2 ctdb main protocol code
4 Copyright (C) Andrew Tridgell 2006
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, see <http://www.gnu.org/licenses/>.
21 #include "lib/tdb/include/tdb.h"
22 #include "lib/events/events.h"
23 #include "lib/util/dlinklist.h"
24 #include "system/network.h"
25 #include "system/filesys.h"
26 #include "../include/ctdb_private.h"
29 choose the transport we will use
31 int ctdb_set_transport(struct ctdb_context *ctdb, const char *transport)
33 ctdb->transport = talloc_strdup(ctdb, transport);
38 Check whether an ip is a valid node ip
39 Returns the node id for this ip address or -1
41 int ctdb_ip_to_nodeid(struct ctdb_context *ctdb, const char *nodeip)
45 for (nodeid=0;nodeid<ctdb->num_nodes;nodeid++) {
46 if (!strcmp(ctdb->nodes[nodeid]->address.address, nodeip)) {
55 choose the recovery lock file
57 int ctdb_set_recovery_lock_file(struct ctdb_context *ctdb, const char *file)
59 ctdb->recovery_lock_file = talloc_strdup(ctdb, file);
64 set the directory for the local databases
66 int ctdb_set_tdb_dir(struct ctdb_context *ctdb, const char *dir)
68 ctdb->db_directory = talloc_strdup(ctdb, dir);
69 if (ctdb->db_directory == NULL) {
76 set the directory for the persistent databases
78 int ctdb_set_tdb_dir_persistent(struct ctdb_context *ctdb, const char *dir)
80 ctdb->db_directory_persistent = talloc_strdup(ctdb, dir);
81 if (ctdb->db_directory_persistent == NULL) {
88 add a node to the list of active nodes
90 static int ctdb_add_node(struct ctdb_context *ctdb, char *nstr)
92 struct ctdb_node *node, **nodep;
94 nodep = talloc_realloc(ctdb, ctdb->nodes, struct ctdb_node *, ctdb->num_nodes+1);
95 CTDB_NO_MEMORY(ctdb, nodep);
98 nodep = &ctdb->nodes[ctdb->num_nodes];
99 (*nodep) = talloc_zero(ctdb->nodes, struct ctdb_node);
100 CTDB_NO_MEMORY(ctdb, *nodep);
103 if (ctdb_parse_address(ctdb, node, nstr, &node->address) != 0) {
107 node->name = talloc_asprintf(node, "%s:%u",
108 node->address.address,
110 /* this assumes that the nodes are kept in sorted order, and no gaps */
111 node->pnn = ctdb->num_nodes;
113 /* nodes start out disconnected and unhealthy */
114 node->flags = (NODE_FLAGS_DISCONNECTED | NODE_FLAGS_UNHEALTHY);
116 if (ctdb->address.address &&
117 ctdb_same_address(&ctdb->address, &node->address)) {
118 ctdb->pnn = node->pnn;
119 node->flags &= ~NODE_FLAGS_DISCONNECTED;
123 node->dead_count = 0;
129 setup the node list from a file
131 int ctdb_set_nlist(struct ctdb_context *ctdb, const char *nlist)
137 talloc_free(ctdb->node_list_file);
138 ctdb->node_list_file = talloc_strdup(ctdb, nlist);
140 lines = file_lines_load(nlist, &nlines, ctdb);
142 ctdb_set_error(ctdb, "Failed to load nlist '%s'\n", nlist);
145 while (nlines > 0 && strcmp(lines[nlines-1], "") == 0) {
149 for (i=0;i<nlines;i++) {
150 if (ctdb_add_node(ctdb, lines[i]) != 0) {
156 /* initialize the vnn mapping table now that we have num_nodes setup */
157 ctdb->vnn_map = talloc(ctdb, struct ctdb_vnn_map);
158 CTDB_NO_MEMORY(ctdb, ctdb->vnn_map);
160 ctdb->vnn_map->generation = INVALID_GENERATION;
161 ctdb->vnn_map->size = ctdb->num_nodes;
162 ctdb->vnn_map->map = talloc_array(ctdb->vnn_map, uint32_t, ctdb->vnn_map->size);
163 CTDB_NO_MEMORY(ctdb, ctdb->vnn_map->map);
165 for(i=0;i<ctdb->vnn_map->size;i++) {
166 ctdb->vnn_map->map[i] = i;
175 setup the local node address
177 int ctdb_set_address(struct ctdb_context *ctdb, const char *address)
179 if (ctdb_parse_address(ctdb, ctdb, address, &ctdb->address) != 0) {
183 ctdb->name = talloc_asprintf(ctdb, "%s:%u",
184 ctdb->address.address,
191 return the number of active nodes
193 uint32_t ctdb_get_num_active_nodes(struct ctdb_context *ctdb)
197 for (i=0;i<ctdb->vnn_map->size;i++) {
198 struct ctdb_node *node = ctdb->nodes[ctdb->vnn_map->map[i]];
199 if (!(node->flags & NODE_FLAGS_INACTIVE)) {
208 called when we need to process a packet. This can be a requeued packet
209 after a lockwait, or a real packet from another node
211 void ctdb_input_pkt(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
215 /* place the packet as a child of the tmp_ctx. We then use
216 talloc_free() below to free it. If any of the calls want
217 to keep it, then they will steal it somewhere else, and the
218 talloc_free() will only free the tmp_ctx */
219 tmp_ctx = talloc_new(ctdb);
220 talloc_steal(tmp_ctx, hdr);
222 DEBUG(DEBUG_DEBUG,(__location__ " ctdb request %u of type %u length %u from "
223 "node %u to %u\n", hdr->reqid, hdr->operation, hdr->length,
224 hdr->srcnode, hdr->destnode));
226 switch (hdr->operation) {
228 case CTDB_REPLY_CALL:
229 case CTDB_REQ_DMASTER:
230 case CTDB_REPLY_DMASTER:
231 /* we dont allow these calls when banned */
232 if (ctdb->nodes[ctdb->pnