_id = ctdb_db->db_id;
441 c->callid = call->call_id;
443 c->keylen = call->key.dsize;
444 c->calldatalen = call->call_data.dsize;
445 memcpy(&c->data[0], call->key.dptr, call->key.dsize);
446 memcpy(&c->data[call->key.dsize],
447 call->call_data.dptr, call->call_data.dsize);
448 *(state->call) = *call;
449 state->call->call_data.dptr = &c->data[call->key.dsize];
450 state->call->key.dptr = &c->data[0];
452 state->state = CTDB_CALL_WAIT;
455 ctdb_client_queue_pkt(ctdb, &c->hdr);
462 full ctdb_call. Equivalent to a ctdb_call_send() followed by a ctdb_call_recv()
464 int ctdb_call(struct ctdb_db_context *ctdb_db, struct ctdb_call *call)
466 struct ctdb_client_call_state *state;
468 state = ctdb_call_send(ctdb_db, call);
469 return ctdb_call_recv(state, call);
474 tell the daemon what messaging srvid we will use, and register the message
475 handler function in the client
477 int ctdb_set_message_handler(struct ctdb_context *ctdb, uint64_t srvid,
478 ctdb_message_fn_t handler,
485 res = ctdb_control(ctdb, CTDB_CURRENT_NODE, srvid, CTDB_CONTROL_REGISTER_SRVID, 0,
486 tdb_null, NULL, NULL, &status, NULL, NULL);
487 if (res != 0 || status != 0) {
488 DEBUG(DEBUG_ERR,("Failed to register srvid %llu\n", (unsigned long long)srvid));
492 /* also need to register the handler with our own ctdb structure */
493 return ctdb_register_message_handler(ctdb, ctdb, srvid, handler, private_data);
497 tell the daemon we no longer want a srvid
499 int ctdb_remove_message_handler(struct ctdb_context *ctdb, uint64_t srvid, void *private_data)
504 res = ctdb_control(ctdb, CTDB_CURRENT_NODE, srvid, CTDB_CONTROL_DEREGISTER_SRVID, 0,
505 tdb_null, NULL, NULL, &status, NULL, NULL);
506 if (res != 0 || status != 0) {
507 DEBUG(DEBUG_ERR,("Failed to deregister srvid %llu\n", (unsigned long long)srvid));
511 /* also need to register the handler with our own ctdb structure */
512 ctdb_deregister_message_handler(ctdb, srvid, private_data);
518 send a message - from client context
520 int ctdb_send_message(struct ctdb_context *ctdb, uint32_t pnn,
521 uint64_t srvid, TDB_DATA data)
523 struct ctdb_req_message *r;
526 len = offsetof(struct ctdb_req_message, data) + data.dsize;
527 r = ctdbd_allocate_pkt(ctdb, ctdb, CTDB_REQ_MESSAGE,
528 len, struct ctdb_req_message);
529 CTDB_NO_MEMORY(ctdb, r);
531 r->hdr.destnode = pnn;
533 r->datalen = data.dsize;
534 memcpy(&r->data[0], data.dptr, data.dsize);
536 res = ctdb_client_queue_pkt(ctdb, &r->hdr);
547 cancel a ctdb_fetch_lock operation, releasing the lock
549 static int fetch_lock_destructor(struct ctdb_record_handle *h)
551 ctdb_ltdb_unlock(h->ctdb_db, h->key);
556 force the migration of a record to this node
558 static int ctdb_client_force_migration(struct ctdb_db_context *ctdb_db, TDB_DATA key)
560 struct ctdb_call call;
562 call.call_id = CTDB_NULL_FUNC;
564 call.flags = CTDB_IMMEDIATE_MIGRATION;
565 return ctdb_call(ctdb_db, &call);
569 get a lock on a record, and return the records data. Blocks until it gets the lock
571 struct ctdb_record_handle *ctdb_fetch_lock(struct ctdb_db_context *ctdb_db, TALLOC_CTX *mem_ctx,
572 TDB_DATA key, TDB_DATA *data)
575 struct ctdb_record_handle *h;
578 procedure is as follows:
580 1) get the chain lock.
581 2) check if we are dmaster
582 3) if we are the dmaster then return handle
583 4) if not dmaster then ask ctdb daemon to make us dmaster, and wait for
585 5) when we get the reply, goto (1)
588 h = talloc_zero(mem_ctx, struct ctdb_record_handle);
593 h->ctdb_db = ctdb_db;
595 h->key.dptr = talloc_memdup(h, key.dptr, key.dsize);
596 if (h->key.dptr == NULL) {
602 DEBUG(DEBUG_DEBUG,("ctdb_fetch_lock: key=%*.*s\n", (int)key.dsize, (int)key.dsize,
603 (const char *)key.dptr));
606 /* step 1 - get the chain lock */
607 ret = ctdb_ltdb_lock(ctdb_db, key);
609 DEBUG(DEBUG_ERR, (__location__ " failed to lock ltdb record\n"));
614 DEBUG(DEBUG_DEBUG,("ctdb_fetch_lock: got chain lock\n"));
616 talloc_set_destructor(h, fetch_lock_destructor);
618 ret = ctdb_ltdb_fetch(ctdb_db, key, &h->header, h, data);
620 /* when torturing, ensure we test the remote path */
621 if ((ctdb_db->ctdb->flags & CTDB_FLAG_TORTURE) &&
623 h->header.dmaster = (uint32_t)-1;
627 DEBUG(DEBUG_DEBUG,("ctdb_fetch_lock: done local fetch\n"));
629 if (ret != 0 || h->header.dmaster != ctdb_db->ctdb->pnn) {
630 ctdb_ltdb_unlock(ctdb_db, key);
631 ret = ctdb_client_force_migration(ctdb_db, key);
633 DEBUG(DEBUG_DEBUG,("ctdb_fetch_lock: force_migration failed\n"));
640 DEBUG(DEBUG_DEBUG,("ctdb_fetch_lock: we are dmaster - done\n"));
645 store some data to the record that was locked with ctdb_fetch_lock()
647 int ctdb_record_store(struct ctdb_record_handle *h, TDB_DATA data)
649 if (h->ctdb_db->persistent) {
650 DEBUG(DEBUG_ERR, (__location__ " ctdb_record_store prohibited for persistent dbs\n"));
654 return ctdb_ltdb_store(h->ctdb_db, h->key, &h->header, data);
658 non-locking fetch of a record
660 int ctdb_fetch(struct ctdb_db_context *ctdb_db, TALLOC_CTX *mem_ctx,
661 TDB_DATA key, TDB_DATA *data)
663 struct ctdb_call call;
666 call.call_id = CTDB_FETCH_FUNC;
667 call.call_data.dptr = NULL;
668 call.call_data.dsize = 0;
670 ret = ctdb_call(ctdb_db, &call);
673 *data = call.reply_data;
674 talloc_steal(mem_ctx, data->dptr);
683 called when a control completes or timesout to invoke the callback
684 function the user provided
686 static void invoke_control_callback(struct event_context *ev, struct timed_event *te,
687 struct timeval t, void *private_data)
689 struct ctdb_client_control_state *state;
690 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
693 state = talloc_get_type(private_data, struct ctdb_client_control_state);
694 talloc_steal(tmp_ctx, state);
696 ret = ctdb_control_recv(state->ctdb, state, state,
701 talloc_free(tmp_ctx);
705 called when a CTDB_REPLY_CONTROL packet comes in in the client
707 This packet comes in response to a CTDB_REQ_CONTROL request packet. It
708 contains any reply data from the control
710 static void ctdb_client_reply_control(struct ctdb_context *ctdb,
711 struct ctdb_req_header *hdr)
713 struct ctdb_reply_control *c = (struct ctdb_reply_control *)hdr;
714 struct ctdb_client_control_state *state;
716 state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_client_control_state);
718 DEBUG(DEBUG_ERR,(__location__ " reqid %u not found\n", hdr->reqid));
722 if (hdr->reqid != state->reqid) {
723 /* we found a record but it was the wrong one */
724 DEBUG(DEBUG_ERR, ("Dropped orphaned reply control with reqid:%u\n",hdr->reqid));
728 state->outdata.dptr = c->data;
729 state->outdata.dsize = c->datalen;
730 state->status = c->status;
732 state->errormsg = talloc_strndup(state,
733 (char *)&c->data[c->datalen],
737 /* state->outdata now uses resources from c so we dont want c
738 to just dissappear from under us while state is still alive
740 talloc_steal(state, c);
742 state->state = CTDB_CONTROL_DONE;
744 /* if we had a callback registered for this control, pull the response
745 and call the callback.
747 if (state->async.fn) {
748 event_add_timed(ctdb->ev, state, timeval_zero(), invoke_control_callback, state);
754 destroy a ctdb_control in client
756 static int ctdb_control_destructor(struct ctdb_client_control_state *state)
758 ctdb_reqid_remove(state->ctdb, state->reqid);
763 /* time out handler for ctdb_control */
764 static void control_timeout_func(struct event_context *ev, struct timed_event *te,
765 struct timeval t, void *private_data)
767 struct ctdb_client_control_state *state = talloc_get_type(private_data, struct ctdb_client_control_state);
769 DEBUG(DEBUG_ERR,(__location__ " control timed out. reqid:%u opcode:%u "
770 "dstnode:%u\n", state->reqid, state->c->opcode,
771 state->c->hdr.destnode));
773 state->state = CTDB_CONTROL_TIMEOUT;
775 /* if we had a callback registered for this control, pull the response
776 and call the callback.
778 if (state->async.fn) {
779 event_add_timed(state->ctdb->ev, state, timeval_zero(), invoke_control_callback, state);
783 /* async version of send control request */
784 struct ctdb_client_control_state *ctdb_control_send(struct ctdb_context *ctdb,
785 uint32_t destnode, uint64_t srvid,
786 uint32_t opcode, uint32_t flags, TDB_DATA data,
788 struct timeval *timeout,
791 struct ctdb_client_control_state *state;
793 struct ctdb_req_control *c;
800 /* if the domain socket is not yet open, open it */
801 if (ctdb->daemon.sd==-1) {
802 ctdb_socket_connect(ctdb);
805 state = talloc_zero(mem_ctx, struct ctdb_client_control_state);
806 CTDB_NO_MEMORY_NULL(ctdb, state);
809 state->reqid = ctdb_reqid_new(ctdb, state);
810 state->state = CTDB_CONTROL_WAIT;
811 state->errormsg = NULL;
813 talloc_set_destructor(state, ctdb_control_destructor);
815 len = offsetof(struct ctdb_req_control, data) + data.dsize;
816 c = ctdbd_allocate_pkt(ctdb, state, CTDB_REQ_CONTROL,
817 len, struct ctdb_req_control);
819 CTDB_NO_MEMORY_NULL(ctdb, c);
820 c->hdr.reqid = state->reqid;
821 c->hdr.destnode = destnode;
826 c->datalen = data.dsize;
828 memcpy(&c->data[0], data.dptr, data.dsize);
832 if (timeout && !timeval_is_zero(timeout)) {
833 event_add_timed(ctdb->ev, state, *timeout, control_timeout_func, state);
836 ret = ctdb_client_queue_pkt(ctdb, &(c->hdr));
842 if (flags & CTDB_CTRL_FLAG_NOREPLY) {
851 /* async version of receive control reply */
852 int ctdb_control_recv(struct ctdb_context *ctdb,
853 struct ctdb_client_control_state *state,
855 TDB_DATA *outdata, int32_t *status, char **errormsg)
859 if (status != NULL) {
862 if (errormsg != NULL) {
870 /* prevent double free of state */
871 tmp_ctx = talloc_new(ctdb);
872 talloc_steal(tmp_ctx, state);
874 /* loop one event at a time until we either timeout or the control
877 while (state->state == CTDB_CONTROL_WAIT) {
878 event_loop_once(ctdb->ev);
881 if (state->state != CTDB_CONTROL_DONE) {
882 DEBUG(DEBUG_ERR,(__location__ " ctdb_control_recv failed\n"));
883 if (state->async.fn) {
884 state->async.fn(state);
886 talloc_free(tmp_ctx);
890 if (state->errormsg) {
891 DEBUG(DEBUG_ERR,("ctdb_control error: '%s'\n", state->errormsg));
893 (*errormsg) = talloc_move(mem_ctx, &state->errormsg);
895 if (state->async.fn) {
896 state->async.fn(state);
898 talloc_free(tmp_ctx);
903 *outdata = state->outdata;
904 outdata->dptr = talloc_memdup(mem_ctx, outdata->dptr, outdata->dsize);
908 *status = state->status;
911 if (state->async.fn) {
912 state->async.fn(state);
915 talloc_free(tmp_ctx);
922 send a ctdb control message
923 timeout specifies how long we should wait for a reply.
924 if timeout is NULL we wait indefinitely
926 int ctdb_control(struct ctdb_context *ctdb, uint32_t destnode, uint64_t srvid,
927 uint32_t opcode, uint32_t flags, TDB_DATA data,
928 TALLOC_CTX *mem_ctx, TDB_DATA *outdata, int32_t *status,
929 struct timeval *timeout,
932 struct ctdb_client_control_state *state;
934 state = ctdb_control_send(ctdb, destnode, srvid, opcode,
935 flags, data, mem_ctx,
937 return ctdb_control_recv(ctdb, state, mem_ctx, outdata, status,
945 a process exists call. Returns 0 if process exists, -1 otherwise
947 int ctdb_ctrl_process_exists(struct ctdb_context *ctdb, uint32_t destnode, pid_t pid)
953 data.dptr = (uint8_t*)&pid;
954 data.dsize = sizeof(pid);
956 ret = ctdb_control(ctdb, destnode, 0,
957 CTDB_CONTROL_PROCESS_EXISTS, 0, data,
958 NULL, NULL, &status, NULL, NULL);
960 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for process_exists failed\n"));
968 get remote statistics
970 int ctdb_ctrl_statistics(struct ctdb_context *ctdb, uint32_t destnode, struct ctdb_statistics *status)
976 ret = ctdb_control(ctdb, destnode, 0,
977 CTDB_CONTROL_STATISTICS, 0, tdb_null,
978 ctdb, &data, &res, NULL, NULL);
979 if (ret != 0 || res != 0) {
980 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for statistics failed\n"));
984 if (data.dsize != sizeof(struct ctdb_statistics)) {
985 DEBUG(DEBUG_ERR,(__location__ " Wrong statistics size %u - expected %u\n",
986 (unsigned)data.dsize, (unsigned)sizeof(struct ctdb_statistics)));
990 *status = *(struct ctdb_statistics *)data.dptr;
991 talloc_free(data.dptr);
997 shutdown a remote ctdb node
999 int ctdb_ctrl_shutdown(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
1001 struct ctdb_client_control_state *state;
1003 state = ctdb_control_send(ctdb, destnode, 0,
1004 CTDB_CONTROL_SHUTDOWN, 0, tdb_null,
1005 NULL, &timeout, NULL);
1006 if (state == NULL) {
1007 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for shutdown failed\n"));
1015 get vnn map from a remote node
1017 int ctdb_ctrl_getvnnmap(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, TALLOC_CTX *mem_ctx, struct ctdb_vnn_map **vnnmap)
1022 struct ctdb_vnn_map_wire *map;
1024 ret = ctdb_control(ctdb, destnode, 0,
1025 CTDB_CONTROL_GETVNNMAP, 0, tdb_null,
1026 mem_ctx, &outdata, &res, &timeout, NULL);
1027 if (ret != 0 || res != 0) {
1028 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getvnnmap failed\n"));
1032 map = (struct ctdb_vnn_map_wire *)outdata.dptr;
1033 if (outdata.dsize < offsetof(struct ctdb_vnn_map_wire, map) ||
1034 outdata.dsize != map->size*sizeof(uint32_t) + offsetof(struct ctdb_vnn_map_wire, map)) {
1035 DEBUG(DEBUG_ERR,("Bad vnn map size received in ctdb_ctrl_getvnnmap\n"));
1039 (*vnnmap) = talloc(mem_ctx, struct ctdb_vnn_map);
1040 CTDB_NO_MEMORY(ctdb, *vnnmap);
1041 (*vnnmap)->generation = map->generation;
1042 (*vnnmap)->size = map->size;
1043 (*vnnmap)->map = talloc_array(*vnnmap, uint32_t, map->size);
1045 CTDB_NO_MEMORY(ctdb, (*vnnmap)->map);
1046 memcpy((*vnnmap)->map, map->map, sizeof(uint32_t)*map->size);
1047 talloc_free(outdata.dptr);
1054 get the recovery mode of a remote node
1056 struct ctdb_client_control_state *
1057 ctdb_ctrl_getrecmode_send(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode)
1059 return ctdb_control_send(ctdb, destnode, 0,
1060 CTDB_CONTROL_GET_RECMODE, 0, tdb_null,
1061 mem_ctx, &timeout, NULL);
1064 int ctdb_ctrl_getrecmode_recv(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state, uint32_t *recmode)
1069 ret = ctdb_control_recv(ctdb, state, mem_ctx, NULL, &res, NULL);
1071 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_getrecmode_recv failed\n"));
1076 *recmode = (uint32_t)res;
1082 int ctdb_ctrl_getrecmode(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode, uint32_t *recmode)
1084 struct ctdb_client_control_state *state;
1086 state = ctdb_ctrl_getrecmode_send(ctdb, mem_ctx, timeout, destnode);
1087 return ctdb_ctrl_getrecmode_recv(ctdb, mem_ctx, state, recmode);
1094 set the recovery mode of a remote node
1096 int ctdb_ctrl_setrecmode(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t recmode)
1102 data.dsize = sizeof(uint32_t);
1103 data.dptr = (unsigned char *)&recmode;
1105 ret = ctdb_control(ctdb, destnode, 0,
1106 CTDB_CONTROL_SET_RECMODE, 0, data,
1107 NULL, NULL, &res, &timeout, NULL);
1108 if (ret != 0 || res != 0) {
1109 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setrecmode failed\n"));
1119 get the recovery master of a remote node
1121 struct ctdb_client_control_state *
1122 ctdb_ctrl_getrecmaster_send(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx,
1123 struct timeval timeout, uint32_t destnode)
1125 return ctdb_control_send(ctdb, destnode, 0,
1126 CTDB_CONTROL_GET_RECMASTER, 0, tdb_null,
1127 mem_ctx, &timeout, NULL);
1130 int ctdb_ctrl_getrecmaster_recv(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state, uint32_t *recmaster)
1135 ret = ctdb_control_recv(ctdb, state, mem_ctx, NULL, &res, NULL);
1137 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_getrecmaster_recv failed\n"));
1142 *recmaster = (uint32_t)res;
1148 int ctdb_ctrl_getrecmaster(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode, uint32_t *recmaster)
1150 struct ctdb_client_control_state *state;
1152 state = ctdb_ctrl_getrecmaster_send(ctdb, mem_ctx, timeout, destnode);
1153 return ctdb_ctrl_getrecmaster_recv(ctdb, mem_ctx, state, recmaster);
1158 set the recovery master of a remote node
1160 int ctdb_ctrl_setrecmaster(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t recmaster)
1167 data.dsize = sizeof(uint32_t);
1168 data.dptr = (unsigned char *)&recmaster;
1170 ret = ctdb_control(ctdb, destnode, 0,
1171 CTDB_CONTROL_SET_RECMASTER, 0, data,
1172 NULL, NULL, &res, &timeout, NULL);
1173 if (ret != 0 || res != 0) {
1174 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setrecmaster failed\n"));
1183 get a list of databases off a remote node
1185 int ctdb_ctrl_getdbmap(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode,
1186 TALLOC_CTX *mem_ctx, struct ctdb_dbid_map **dbmap)
1192 ret = ctdb_control(ctdb, destnode, 0,
1193 CTDB_CONTROL_GET_DBMAP, 0, tdb_null,
1194 mem_ctx, &outdata, &res, &timeout, NULL);
1195 if (ret != 0 || res != 0) {
1196 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getdbmap failed ret:%d res:%d\n", ret, res));
1200 *dbmap = (struct ctdb_dbid_map *)talloc_memdup(mem_ctx, outdata.dptr, outdata.dsize);
1201 talloc_free(outdata.dptr);
1207 get a list of nodes (vnn and flags ) from a remote node
1209 int ctdb_ctrl_getnodemap(struct ctdb_context *ctdb,
1210 struct timeval timeout, uint32_t destnode,
1211 TALLOC_CTX *mem_ctx, struct ctdb_node_map **nodemap)
1217 ret = ctdb_control(ctdb, destnode, 0,
1218 CTDB_CONTROL_GET_NODEMAP, 0, tdb_null,
1219 mem_ctx, &outdata, &res, &timeout, NULL);
1220 if (ret == 0 && res == -1 && outdata.dsize == 0) {
1221 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getnodes failed, falling back to ipv4-only control\n"));
1222 return ctdb_ctrl_getnodemapv4(ctdb, timeout, destnode, mem_ctx, nodemap);
1224 if (ret != 0 || res != 0 || outdata.dsize == 0) {
1225 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getnodes failed ret:%d res:%d\n", ret, res));
1229 *nodemap = (struct ctdb_node_map *)talloc_memdup(mem_ctx, outdata.dptr, outdata.dsize);
1230 talloc_free(outdata.dptr);
1236 old style ipv4-only get a list of nodes (vnn and flags ) from a remote node
1238 int ctdb_ctrl_getnodemapv4(struct ctdb_context *ctdb,
1239 struct timeval timeout, uint32_t destnode,
1240 TALLOC_CTX *mem_ctx, struct ctdb_node_map **nodemap)
1244 struct ctdb_node_mapv4 *nodemapv4;
1247 ret = ctdb_control(ctdb, destnode, 0,
1248 CTDB_CONTROL_GET_NODEMAPv4, 0, tdb_null,
1249 mem_ctx, &outdata, &res, &timeout, NULL);
1250 if (ret != 0 || res != 0 || outdata.dsize == 0) {
1251 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getnodesv4 failed ret:%d res:%d\n", ret, res));
1255 nodemapv4 = (struct ctdb_node_mapv4 *)outdata.dptr;
1257 len = offsetof(struct ctdb_node_map, nodes) + nodemapv4->num*sizeof(struct ctdb_node_and_flags);
1258 (*nodemap) = talloc_zero_size(mem_ctx, len);
1259 CTDB_NO_MEMORY(ctdb, (*nodemap));
1261 (*nodemap)->num = nodemapv4->num;
1262 for (i=0; i<nodemapv4->num; i++) {
1263 (*nodemap)->nodes[i].pnn = nodemapv4->nodes[i].pnn;
1264 (*nodemap)->nodes[i].flags = nodemapv4->nodes[i].flags;
1265 (*nodemap)->nodes[i].addr.ip = nodemapv4->nodes[i].sin;
1266 (*nodemap)->nodes[i].addr.sa.sa_family = AF_INET;
1269 talloc_free(outdata.dptr);
1275 drop the transport, reload the nodes file and restart the transport
1277 int ctdb_ctrl_reload_nodes_file(struct ctdb_context *ctdb,
1278 struct timeval timeout, uint32_t destnode)
1283 ret = ctdb_control(ctdb, destnode, 0,
1284 CTDB_CONTROL_RELOAD_NODES_FILE, 0, tdb_null,
1285 NULL, NULL, &res, &timeout, NULL);
1286 if (ret != 0 || res != 0) {
1287 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for reloadnodesfile failed\n"));
1296 set vnn map on a node
1298 int ctdb_ctrl_setvnnmap(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode,
1299 TALLOC_CTX *mem_ctx, struct ctdb_vnn_map *vnnmap)
1304 struct ctdb_vnn_map_wire *map;
1307 len = offsetof(struct ctdb_vnn_map_wire, map) + sizeof(uint32_t)*vnnmap->size;
1308 map = talloc_size(mem_ctx, len);
1309 CTDB_NO_MEMORY(ctdb, map);
1311 map->generation = vnnmap->generation;
1312 map->size = vnnmap->size;
1313 memcpy(map->map, vnnmap->map, sizeof(uint32_t)*map->size);
1316 data.dptr = (uint8_t *)map;
1318 ret = ctdb_control(ctdb, destnode, 0,
1319 CTDB_CONTROL_SETVNNMAP, 0, data,
1320 NULL, NULL, &res, &timeout, NULL);
1321 if (ret != 0 || res != 0) {
1322 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setvnnmap failed\n"));
1333 async send for pull database
1335 struct ctdb_client_control_state *ctdb_ctrl_pulldb_send(
1336 struct ctdb_context *ctdb, uint32_t destnode, uint32_t dbid,
1337 uint32_t lmaster, TALLOC_CTX *mem_ctx, struct timeval timeout)
1340 struct ctdb_control_pulldb *pull;
1341 struct ctdb_client_control_state *state;
1343 pull = talloc(mem_ctx, struct ctdb_control_pulldb);
1344 CTDB_NO_MEMORY_NULL(ctdb, pull);
1347 pull->lmaster = lmaster;
1349 indata.dsize = sizeof(struct ctdb_control_pulldb);
1350 indata.dptr = (unsigned char *)pull;
1352 state = ctdb_control_send(ctdb, destnode, 0,
1353 CTDB_CONTROL_PULL_DB, 0, indata,
1354 mem_ctx, &timeout, NULL);
1361 async recv for pull database
1363 int ctdb_ctrl_pulldb_recv(
1364 struct ctdb_context *ctdb,
1365 TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state,
1371 ret = ctdb_control_recv(ctdb, state, mem_ctx, outdata, &res, NULL);
1372 if ( (ret != 0) || (res != 0) ){
1373 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_pulldb_recv failed\n"));
1381 pull all keys and records for a specific database on a node
1383 int ctdb_ctrl_pulldb(struct ctdb_context *ctdb, uint32_t destnode,
1384 uint32_t dbid, uint32_t lmaster,
1385 TALLOC_CTX *mem_ctx, struct timeval timeout,
1388 struct ctdb_client_control_state *state;
1390 state = ctdb_ctrl_pulldb_send(ctdb, destnode, dbid, lmaster, mem_ctx,
1393 return ctdb_ctrl_pulldb_recv(ctdb, mem_ctx, state, outdata);
1398 change dmaster for all keys in the database to the new value
1400 int ctdb_ctrl_setdmaster(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode,
1401 TALLOC_CTX *mem_ctx, uint32_t dbid, uint32_t dmaster)
1407 indata.dsize = 2*sizeof(uint32_t);
1408 indata.dptr = (unsigned char *)talloc_array(mem_ctx, uint32_t, 2);
1410 ((uint32_t *)(&indata.dptr[0]))[0] = dbid;
1411 ((uint32_t *)(&indata.dptr[0]))[1] = dmaster;
1413 ret = ctdb_control(ctdb, destnode, 0,
1414 CTDB_CONTROL_SET_DMASTER, 0, indata,
1415 NULL, NULL, &res, &timeout, NULL);
1416 if (ret != 0 || res != 0) {
1417 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setdmaster failed\n"));
1425 ping a node, return number of clients connected
1427 int ctdb_ctrl_ping(struct ctdb_context *ctdb, uint32_t destnode)
1432 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_PING, 0,
1433 tdb_null, NULL, NULL, &res, NULL, NULL);
1441 find the real path to a ltdb
1443 int ctdb_ctrl_getdbpath(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t dbid, TALLOC_CTX *mem_ctx,
1450 data.dptr = (uint8_t *)&dbid;
1451 data.dsize = sizeof(dbid);
1453 ret = ctdb_control(ctdb, destnode, 0,
1454 CTDB_CONTROL_GETDBPATH, 0, data,
1455 mem_ctx, &data, &res, &timeout, NULL);
1456 if (ret != 0 || res != 0) {
1460 (*path) = talloc_strndup(mem_ctx, (const char *)data.dptr, data.dsize);
1461 if ((*path) == NULL) {
1465 talloc_free(data.dptr);
1471 find the name of a db
1473 int ctdb_ctrl_getdbname(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t dbid, TALLOC_CTX *mem_ctx,
1480 data.dptr = (uint8_t *)&dbid;
1481 data.dsize = sizeof(dbid);
1483 ret = ctdb_control(ctdb, destnode, 0,
1484 CTDB_CONTROL_GET_DBNAME, 0, data,
1485 mem_ctx, &data, &res, &timeout, NULL);
1486 if (ret != 0 || res != 0) {
1490 (*name) = talloc_strndup(mem_ctx, (const char *)data.dptr, data.dsize);
1491 if ((*name) == NULL) {
1495 talloc_free(data.dptr);
1501 get the health status of a db
1503 int ctdb_ctrl_getdbhealth(struct ctdb_context *ctdb,
1504 struct timeval timeout,
1506 uint32_t dbid, TALLOC_CTX *mem_ctx,
1507 const char **reason)
1513 data.dptr = (uint8_t *)&dbid;
1514 data.dsize = sizeof(dbid);
1516 ret = ctdb_control(ctdb, destnode, 0,
1517 CTDB_CONTROL_DB_GET_HEALTH, 0, data,
1518 mem_ctx, &data, &res, &timeout, NULL);
1519 if (ret != 0 || res != 0) {
1523 if (data.dsize == 0) {
1528 (*reason) = talloc_strndup(mem_ctx, (const char *)data.dptr, data.dsize);
1529 if ((*reason) == NULL) {
1533 talloc_free(data.dptr);
1541 int ctdb_ctrl_createdb(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode,
1542 TALLOC_CTX *mem_ctx, const char *name, bool persistent)
1548 data.dptr = discard_const(name);
1549 data.dsize = strlen(name)+1;
1551 ret = ctdb_control(ctdb, destnode, 0,
1552 persistent?CTDB_CONTROL_DB_ATTACH_PERSISTENT:CTDB_CONTROL_DB_ATTACH,
1554 mem_ctx, &data, &res, &timeout, NULL);
1556 if (ret != 0 || res != 0) {
1564 get debug level on a node
1566 int ctdb_ctrl_get_debuglevel(struct ctdb_context *ctdb, uint32_t destnode, int32_t *level)
1572 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_GET_DEBUG, 0, tdb_null,
1573 ctdb, &data, &res, NULL, NULL);
1574 if (ret != 0 || res != 0) {
1577 if (data.dsize != sizeof(int32_t)) {
1578 DEBUG(DEBUG_ERR,("Bad control reply size in ctdb_get_debuglevel (got %u)\n",
1579 (unsigned)data.dsize));
1582 *level = *(int32_t *)data.dptr;
1583 talloc_free(data.dptr);
1588 set debug level on a node
1590 int ctdb_ctrl_set_debuglevel(struct ctdb_context *ctdb, uint32_t destnode, int32_t level)
1596 data.dptr = (uint8_t *)&level;
1597 data.dsize = sizeof(level);
1599 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_SET_DEBUG, 0, data,
1600 NULL, NULL, &res, NULL, NULL);
1601 if (ret != 0 || res != 0) {
1609 get a list of connected nodes
1611 uint32_t *ctdb_get_connected_nodes(struct ctdb_context *ctdb,
1612 struct timeval timeout,
1613 TALLOC_CTX *mem_ctx,
1614 uint32_t *num_nodes)
1616 struct ctdb_node_map *map=NULL;
1622 ret = ctdb_ctrl_getnodemap(ctdb, timeout, CTDB_CURRENT_NODE, mem_ctx, &map);
1627 nodes = talloc_array(mem_ctx, uint32_t, map->num);
1628 if (nodes == NULL) {
1632 for (i=0;i<map->num;i++) {
1633 if (!(map->nodes[i].flags & NODE_FLAGS_DISCONNECTED)) {
1634 nodes[*num_nodes] = map->nodes[i].pnn;
1646 int ctdb_statistics_reset(struct ctdb_context *ctdb, uint32_t destnode)
1651 ret = ctdb_control(ctdb, destnode, 0,
1652 CTDB_CONTROL_STATISTICS_RESET, 0, tdb_null,
1653 NULL, NULL, &res, NULL, NULL);
1654 if (ret != 0 || res != 0) {
1655 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for reset statistics failed\n"));
1662 this is the dummy null procedure that all databases support
1664 static int ctdb_null_func(struct ctdb_call_info *call)
1670 this is a plain fetch procedure that all databases support
1672 static int ctdb_fetch_func(struct ctdb_call_info *call)
1674 call->reply_data = &call->record_data;
1679 attach to a specific database - client call
1681 struct ctdb_db_context *ctdb_attach(struct ctdb_context *ctdb, const char *name, bool persistent, uint32_t tdb_flags)
1683 struct ctdb_db_context *ctdb_db;
1688 ctdb_db = ctdb_db_handle(ctdb, name);
1693 ctdb_db = talloc_zero(ctdb, struct ctdb_db_context);
1694 CTDB_NO_MEMORY_NULL(ctdb, ctdb_db);
1696 ctdb_db->ctdb = ctdb;
1697 ctdb_db->db_name = talloc_strdup(ctdb_db, name);
1698 CTDB_NO_MEMORY_NULL(ctdb, ctdb_db->db_name);
1700 data.dptr = discard_const(name);
1701 data.dsize = strlen(name)+1;
1703 /* tell ctdb daemon to attach */
1704 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, tdb_flags,
1705 persistent?CTDB_CONTROL_DB_ATTACH_PERSISTENT:CTDB_CONTROL_DB_ATTACH,
1706 0, data, ctdb_db, &data, &res, NULL, NULL);
1707 if (ret != 0 || res != 0 || data.dsize != sizeof(uint32_t)) {
1708 DEBUG(DEBUG_ERR,("Failed to attach to database '%s'\n", name));
1709 talloc_free(ctdb_db);
1713 ctdb_db->db_id = *(uint32_t *)data.dptr;
1714 talloc_free(data.dptr);
1716 ret = ctdb_ctrl_getdbpath(ctdb, timeval_current_ofs(2, 0), CTDB_CURRENT_NODE, ctdb_db->db_id, ctdb_db, &ctdb_db->db_path);
1718 DEBUG(DEBUG_ERR,("Failed to get dbpath for database '%s'\n", name));
1719 talloc_free(ctdb_db);
1723 tdb_flags = persistent?TDB_DEFAULT:TDB_NOSYNC;
1724 if (!ctdb->do_setsched) {
1725 tdb_flags |= TDB_NOMMAP;
1727 tdb_flags |= TDB_DISALLOW_NESTING;
1729 ctdb_db->ltdb = tdb_wrap_open(ctdb, ctdb_db->db_path, 0, tdb_flags, O_RDWR, 0);
1730 if (ctdb_db->ltdb == NULL) {
1731 ctdb_set_error(ctdb, "Failed to open tdb '%s'\n", ctdb_db->db_path);
1732 talloc_free(ctdb_db);
1736 ctdb_db->persistent = persistent;
1738 DLIST_ADD(ctdb->db_list, ctdb_db);
1740 /* add well known functions */
1741 ctdb_set_call(ctdb_db, ctdb_null_func, CTDB_NULL_FUNC);
1742 ctdb_set_call(ctdb_db, ctdb_fetch_func, CTDB_FETCH_FUNC);
1749 setup a call for a database
1751 int ctdb_set_call(struct ctdb_db_context *ctdb_db, ctdb_fn_t fn, uint32_t id)
1753 struct ctdb_registered_call *call;
1758 struct ctdb_control_set_call c;
1761 /* this is no longer valid with the separate daemon architecture */
1762 c.db_id = ctdb_db->db_id;
1766 data.dptr = (uint8_t *)&c;
1767 data.dsize = sizeof(c);
1769 ret = ctdb_control(ctdb_db->ctdb, CTDB_CURRENT_NODE, 0, CTDB_CONTROL_SET_CALL, 0,
1770 data, NULL, NULL, &status, NULL, NULL);
1771 if (ret != 0 || status != 0) {
1772 DEBUG(DEBUG_ERR,("ctdb_set_call failed for call %u\n", id));
1777 /* also register locally */
1778 call = talloc(ctdb_db, struct ctdb_registered_call);
1782 DLIST_ADD(ctdb_db->calls, call);
1787 struct traverse_state {
1790 ctdb_traverse_func fn;
1795 called on each key during a ctdb_traverse
1797 static void traverse_handler(struct ctdb_context *ctdb, uint64_t srvid, TDB_DATA data, void *p)
1799 struct traverse_state *state = (struct traverse_state *)p;
1800 struct ctdb_rec_data *d = (struct ctdb_rec_data *)data.dptr;
1803 if (data.dsize < sizeof(uint32_t) ||
1804 d->length != data.dsize) {
1805 DEBUG(DEBUG_ERR,("Bad data size %u in traverse_handler\n", (unsigned)data.dsize));
1810 key.dsize = d->keylen;
1811 key.dptr = &d->data[0];
1812 data.dsize = d->datalen;
1813 data.dptr = &d->data[d->keylen];
1815 if (key.dsize == 0 && data.dsize == 0) {
1816 /* end of traverse */
1821 if (data.dsize == sizeof(struct ctdb_ltdb_header)) {
1822 /* empty records are deleted records in ctdb */
1826 if (state->fn(ctdb, key, data, state->private_data) != 0) {
1835 start a cluster wide traverse, calling the supplied fn on each record
1836 return the number of records traversed, or -1 on error
1838 int ctdb_traverse(struct ctdb_db_context *ctdb_db, ctdb_traverse_func fn, void *private_data)
1841 struct ctdb_traverse_start t;
1844 uint64_t srvid = (getpid() | 0xFLL<<60);
1845 struct traverse_state state;
1849 state.private_data = private_data;
1852 ret = ctdb_set_message_handler(ctdb_db->ctdb, srvid, traverse_handler, &state);
1854 DEBUG(DEBUG_ERR,("Failed to setup traverse handler\n"));
1858 t.db_id = ctdb_db->db_id;
1862 data.dptr = (uint8_t *)&t;
1863 data.dsize = sizeof(t);
1865 ret = ctdb_control(ctdb_db->ctdb, CTDB_CURRENT_NODE, 0, CTDB_CONTROL_TRAVERSE_START, 0,
1866 data, NULL, NULL, &status, NULL, NULL);
1867 if (ret != 0 || status != 0) {
1868 DEBUG(DEBUG_ERR,("ctdb_traverse_all failed\n"));
1869 ctdb_remove_message_handler(ctdb_db->ctdb, srvid, &state);
1873 while (!state.done) {
1874 event_loop_once(ctdb_db->ctdb->ev);
1877 ret = ctdb_remove_message_handler(ctdb_db->ctdb, srvid, &state);
1879 DEBUG(DEBUG_ERR,("Failed to remove ctdb_traverse handler\n"));
1886 #define ISASCII(x) ((x>31)&&(x<128))
1888 called on each key during a catdb
1890 int ctdb_dumpdb_record(struct ctdb_context *ctdb, TDB_DATA key, TDB_DATA data, void *p)
1893 FILE *f = (FILE *)p;
1894 struct ctdb_ltdb_header *h = (struct ctdb_ltdb_header *)data.dptr;
1896 fprintf(f, "key(%u) = \"", (unsigned)key.dsize);
1897 for (i=0;i<key.dsize;i++) {
1898 if (ISASCII(key.dptr[i])) {
1899 fprintf(f, "%c", key.dptr[i]);
1901 fprintf(f, "\\%02X", key.dptr[i]);
1906 fprintf(f, "dmaster: %u\n", h->dmaster);
1907 fprintf(f, "rsn: %llu\n", (unsigned long long)h->rsn);
1909 fprintf(f, "data(%u) = \"", (unsigned)data.dsize);
1910 for (i=sizeof(*h);i<data.dsize;i++) {
1911 if (ISASCII(data.dptr[i])) {
1912 fprintf(f, "%c", data.dptr[i]);
1914 fprintf(f, "\\%02X", data.dptr[i]);
1925 convenience function to list all keys to stdout
1927 int ctdb_dump_db(struct ctdb_db_context *ctdb_db, FILE *f)
1929 return ctdb_traverse(ctdb_db, ctdb_dumpdb_record, f);
1933 get the pid of a ctdb daemon
1935 int ctdb_ctrl_getpid(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t *pid)
1940 ret = ctdb_control(ctdb, destnode, 0,
1941 CTDB_CONTROL_GET_PID, 0, tdb_null,
1942 NULL, NULL, &res, &timeout, NULL);
1944 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getpid failed\n"));
1955 async freeze send control
1957 struct ctdb_client_control_state *
1958 ctdb_ctrl_freeze_send(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode, uint32_t priority)
1960 return ctdb_control_send(ctdb, destnode, priority,
1961 CTDB_CONTROL_FREEZE, 0, tdb_null,
1962 mem_ctx, &timeout, NULL);
1966 async freeze recv control
1968 int ctdb_ctrl_freeze_recv(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state)
1973 ret = ctdb_control_recv(ctdb, state, mem_ctx, NULL, &res, NULL);
1974 if ( (ret != 0) || (res != 0) ){
1975 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_freeze_recv failed\n"));
1983 freeze databases of a certain priority
1985 int ctdb_ctrl_freeze_priority(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t priority)
1987 TALLOC_CTX *tmp_ctx = talloc_new(ctdb);
1988 struct ctdb_client_control_state *state;
1991 state = ctdb_ctrl_freeze_send(ctdb, tmp_ctx, timeout, destnode, priority);
1992 ret = ctdb_ctrl_freeze_recv(ctdb, tmp_ctx, state);
1993 talloc_free(tmp_ctx);
1998 /* Freeze all databases */
1999 int ctdb_ctrl_freeze(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2003 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
2004 if (ctdb_ctrl_freeze_priority(ctdb, timeout, destnode, i) != 0) {
2012 thaw databases of a certain priority
2014 int ctdb_ctrl_thaw_priority(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t priority)
2019 ret = ctdb_control(ctdb, destnode, priority,
2020 CTDB_CONTROL_THAW, 0, tdb_null,
2021 NULL, NULL, &res, &timeout, NULL);
2022 if (ret != 0 || res != 0) {
2023 DEBUG(DEBUG_ERR,(__location__ " ctdb_control thaw failed\n"));
2030 /* thaw all databases */
2031 int ctdb_ctrl_thaw(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2033 return ctdb_ctrl_thaw_priority(ctdb, timeout, destnode, 0);
2037 get pnn of a node, or -1
2039 int ctdb_ctrl_getpnn(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2044 ret = ctdb_control(ctdb, destnode, 0,
2045 CTDB_CONTROL_GET_PNN, 0, tdb_null,
2046 NULL, NULL, &res, &timeout, NULL);
2048 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getpnn failed\n"));
2056 get the monitoring mode of a remote node
2058 int ctdb_ctrl_getmonmode(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t *monmode)
2063 ret = ctdb_control(ctdb, destnode, 0,
2064 CTDB_CONTROL_GET_MONMODE, 0, tdb_null,
2065 NULL, NULL, &res, &timeout, NULL);
2067 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getmonmode failed\n"));
2078 set the monitoring mode of a remote node to active
2080 int ctdb_ctrl_enable_monmode(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2085 ret = ctdb_control(ctdb, destnode, 0,
2086 CTDB_CONTROL_ENABLE_MONITOR, 0, tdb_null,
2087 NULL, NULL,NULL, &timeout, NULL);
2089 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for enable_monitor failed\n"));
2099 set the monitoring mode of a remote node to disable
2101 int ctdb_ctrl_disable_monmode(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2106 ret = ctdb_control(ctdb, destnode, 0,
2107 CTDB_CONTROL_DISABLE_MONITOR, 0, tdb_null,
2108 NULL, NULL, NULL, &timeout, NULL);
2110 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for disable_monitor failed\n"));
2122 sent to a node to make it take over an ip address
2124 int ctdb_ctrl_takeover_ip(struct ctdb_context *ctdb, struct timeval timeout,
2125 uint32_t destnode, struct ctdb_public_ip *ip)
2128 struct ctdb_public_ipv4 ipv4;
2132 if (ip->addr.sa.sa_family == AF_INET) {
2134 ipv4.sin = ip->addr.ip;
2136 data.dsize = sizeof(ipv4);
2137 data.dptr = (uint8_t *)&ipv4;
2139 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_TAKEOVER_IPv4, 0, data, NULL,
2140 NULL, &res, &timeout, NULL);
2142 data.dsize = sizeof(*ip);
2143 data.dptr = (uint8_t *)ip;
2145 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_TAKEOVER_IP, 0, data, NULL,
2146 NULL, &res, &timeout, NULL);
2149 if (ret != 0 || res != 0) {
2150 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for takeover_ip failed\n"));
2159 sent to a node to make it release an ip address
2161 int ctdb_ctrl_release_ip(struct ctdb_context *ctdb, struct timeval timeout,
2162 uint32_t destnode, struct ctdb_public_ip *ip)
2165 struct ctdb_public_ipv4 ipv4;
2169 if (ip->addr.sa.sa_family == AF_INET) {
2171 ipv4.sin = ip->addr.ip;
2173 data.dsize = sizeof(ipv4);
2174 data.dptr = (uint8_t *)&ipv4;
2176 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_RELEASE_IPv4, 0, data, NULL,
2177 NULL, &res, &timeout, NULL);
2179 data.dsize = sizeof(*ip);
2180 data.dptr = (uint8_t *)ip;
2182 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_RELEASE_IP, 0, data, NULL,
2183 NULL, &res, &timeout, NULL);
2186 if (ret != 0 || res != 0) {
2187 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for release_ip failed\n"));
2198 int ctdb_ctrl_get_tunable(struct ctdb_context *ctdb,
2199 struct timeval timeout,
2201 const char *name, uint32_t *value)
2203 struct ctdb_control_get_tunable *t;
2204 TDB_DATA data, outdata;
2208 data.dsize = offsetof(struct ctdb_control_get_tunable, name) + strlen(name) + 1;
2209 data.dptr = talloc_size(ctdb, data.dsize);
2210 CTDB_NO_MEMORY(ctdb, data.dptr);
2212 t = (struct ctdb_control_get_tunable *)data.dptr;
2213 t->length = strlen(name)+1;
2214 memcpy(t->name, name, t->length);
2216 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_GET_TUNABLE, 0, data, ctdb,
2217 &outdata, &res, &timeout, NULL);
2218 talloc_free(data.dptr);
2219 if (ret != 0 || res != 0) {
2220 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for get_tunable failed\n"));
2224 if (outdata.dsize != sizeof(uint32_t)) {
2225 DEBUG(DEBUG_ERR,("Invalid return data in get_tunable\n"));
2226 talloc_free(outdata.dptr);
2230 *value = *(uint32_t *)outdata.dptr;
2231 talloc_free(outdata.dptr);
2239 int ctdb_ctrl_set_tunable(struct ctdb_context *ctdb,
2240 struct timeval timeout,
2242 const char *name, uint32_t value)
2244 struct ctdb_control_set_tunable *t;
2249 data.dsize = offsetof(struct ctdb_control_set_tunable, name) + strlen(name) + 1;
2250 data.dptr = talloc_size(ctdb, data.dsize);
2251 CTDB_NO_MEMORY(ctdb, data.dptr);
2253 t = (struct ctdb_control_set_tunable *)data.dptr;
2254 t->length = strlen(name)+1;
2255 memcpy(t->name, name, t->length);
2258 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_SET_TUNABLE, 0, data, NULL,
2259 NULL, &res, &timeout, NULL);
2260 talloc_free(data.dptr);
2261 if (ret != 0 || res != 0) {
2262 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for set_tunable failed\n"));
2272 int ctdb_ctrl_list_tunables(struct ctdb_context *ctdb,
2273 struct timeval timeout,
2275 TALLOC_CTX *mem_ctx,
2276 const char ***list, uint32_t *count)
2281 struct ctdb_control_list_tunable *t;
2284 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_LIST_TUNABLES, 0, tdb_null,
2285 mem_ctx, &outdata, &res, &timeout, NULL);
2286 if (ret != 0 || res != 0) {
2287 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for list_tunables failed\n"));
2291 t = (struct ctdb_control_list_tunable *)outdata.dptr;
2292 if (outdata.dsize < offsetof(struct ctdb_control_list_tunable, data) ||
2293 t->length > outdata.dsize-offsetof(struct ctdb_control_list_tunable, data)) {
2294 DEBUG(DEBUG_ERR,("Invalid data in list_tunables reply\n"));
2295 talloc_free(outdata.dptr);
2299 p = talloc_strndup(mem_ctx, (char *)t->data, t->length);
2300 CTDB_NO_MEMORY(ctdb, p);
2302 talloc_free(outdata.dptr);
2307 for (s=strtok_r(p, ":", &ptr); s; s=strtok_r(NULL, ":", &ptr)) {
2308 (*list) = talloc_realloc(mem_ctx, *list, const char *, 1+(*count));
2309 CTDB_NO_MEMORY(ctdb, *list);
2310 (*list)[*count] = talloc_strdup(*list, s);
2311 CTDB_NO_MEMORY(ctdb, (*list)[*count]);
2321 int ctdb_ctrl_get_public_ips(struct ctdb_context *ctdb,
2322 struct timeval timeout, uint32_t destnode,
2323 TALLOC_CTX *mem_ctx, struct ctdb_all_public_ips **ips)
2329 ret = ctdb_control(ctdb, destnode, 0,
2330 CTDB_CONTROL_GET_PUBLIC_IPS, 0, tdb_null,
2331 mem_ctx, &outdata, &res, &timeout, NULL);
2332 if (ret == 0 && res == -1) {
2333 DEBUG(DEBUG_ERR,(__location__ " ctdb_control to get public ips failed, falling back to ipv4-only version\n"));
2334 return ctdb_ctrl_get_public_ipsv4(ctdb, timeout, destnode, mem_ctx, ips);
2336 if (ret != 0 || res != 0) {
2337 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getpublicips failed ret:%d res:%d\n", ret, res));
2341 *ips = (struct ctdb_all_public_ips *)talloc_memdup(mem_ctx, outdata.dptr, outdata.dsize);
2342 talloc_free(outdata.dptr);
2347 int ctdb_ctrl_get_public_ipsv4(struct ctdb_context *ctdb,
2348 struct timeval timeout, uint32_t destnode,
2349 TALLOC_CTX *mem_ctx, struct ctdb_all_public_ips **ips)
2354 struct ctdb_all_public_ipsv4 *ipsv4;
2356 ret = ctdb_control(ctdb, destnode, 0,
2357 CTDB_CONTROL_GET_PUBLIC_IPSv4, 0, tdb_null,
2358 mem_ctx, &outdata, &res, &timeout, NULL);
2359 if (ret != 0 || res != 0) {
2360 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getpublicips failed\n"));
2364 ipsv4 = (struct ctdb_all_public_ipsv4 *)outdata.dptr;
2365 len = offsetof(struct ctdb_all_public_ips, ips) +
2366 ipsv4->num*sizeof(struct ctdb_public_ip);
2367 *ips = talloc_zero_size(mem_ctx, len);
2368 CTDB_NO_MEMORY(ctdb, *ips);
2369 (*ips)->num = ipsv4->num;
2370 for (i=0; i<ipsv4->num; i++) {
2371 (*ips)->ips[i].pnn = ipsv4->ips[i].pnn;
2372 (*ips)->ips[i].addr.ip = ipsv4->ips[i].sin;
2375 talloc_free(outdata.dptr);
2381 set/clear the permanent disabled bit on a remote node
2383 int ctdb_ctrl_modflags(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode,
2384 uint32_t set, uint32_t clear)
2388 struct ctdb_node_map *nodemap=NULL;
2389 struct ctdb_node_flag_change c;
2390 TALLOC_CTX *tmp_ctx = talloc_new(ctdb);
2395 /* find the recovery master */
2396 ret = ctdb_ctrl_getrecmaster(ctdb, tmp_ctx, timeout, CTDB_CURRENT_NODE, &recmaster);
2398 DEBUG(DEBUG_ERR, (__location__ " Unable to get recmaster from local node\n"));
2399 talloc_free(tmp_ctx);
2404 /* read the node flags from the recmaster */
2405 ret = ctdb_ctrl_getnodemap(ctdb, timeout, recmaster, tmp_ctx, &nodemap);
2407 DEBUG(DEBUG_ERR, (__location__ " Unable to get nodemap from node %u\n", destnode));
2408 talloc_free(tmp_ctx);
2411 if (destnode >= nodemap->num) {
2412 DEBUG(DEBUG_ERR,(__location__ " Nodemap from recmaster does not contain node %d\n", destnode));
2413 talloc_free(tmp_ctx);
2418 c.old_flags = nodemap->nodes[destnode].flags;
2419 c.new_flags = c.old_flags;
2421 c.new_flags &= ~clear;
2423 data.dsize = sizeof(c);
2424 data.dptr = (unsigned char *)&c;
2426 /* send the flags update to all connected nodes */
2427 nodes = list_of_connected_nodes(ctdb, nodemap, tmp_ctx, true);
2429 if (ctdb_client_async_control(ctdb, CTDB_CONTROL_MODIFY_FLAGS,
2431 timeout, false, data,
2434 DEBUG(DEBUG_ERR, (__location__ " Unable to update nodeflags on remote nodes\n"));
2436 talloc_free(tmp_ctx);
2440 talloc_free(tmp_ctx);
2448 int ctdb_ctrl_get_all_tunables(struct ctdb_context *ctdb,
2449 struct timeval timeout,
2451 struct ctdb_tunable *tunables)
2457 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_GET_ALL_TUNABLES, 0, tdb_null, ctdb,
2458 &outdata, &res, &timeout, NULL);
2459 if (ret != 0 || res != 0) {
2460 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for get all tunables failed\n"));
2464 if (outdata.dsize != sizeof(*tunables)) {
2465 DEBUG(DEBUG_ERR,(__location__ " bad data size %u in ctdb_ctrl_get_all_tunables should be %u\n",
2466 (unsigned)outdata.dsize, (unsigned)sizeof(*tunables)));
2470 *tunables = *(struct ctdb_tunable *)outdata.dptr;
2471 talloc_free(outdata.dptr);
2476 add a public address to a node
2478 int ctdb_ctrl_add_public_ip(struct ctdb_context *ctdb,
2479 struct timeval timeout,
2481 struct ctdb_control_ip_iface *pub)
2487 data.dsize = offsetof(struct ctdb_control_ip_iface, iface) + pub->len;
2488 data.dptr = (unsigned char *)pub;
2490 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_ADD_PUBLIC_IP, 0, data, NULL,
2491 NULL, &res, &timeout, NULL);
2492 if (ret != 0 || res != 0) {
2493 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for add_public_ip failed\n"));
2501 delete a public address from a node
2503 int ctdb_ctrl_del_public_ip(struct ctdb_context *ctdb,
2504 struct timeval timeout,
2506 struct ctdb_control_ip_iface *pub)
2512 data.dsize = offsetof(struct ctdb_control_ip_iface, iface) + pub->len;
2513 data.dptr = (unsigned char *)pub;
2515 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_DEL_PUBLIC_IP, 0, data, NULL,
2516 NULL, &res, &timeout, NULL);
2517 if (ret != 0 || res != 0) {
2518 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for del_public_ip failed\n"));
2526 kill a tcp connection
2528 int ctdb_ctrl_killtcp(struct ctdb_context *ctdb,
2529 struct timeval timeout,
2531 struct ctdb_control_killtcp *killtcp)
2537 data.dsize = sizeof(struct ctdb_control_killtcp);
2538 data.dptr = (unsigned char *)killtcp;
2540 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_KILL_TCP, 0, data, NULL,
2541 NULL, &res, &timeout, NULL);
2542 if (ret != 0 || res != 0) {
2543 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for killtcp failed\n"));
2553 int ctdb_ctrl_gratious_arp(struct ctdb_context *ctdb,
2554 struct timeval timeout,
2556 ctdb_sock_addr *addr,
2562 struct ctdb_control_gratious_arp *gratious_arp;
2563 TALLOC_CTX *tmp_ctx = talloc_new(ctdb);
2566 len = strlen(ifname)+1;
2567 gratious_arp = talloc_size(tmp_ctx,
2568 offsetof(struct ctdb_control_gratious_arp, iface) + len);
2569 CTDB_NO_MEMORY(ctdb, gratious_arp);
2571 gratious_arp->addr = *addr;
2572 gratious_arp->len = len;
2573 memcpy(&gratious_arp->iface[0], ifname, len);
2576 data.dsize = offsetof(struct ctdb_control_gratious_arp, iface) + len;
2577 data.dptr = (unsigned char *)gratious_arp;
2579 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_SEND_GRATIOUS_ARP, 0, data, NULL,
2580 NULL, &res, &timeout, NULL);
2581 if (ret != 0 || res != 0) {
2582 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for gratious_arp failed\n"));
2583 talloc_free(tmp_ctx);
2587 talloc_free(tmp_ctx);
2592 get a list of all tcp tickles that a node knows about for a particular vnn
2594 int ctdb_ctrl_get_tcp_tickles(struct ctdb_context *ctdb,
2595 struct timeval timeout, uint32_t destnode,
2596 TALLOC_CTX *mem_ctx,
2597 ctdb_sock_addr *addr,
2598 struct ctdb_control_tcp_tickle_list **list)
2601 TDB_DATA data, outdata;
2604 data.dptr = (uint8_t*)addr;
2605 data.dsize = sizeof(ctdb_sock_addr);
2607 ret = ctdb_control(ctdb, destnode, 0,
2608 CTDB_CONTROL_GET_TCP_TICKLE_LIST, 0, data,
2609 mem_ctx, &outdata, &status, NULL, NULL);
2610 if (ret != 0 || status != 0) {
2611 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for get tcp tickles failed\n"));
2615 *list = (struct ctdb_control_tcp_tickle_list *)outdata.dptr;
2621 register a server id
2623 int ctdb_ctrl_register_server_id(struct ctdb_context *ctdb,
2624 struct timeval timeout,
2625 struct ctdb_server_id *id)
2631 data.dsize = sizeof(struct ctdb_server_id);
2632 data.dptr = (unsigned char *)id;
2634 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, 0,
2635 CTDB_CONTROL_REGISTER_SERVER_ID,
2637 NULL, &res, &timeout, NULL);
2638 if (ret != 0 || res != 0) {
2639 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for register server id failed\n"));
2647 unregister a server id
2649 int ctdb_ctrl_unregister_server_id(struct ctdb_context *ctdb,
2650 struct timeval timeout,
2651 struct ctdb_server_id *id)
2657 data.dsize = sizeof(struct ctdb_server_id);
2658 data.dptr = (unsigned char *)id;
2660 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, 0,
2661 CTDB_CONTROL_UNREGISTER_SERVER_ID,
2663 NULL, &res, &timeout, NULL);
2664 if (ret != 0 || res != 0) {
2665 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for unregister server id failed\n"));
2674 check if a server id exists
2676 if a server id does exist, return *status == 1, otherwise *status == 0
2678 int ctdb_ctrl_check_server_id(struct ctdb_context *ctdb,
2679 struct timeval timeout,
2681 struct ctdb_server_id *id,
2688 data.dsize = sizeof(struct ctdb_server_id);
2689 data.dptr = (unsigned char *)id;
2691 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_CHECK_SERVER_ID,
2693 NULL, &res, &timeout, NULL);
2695 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for check server id failed\n"));
2709 get the list of server ids that are registered on a node
2711 int ctdb_ctrl_get_server_id_list(struct ctdb_context *ctdb,
2712 TALLOC_CTX *mem_ctx,
2713 struct timeval timeout, uint32_t destnode,
2714 struct ctdb_server_id_list **svid_list)
2720 ret = ctdb_control(ctdb, destnode, 0,
2721 CTDB_CONTROL_GET_SERVER_ID_LIST, 0, tdb_null,
2722 mem_ctx, &outdata, &res, &timeout, NULL);
2723 if (ret != 0 || res != 0) {
2724 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for get_server_id_list failed\n"));
2728 *svid_list = (struct ctdb_server_id_list *)talloc_steal(mem_ctx, outdata.dptr);
2734 initialise the ctdb daemon for client applications
2736 NOTE: In current code the daemon does not fork. This is for testing purposes only
2737 and to simplify the code.
2739 struct ctdb_context *ctdb_init(struct event_context *ev)
2742 struct ctdb_context *ctdb;
2744 ctdb = talloc_zero(ev, struct ctdb_context);
2746 DEBUG(DEBUG_ERR,(__location__ " talloc_zero failed.\n"));
2750 ctdb->idr = idr_init(ctdb);
2751 CTDB_NO_MEMORY_NULL(ctdb, ctdb->idr);
2753 ret = ctdb_set_socketname(ctdb, CTDB_PATH);
2755 DEBUG(DEBUG_ERR,(__location__ " ctdb_set_socketname failed.\n"));
2767 void ctdb_set_flags(struct ctdb_context *ctdb, unsigned flags)
2769 ctdb->flags |= flags;
2773 setup the local socket name
2775 int ctdb_set_socketname(struct ctdb_context *ctdb, const char *socketname)
2777 ctdb->daemon.name = talloc_strdup(ctdb, socketname);
2778 CTDB_NO_MEMORY(ctdb, ctdb->daemon.name);
2784 return the pnn of this node
2786 uint32_t ctdb_get_pnn(struct ctdb_context *ctdb)
2793 get the uptime of a remote node
2795 struct ctdb_client_control_state *
2796 ctdb_ctrl_uptime_send(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode)
2798 return ctdb_control_send(ctdb, destnode, 0,
2799 CTDB_CONTROL_UPTIME, 0, tdb_null,
2800 mem_ctx, &timeout, NULL);
2803 int ctdb_ctrl_uptime_recv(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state, struct ctdb_uptime **uptime)
2809 ret = ctdb_control_recv(ctdb, state, mem_ctx, &outdata, &res, NULL);
2810 if (ret != 0 || res != 0) {
2811 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_uptime_recv failed\n"));
2815 *uptime = (struct ctdb_uptime *)talloc_steal(mem_ctx, outdata.dptr);
2820 int ctdb_ctrl_uptime(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode, struct ctdb_uptime **uptime)
2822 struct ctdb_client_control_state *state;
2824 state = ctdb_ctrl_uptime_send(ctdb, mem_ctx, timeout, destnode);
2825 return ctdb_ctrl_uptime_recv(ctdb, mem_ctx, state, uptime);
2829 send a control to execute the "recovered" event script on a node
2831 int ctdb_ctrl_end_recovery(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
2836 ret = ctdb_control(ctdb, destnode, 0,
2837 CTDB_CONTROL_END_RECOVERY, 0, tdb_null,
2838 NULL, NULL, &status, &timeout, NULL);
2839 if (ret != 0 || status != 0) {
2840 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for end_recovery failed\n"));
2848 callback for the async helpers used when sending the same control
2849 to multiple nodes in parallell.
2851 static void async_callback(struct ctdb_client_control_state *state)
2853 struct client_async_data *data = talloc_get_type(state->async.private_data, struct client_async_data);
2854 struct ctdb_context *ctdb = talloc_get_type(state->ctdb, struct ctdb_context);
2858 uint32_t destnode = state->c->hdr.destnode;
2860 /* one more node has responded with recmode data */
2863 /* if we failed to push the db, then return an error and let
2864 the main loop try again.
2866 if (state->state != CTDB_CONTROL_DONE) {
2867 if ( !data->dont_log_errors) {
2868 DEBUG(DEBUG_ERR,("Async operation failed with state %d, opcode:%u\n", state->state, data->opcode));
2871 if (data->fail_callback) {
2872 data->fail_callback(ctdb, destnode, res, outdata,
2873 data->callback_data);
2878 state->async.fn = NULL;
2880 ret = ctdb_control_recv(ctdb, state, data, &outdata, &res, NULL);
2881 if ((ret != 0) || (res != 0)) {
2882 if ( !data->dont_log_errors) {
2883 DEBUG(DEBUG_ERR,("Async operation failed with ret=%d res=%d opcode=%u\n", ret, (int)res, data->opcode));
2886 if (data->fail_callback) {
2887 data->fail_callback(ctdb, destnode, res, outdata,
2888 data->callback_data);
2891 if ((ret == 0) && (data->callback != NULL)) {
2892 data->callback(ctdb, destnode, res, outdata,
2893 data->callback_data);
2898 void ctdb_client_async_add(struct client_async_data *data, struct ctdb_client_control_state *state)
2900 /* set up the callback functions */
2901 state->async.fn = async_callback;
2902 state->async.private_data = data;
2904 /* one more control to wait for to complete */
2909 /* wait for up to the maximum number of seconds allowed
2910 or until all nodes we expect a response from has replied
2912 int ctdb_client_async_wait(struct ctdb_context *ctdb, struct client_async_data *data)
2914 while (data->count > 0) {
2915 event_loop_once(ctdb->ev);
2917 if (data->fail_count != 0) {
2918 if (!data->dont_log_errors) {
2919 DEBUG(DEBUG_ERR,("Async wait failed - fail_count=%u\n",
2929 perform a simple control on the listed nodes
2930 The control cannot return data
2932 int ctdb_client_async_control(struct ctdb_context *ctdb,
2933 enum ctdb_controls opcode,
2936 struct timeval timeout,
2937 bool dont_log_errors,
2939 client_async_callback client_callback,
2940 client_async_callback fail_callback,
2941 void *callback_data)
2943 struct client_async_data *async_data;
2944 struct ctdb_client_control_state *state;
2947 async_data = talloc_zero(ctdb, struct client_async_data);
2948 CTDB_NO_MEMORY_FATAL(ctdb, async_data);
2949 async_data->dont_log_errors = dont_log_errors;
2950 async_data->callback = client_callback;
2951 async_data->fail_callback = fail_callback;
2952 async_data->callback_data = callback_data;
2953 async_data->opcode = opcode;
2955 num_nodes = talloc_get_size(nodes) / sizeof(uint32_t);
2957 /* loop over all nodes and send an async control to each of them */
2958 for (j=0; j<num_nodes; j++) {
2959 uint32_t pnn = nodes[j];
2961 state = ctdb_control_send(ctdb, pnn, srvid, opcode,
2962 0, data, async_data, &timeout, NULL);
2963 if (state == NULL) {
2964 DEBUG(DEBUG_ERR,(__location__ " Failed to call async control %u\n", (unsigned)opcode));
2965 talloc_free(async_data);
2969 ctdb_client_async_add(async_data, state);
2972 if (ctdb_client_async_wait(ctdb, async_data) != 0) {
2973 talloc_free(async_data);
2977 talloc_free(async_data);
2981 uint32_t *list_of_vnnmap_nodes(struct ctdb_context *ctdb,
2982 struct ctdb_vnn_map *vnn_map,
2983 TALLOC_CTX *mem_ctx,
2986 int i, j, num_nodes;
2989 for (i=num_nodes=0;i<vnn_map->size;i++) {
2990 if (vnn_map->map[i] == ctdb->pnn && !include_self) {
2996 nodes = talloc_array(mem_ctx, uint32_t, num_nodes);
2997 CTDB_NO_MEMORY_FATAL(ctdb, nodes);
2999 for (i=j=0;i<vnn_map->size;i++) {
3000 if (vnn_map->map[i] == ctdb->pnn && !include_self) {
3003 nodes[j++] = vnn_map->map[i];
3009 uint32_t *list_of_active_nodes(struct ctdb_context *ctdb,
3010 struct ctdb_node_map *node_map,
3011 TALLOC_CTX *mem_ctx,
3014 int i, j, num_nodes;
3017 for (i=num_nodes=0;i<node_map->num;i++) {
3018 if (node_map->nodes[i].flags & NODE_FLAGS_INACTIVE) {
3021 if (node_map->nodes[i].pnn == ctdb->pnn && !include_self) {
3027 nodes = talloc_array(mem_ctx, uint32_t, num_nodes);
3028 CTDB_NO_MEMORY_FATAL(ctdb, nodes);
3030 for (i=j=0;i<node_map->num;i++) {
3031 if (node_map->nodes[i].flags & NODE_FLAGS_INACTIVE) {
3034 if (node_map->nodes[i].pnn == ctdb->pnn && !include_self) {
3037 nodes[j++] = node_map->nodes[i].pnn;
3043 uint32_t *list_of_active_nodes_except_pnn(struct ctdb_context *ctdb,
3044 struct ctdb_node_map *node_map,
3045 TALLOC_CTX *mem_ctx,
3048 int i, j, num_nodes;
3051 for (i=num_nodes=0;i<node_map->num;i++) {
3052 if (node_map->nodes[i].flags & NODE_FLAGS_INACTIVE) {
3055 if (node_map->nodes[i].pnn == pnn) {
3061 nodes = talloc_array(mem_ctx, uint32_t, num_nodes);
3062 CTDB_NO_MEMORY_FATAL(ctdb, nodes);
3064 for (i=j=0;i<node_map->num;i++) {
3065 if (node_map->nodes[i].flags & NODE_FLAGS_INACTIVE) {
3068 if (node_map->nodes[i].pnn == pnn) {
3071 nodes[j++] = node_map->nodes[i].pnn;
3077 uint32_t *list_of_connected_nodes(struct ctdb_context *ctdb,
3078 struct ctdb_node_map *node_map,
3079 TALLOC_CTX *mem_ctx,
3082 int i, j, num_nodes;
3085 for (i=num_nodes=0;i<node_map->num;i++) {
3086 if (node_map->nodes[i].flags & NODE_FLAGS_DISCONNECTED) {
3089 if (node_map->nodes[i].pnn == ctdb->pnn && !include_self) {
3095 nodes = talloc_array(mem_ctx, uint32_t, num_nodes);
3096 CTDB_NO_MEMORY_FATAL(ctdb, nodes);
3098 for (i=j=0;i<node_map->num;i++) {
3099 if (node_map->nodes[i].flags & NODE_FLAGS_DISCONNECTED) {
3102 if (node_map->nodes[i].pnn == ctdb->pnn && !include_self) {
3105 nodes[j++] = node_map->nodes[i].pnn;
3112 this is used to test if a pnn lock exists and if it exists will return
3113 the number of connections that pnn has reported or -1 if that recovery
3114 daemon is not running.
3117 ctdb_read_pnn_lock(int fd, int32_t pnn)
3122 lock.l_type = F_WRLCK;
3123 lock.l_whence = SEEK_SET;
3128 if (fcntl(fd, F_GETLK, &lock) != 0) {
3129 DEBUG(DEBUG_ERR, (__location__ " F_GETLK failed with %s\n", strerror(errno)));
3133 if (lock.l_type == F_UNLCK) {
3137 if (pread(fd, &c, 1, pnn) == -1) {
3138 DEBUG(DEBUG_CRIT,(__location__ " failed read pnn count - %s\n", strerror(errno)));
3146 get capabilities of a remote node
3148 struct ctdb_client_control_state *
3149 ctdb_ctrl_getcapabilities_send(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct timeval timeout, uint32_t destnode)
3151 return ctdb_control_send(ctdb, destnode, 0,
3152 CTDB_CONTROL_GET_CAPABILITIES, 0, tdb_null,
3153 mem_ctx, &timeout, NULL);
3156 int ctdb_ctrl_getcapabilities_recv(struct ctdb_context *ctdb, TALLOC_CTX *mem_ctx, struct ctdb_client_control_state *state, uint32_t *capabilities)
3162 ret = ctdb_control_recv(ctdb, state, mem_ctx, &outdata, &res, NULL);
3163 if ( (ret != 0) || (res != 0) ) {
3164 DEBUG(DEBUG_ERR,(__location__ " ctdb_ctrl_getcapabilities_recv failed\n"));
3169 *capabilities = *((uint32_t *)outdata.dptr);
3175 int ctdb_ctrl_getcapabilities(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t *capabilities)
3177 struct ctdb_client_control_state *state;
3178 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
3181 state = ctdb_ctrl_getcapabilities_send(ctdb, tmp_ctx, timeout, destnode);
3182 ret = ctdb_ctrl_getcapabilities_recv(ctdb, tmp_ctx, state, capabilities);
3183 talloc_free(tmp_ctx);
3188 * check whether a transaction is active on a given db on a given node
3190 int32_t ctdb_ctrl_transaction_active(struct ctdb_context *ctdb,
3198 indata.dptr = (uint8_t *)&db_id;
3199 indata.dsize = sizeof(db_id);
3201 ret = ctdb_control(ctdb, destnode, 0,
3202 CTDB_CONTROL_TRANS2_ACTIVE,
3203 0, indata, NULL, NULL, &status,
3207 DEBUG(DEBUG_ERR, (__location__ " ctdb control for transaction_active failed\n"));
3215 struct ctdb_transaction_handle {
3216 struct ctdb_db_context *ctdb_db;
3219 * we store the reads and writes done under a transaction:
3220 * - one list stores both reads and writes (m_all),
3221 * - the other just writes (m_write)
3223 struct ctdb_marshall_buffer *m_all;
3224 struct ctdb_marshall_buffer *m_write;
3227 /* start a transaction on a database */
3228 static int ctdb_transaction_destructor(struct ctdb_transaction_handle *h)
3230 tdb_transaction_cancel(h->ctdb_db->ltdb->tdb);
3234 /* start a transaction on a database */
3235 static int ctdb_transaction_fetch_start(struct ctdb_transaction_handle *h)
3237 struct ctdb_record_handle *rh;
3240 struct ctdb_ltdb_header header;
3241 TALLOC_CTX *tmp_ctx;
3242 const char *keyname = CTDB_TRANSACTION_LOCK_KEY;
3244 struct ctdb_db_context *ctdb_db = h->ctdb_db;
3248 key.dptr = discard_const(keyname);
3249 key.dsize = strlen(keyname);
3251 if (!ctdb_db->persistent) {
3252 DEBUG(DEBUG_ERR,(__location__ " Attempted transaction on non-persistent database\n"));
3257 tmp_ctx = talloc_new(h);
3259 rh = ctdb_fetch_lock(ctdb_db, tmp_ctx, key, NULL);
3261 DEBUG(DEBUG_ERR,(__location__ " Failed to fetch_lock database\n"));
3262 talloc_free(tmp_ctx);
3266 status = ctdb_ctrl_transaction_active(ctdb_db->ctdb,
3270 unsigned long int usec = (1000 + random()) % 100000;
3271 DEBUG(DEBUG_DEBUG, (__location__ " transaction is active "
3272 "on db_id[0x%08x]. waiting for %lu "
3274 ctdb_db->db_id, usec));
3275 talloc_free(tmp_ctx);
3281 * store the pid in the database:
3282 * it is not enough that the node is dmaster...
3285 data.dptr = (unsigned char *)&pid;
3286 data.dsize = sizeof(pid_t);
3288 rh->header.dmaster = ctdb_db->ctdb->pnn;
3289 ret = ctdb_ltdb_store(ctdb_db, key, &(rh->header), data);
3291 DEBUG(DEBUG_ERR, (__location__ " Failed to store pid in "
3292 "transaction record\n"));
3293 talloc_free(tmp_ctx);
3299 ret = tdb_transaction_start(ctdb_db->ltdb->tdb);
3301 DEBUG(DEBUG_ERR,(__location__ " Failed to start tdb transaction\n"));
3302 talloc_free(tmp_ctx);
3306 ret = ctdb_ltdb_fetch(ctdb_db, key, &header, tmp_ctx, &data);
3308 DEBUG(DEBUG_ERR,(__location__ " Failed to re-fetch transaction "
3309 "lock record inside transaction\n"));
3310 tdb_transaction_cancel(ctdb_db->ltdb->tdb);
3311 talloc_free(tmp_ctx);
3315 if (header.dmaster != ctdb_db->ctdb->pnn) {
3316 DEBUG(DEBUG_DEBUG,(__location__ " not dmaster any more on "
3317 "transaction lock record\n"));
3318 tdb_transaction_cancel(ctdb_db->ltdb->tdb);
3319 talloc_free(tmp_ctx);
3323 if ((data.dsize != sizeof(pid_t)) || (*(pid_t *)(data.dptr) != pid)) {
3324 DEBUG(DEBUG_DEBUG, (__location__ " my pid is not stored in "
3325 "the transaction lock record\n"));
3326 tdb_transaction_cancel(ctdb_db->ltdb->tdb);
3327 talloc_free(tmp_ctx);
3331 talloc_free(tmp_ctx);
3337 /* start a transaction on a database */
3338 struct ctdb_transaction_handle *ctdb_transaction_start(struct ctdb_db_context *ctdb_db,
3339 TALLOC_CTX *mem_ctx)
3341 struct ctdb_transaction_handle *h;
3344 h = talloc_zero(mem_ctx, struct ctdb_transaction_handle);
3346 DEBUG(DEBUG_ERR,(__location__ " oom for transaction handle\n"));
3350 h->ctdb_db = ctdb_db;
3352 ret = ctdb_transaction_fetch_start(h);
3358 talloc_set_destructor(h, ctdb_transaction_destructor);
3366 fetch a record inside a transaction
3368 int ctdb_transaction_fetch(struct ctdb_transaction_handle *h,
3369 TALLOC_CTX *mem_ctx,
3370 TDB_DATA key, TDB_DATA *data)
3372 struct ctdb_ltdb_header header;
3375 ZERO_STRUCT(header);
3377 ret = ctdb_ltdb_fetch(h->ctdb_db, key, &header, mem_ctx, data);
3378 if (ret == -1 && header.dmaster == (uint32_t)-1) {
3379 /* record doesn't exist yet */
3388 if (!h->in_replay) {
3389 h->m_all = ctdb_marshall_add(h, h->m_all, h->ctdb_db->db_id, 1, key, NULL, *data);
3390 if (h->m_all == NULL) {
3391 DEBUG(DEBUG_ERR,(__location__ " Failed to add to marshalling record\n"));
3400 stores a record inside a transaction
3402 int ctdb_transaction_store(struct ctdb_transaction_handle *h,
3403 TDB_DATA key, TDB_DATA data)
3405 TALLOC_CTX *tmp_ctx = talloc_new(h);
3406 struct ctdb_ltdb_header header;
3410 ZERO_STRUCT(header);
3412 /* we need the header so we can update the RSN */
3413 ret = ctdb_ltdb_fetch(h->ctdb_db, key, &header, tmp_ctx, &olddata);
3414 if (ret == -1 && header.dmaster == (uint32_t)-1) {
3415 /* the record doesn't exist - create one with us as dmaster.
3416 This is only safe because we are in a transaction and this
3417 is a persistent database */
3418 ZERO_STRUCT(header);
3419 } else if (ret != 0) {
3420 DEBUG(DEBUG_ERR,(__location__ " Failed to fetch record\n"));
3421 talloc_free(tmp_ctx);
3425 if (data.dsize == olddata.dsize &&
3426 memcmp(data.dptr, olddata.dptr, data.dsize) == 0) {
3427 /* save writing the same data */
3428 talloc_free(tmp_ctx);
3432 header.dmaster = h->ctdb_db->ctdb->pnn;
3435 if (!h->in_replay) {
3436 h->m_all = ctdb_marshall_add(h, h->m_all, h->ctdb_db->db_id, 0, key, NULL, data);
3437 if (h->m_all == NULL) {
3438 DEBUG(DEBUG_ERR,(__location__ " Failed to add to marshalling record\n"));
3439 talloc_free(tmp_ctx);
3444 h->m_write = ctdb_marshall_add(h, h->m_write, h->ctdb_db->db_id, 0, key, &header, data);
3445 if (h->m_write == NULL) {
3446 DEBUG(DEBUG_ERR,(__location__ " Failed to add to marshalling record\n"));
3447 talloc_free(tmp_ctx);
3451 ret = ctdb_ltdb_store(h->ctdb_db, key, &header, data);
3453 talloc_free(tmp_ctx);
3459 replay a transaction
3461 static int ctdb_replay_transaction(struct ctdb_transaction_handle *h)
3464 struct ctdb_rec_data *rec = NULL;
3466 h->in_replay = true;
3467 talloc_free(h->m_write);
3470 ret = ctdb_transaction_fetch_start(h);
3475 for (i=0;i<h->m_all->count;i++) {
3478 rec = ctdb_marshall_loop_next(h->m_all, rec, NULL, NULL, &key, &data);
3480 DEBUG(DEBUG_ERR, (__location__ " Out of records in ctdb_replay_transaction?\n"));
3484 if (rec->reqid == 0) {
3486 if (ctdb_transaction_store(h, key, data) != 0) {
3491 TALLOC_CTX *tmp_ctx = talloc_new(h);
3493 if (ctdb_transaction_fetch(h, tmp_ctx, key, &data2) != 0) {
3494 talloc_free(tmp_ctx);
3497 if (data2.dsize != data.dsize ||
3498 memcmp(data2.dptr, data.dptr, data.dsize) != 0) {
3499 /* the record has changed on us - we have to give up */
3500 talloc_free(tmp_ctx);
3503 talloc_free(tmp_ctx);
3510 tdb_transaction_cancel(h->ctdb_db->ltdb->tdb);
3516 commit a transaction
3518 int ctdb_transaction_commit(struct ctdb_transaction_handle *h)
3522 struct ctdb_context *ctdb = h->ctdb_db->ctdb;
3523 struct timeval timeout;
3524 enum ctdb_controls failure_control = CTDB_CONTROL_TRANS2_ERROR;
3526 talloc_set_destructor(h, NULL);
3528 /* our commit strategy is quite complex.
3530 - we first try to commit the changes to all other nodes
3532 - if that works, then we commit locally and we are done
3534 - if a commit on another node fails, then we need to cancel
3535 the transaction, then restart the transaction (thus
3536 opening a window of time for a pending recovery to
3537 complete), then replay the transaction, checking all the
3538 reads and writes (checking that reads give the same data,
3539 and writes succeed). Then we retry the transaction to the
3544 if (h->m_write == NULL) {
3545 /* no changes were made */
3546 tdb_transaction_cancel(h->ctdb_db->ltdb->tdb);
3551 /* tell ctdbd to commit to the other nodes */
3552 timeout = timeval_current_ofs(1, 0);
3553 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, h->ctdb_db->db_id,
3554 retries==0?CTDB_CONTROL_TRANS2_COMMIT:CTDB_CONTROL_TRANS2_COMMIT_RETRY, 0,
3555 ctdb_marshall_finish(h->m_write), NULL, NULL, &status,
3557 if (ret != 0 || status != 0) {
3558 tdb_transaction_cancel(h->ctdb_db->ltdb->tdb);
3559 DEBUG(DEBUG_NOTICE, (__location__ " transaction commit%s failed"
3560 ", retrying after 1 second...\n",
3561 (retries==0)?"":"retry "));
3565 failure_control = CTDB_CONTROL_TRANS2_ERROR;
3567 /* work out what error code we will give if we
3568 have to fail the operation */
3569 switch ((enum ctdb_trans2_commit_error)status) {
3570 case CTDB_TRANS2_COMMIT_SUCCESS:
3571 case CTDB_TRANS2_COMMIT_SOMEFAIL:
3572 case CTDB_TRANS2_COMMIT_TIMEOUT:
3573 failure_control = CTDB_CONTROL_TRANS2_ERROR;
3575 case CTDB_TRANS2_COMMIT_ALLFAIL:
3576 failure_control = CTDB_CONTROL_TRANS2_FINISHED;
3581 if (++retries == 100) {
3582 DEBUG(DEBUG_ERR,(__location__ " Giving up transaction on db 0x%08x after %d retries failure_control=%u\n",
3583 h->ctdb_db->db_id, retries, (unsigned)failure_control));
3584 ctdb_control(ctdb, CTDB_CURRENT_NODE, h->ctdb_db->db_id,
3585 failure_control, CTDB_CTRL_FLAG_NOREPLY,
3586 tdb_null, NULL, NULL, NULL, NULL, NULL);
3591 if (ctdb_replay_transaction(h) != 0) {
3592 DEBUG(DEBUG_ERR, (__location__ " Failed to replay "
3593 "transaction on db 0x%08x, "
3594 "failure control =%u\n",
3596 (unsigned)failure_control));
3597 ctdb_control(ctdb, CTDB_CURRENT_NODE, h->ctdb_db->db_id,
3598 failure_control, CTDB_CTRL_FLAG_NOREPLY,
3599 tdb_null, NULL, NULL, NULL, NULL, NULL);
3605 failure_control = CTDB_CONTROL_TRANS2_ERROR;
3608 /* do the real commit locally */
3609 ret = tdb_transaction_commit(h->ctdb_db->ltdb->tdb);
3611 DEBUG(DEBUG_ERR, (__location__ " Failed to commit transaction "
3612 "on db id 0x%08x locally, "
3613 "failure_control=%u\n",
3615 (unsigned)failure_control));
3616 ctdb_control(ctdb, CTDB_CURRENT_NODE, h->ctdb_db->db_id,
3617 failure_control, CTDB_CTRL_FLAG_NOREPLY,
3618 tdb_null, NULL, NULL, NULL, NULL, NULL);
3623 /* tell ctdbd that we are finished with our local commit */
3624 ctdb_control(ctdb, CTDB_CURRENT_NODE, h->ctdb_db->db_id,
3625 CTDB_CONTROL_TRANS2_FINISHED, CTDB_CTRL_FLAG_NOREPLY,
3626 tdb_null, NULL, NULL, NULL, NULL, NULL);
3632 recovery daemon ping to main daemon
3634 int ctdb_ctrl_recd_ping(struct ctdb_context *ctdb)
3639 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, 0, CTDB_CONTROL_RECD_PING, 0, tdb_null,
3640 ctdb, NULL, &res, NULL, NULL);
3641 if (ret != 0 || res != 0) {
3642 DEBUG(DEBUG_ERR,("Failed to send recd ping\n"));
3649 /* when forking the main daemon and the child process needs to connect back
3650 * to the daemon as a client process, this function can be used to change
3651 * the ctdb context from daemon into client mode
3653 int switch_from_server_to_client(struct ctdb_context *ctdb)
3657 /* shutdown the transport */
3658 if (ctdb->methods) {
3659 ctdb->methods->shutdown(ctdb);
3662 /* get a new event context */
3663 talloc_free(ctdb->ev);
3664 ctdb->ev = event_context_init(ctdb);
3666 close(ctdb->daemon.sd);
3667 ctdb->daemon.sd = -1;
3669 /* the client does not need to be realtime */
3670 if (ctdb->do_setsched) {
3671 ctdb_restore_scheduler(ctdb);
3674 /* initialise ctdb */
3675 ret = ctdb_socket_connect(ctdb);
3677 DEBUG(DEBUG_ALERT, (__location__ " Failed to init ctdb client\n"));
3685 get the status of running the monitor eventscripts: NULL means never run.
3687 int ctdb_ctrl_getscriptstatus(struct ctdb_context *ctdb,
3688 struct timeval timeout, uint32_t destnode,
3689 TALLOC_CTX *mem_ctx, enum ctdb_eventscript_call type,
3690 struct ctdb_scripts_wire **script_status)
3693 TDB_DATA outdata, indata;
3695 uint32_t uinttype = type;
3697 indata.dptr = (uint8_t *)&uinttype;
3698 indata.dsize = sizeof(uinttype);
3700 ret = ctdb_control(ctdb, destnode, 0,
3701 CTDB_CONTROL_GET_EVENT_SCRIPT_STATUS, 0, indata,
3702 mem_ctx, &outdata, &res, &timeout, NULL);
3703 if (ret != 0 || res != 0) {
3704 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for getscriptstatus failed ret:%d res:%d\n", ret, res));
3708 if (outdata.dsize == 0) {
3709 *script_status = NULL;
3711 *script_status = (struct ctdb_scripts_wire *)talloc_memdup(mem_ctx, outdata.dptr, outdata.dsize);
3712 talloc_free(outdata.dptr);
3719 tell the main daemon how long it took to lock the reclock file
3721 int ctdb_ctrl_report_recd_lock_latency(struct ctdb_context *ctdb, struct timeval timeout, double latency)
3727 data.dptr = (uint8_t *)&latency;
3728 data.dsize = sizeof(latency);
3730 ret = ctdb_control(ctdb, CTDB_CURRENT_NODE, 0, CTDB_CONTROL_RECD_RECLOCK_LATENCY, 0, data,
3731 ctdb, NULL, &res, NULL, NULL);
3732 if (ret != 0 || res != 0) {
3733 DEBUG(DEBUG_ERR,("Failed to send recd reclock latency\n"));
3741 get the name of the reclock file
3743 int ctdb_ctrl_getreclock(struct ctdb_context *ctdb, struct timeval timeout,
3744 uint32_t destnode, TALLOC_CTX *mem_ctx,
3751 ret = ctdb_control(ctdb, destnode, 0,
3752 CTDB_CONTROL_GET_RECLOCK_FILE, 0, tdb_null,
3753 mem_ctx, &data, &res, &timeout, NULL);
3754 if (ret != 0 || res != 0) {
3758 if (data.dsize == 0) {
3761 *name = talloc_strdup(mem_ctx, discard_const(data.dptr));
3763 talloc_free(data.dptr);
3769 set the reclock filename for a node
3771 int ctdb_ctrl_setreclock(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, const char *reclock)
3777 if (reclock == NULL) {
3781 data.dsize = strlen(reclock) + 1;
3782 data.dptr = discard_const(reclock);
3785 ret = ctdb_control(ctdb, destnode, 0,
3786 CTDB_CONTROL_SET_RECLOCK_FILE, 0, data,
3787 NULL, NULL, &res, &timeout, NULL);
3788 if (ret != 0 || res != 0) {
3789 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setreclock failed\n"));
3799 int ctdb_ctrl_stop_node(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
3804 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_STOP_NODE, 0, tdb_null,
3805 ctdb, NULL, &res, &timeout, NULL);
3806 if (ret != 0 || res != 0) {
3807 DEBUG(DEBUG_ERR,("Failed to stop node\n"));
3817 int ctdb_ctrl_continue_node(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode)
3821 ret = ctdb_control(ctdb, destnode, 0, CTDB_CONTROL_CONTINUE_NODE, 0, tdb_null,
3822 ctdb, NULL, NULL, &timeout, NULL);
3824 DEBUG(DEBUG_ERR,("Failed to continue node\n"));
3832 set the natgw state for a node
3834 int ctdb_ctrl_setnatgwstate(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t natgwstate)
3840 data.dsize = sizeof(natgwstate);
3841 data.dptr = (uint8_t *)&natgwstate;
3843 ret = ctdb_control(ctdb, destnode, 0,
3844 CTDB_CONTROL_SET_NATGWSTATE, 0, data,
3845 NULL, NULL, &res, &timeout, NULL);
3846 if (ret != 0 || res != 0) {
3847 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setnatgwstate failed\n"));
3855 set the lmaster role for a node
3857 int ctdb_ctrl_setlmasterrole(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t lmasterrole)
3863 data.dsize = sizeof(lmasterrole);
3864 data.dptr = (uint8_t *)&lmasterrole;
3866 ret = ctdb_control(ctdb, destnode, 0,
3867 CTDB_CONTROL_SET_LMASTERROLE, 0, data,
3868 NULL, NULL, &res, &timeout, NULL);
3869 if (ret != 0 || res != 0) {
3870 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setlmasterrole failed\n"));
3878 set the recmaster role for a node
3880 int ctdb_ctrl_setrecmasterrole(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t recmasterrole)
3886 data.dsize = sizeof(recmasterrole);
3887 data.dptr = (uint8_t *)&recmasterrole;
3889 ret = ctdb_control(ctdb, destnode, 0,
3890 CTDB_CONTROL_SET_RECMASTERROLE, 0, data,
3891 NULL, NULL, &res, &timeout, NULL);
3892 if (ret != 0 || res != 0) {
3893 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for setrecmasterrole failed\n"));
3900 /* enable an eventscript
3902 int ctdb_ctrl_enablescript(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, const char *script)
3908 data.dsize = strlen(script) + 1;
3909 data.dptr = discard_const(script);
3911 ret = ctdb_control(ctdb, destnode, 0,
3912 CTDB_CONTROL_ENABLE_SCRIPT, 0, data,
3913 NULL, NULL, &res, &timeout, NULL);
3914 if (ret != 0 || res != 0) {
3915 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for enablescript failed\n"));
3922 /* disable an eventscript
3924 int ctdb_ctrl_disablescript(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, const char *script)
3930 data.dsize = strlen(script) + 1;
3931 data.dptr = discard_const(script);
3933 ret = ctdb_control(ctdb, destnode, 0,
3934 CTDB_CONTROL_DISABLE_SCRIPT, 0, data,
3935 NULL, NULL, &res, &timeout, NULL);
3936 if (ret != 0 || res != 0) {
3937 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for disablescript failed\n"));
3945 int ctdb_ctrl_set_ban(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, struct ctdb_ban_time *bantime)
3951 data.dsize = sizeof(*bantime);
3952 data.dptr = (uint8_t *)bantime;
3954 ret = ctdb_control(ctdb, destnode, 0,
3955 CTDB_CONTROL_SET_BAN_STATE, 0, data,
3956 NULL, NULL, &res, &timeout, NULL);
3957 if (ret != 0 || res != 0) {
3958 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for set ban state failed\n"));
3966 int ctdb_ctrl_get_ban(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, TALLOC_CTX *mem_ctx, struct ctdb_ban_time **bantime)
3971 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
3973 ret = ctdb_control(ctdb, destnode, 0,
3974 CTDB_CONTROL_GET_BAN_STATE, 0, tdb_null,
3975 tmp_ctx, &outdata, &res, &timeout, NULL);
3976 if (ret != 0 || res != 0) {
3977 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for set ban state failed\n"));
3978 talloc_free(tmp_ctx);
3982 *bantime = (struct ctdb_ban_time *)talloc_steal(mem_ctx, outdata.dptr);
3983 talloc_free(tmp_ctx);
3989 int ctdb_ctrl_set_db_priority(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, struct ctdb_db_priority *db_prio)
3994 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
3996 data.dptr = (uint8_t*)db_prio;
3997 data.dsize = sizeof(*db_prio);
3999 ret = ctdb_control(ctdb, destnode, 0,
4000 CTDB_CONTROL_SET_DB_PRIORITY, 0, data,
4001 tmp_ctx, NULL, &res, &timeout, NULL);
4002 if (ret != 0 || res != 0) {
4003 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for set_db_priority failed\n"));
4004 talloc_free(tmp_ctx);
4008 talloc_free(tmp_ctx);
4013 int ctdb_ctrl_get_db_priority(struct ctdb_context *ctdb, struct timeval timeout, uint32_t destnode, uint32_t db_id, uint32_t *priority)
4018 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
4020 data.dptr = (uint8_t*)&db_id;
4021 data.dsize = sizeof(db_id);
4023 ret = ctdb_control(ctdb, destnode, 0,
4024 CTDB_CONTROL_GET_DB_PRIORITY, 0, data,
4025 tmp_ctx, NULL, &res, &timeout, NULL);
4026 if (ret != 0 || res < 0) {
4027 DEBUG(DEBUG_ERR,(__location__ " ctdb_control for set_db_priority failed\n"));
4028 talloc_free(tmp_ctx);
4036 talloc_free(tmp_ctx);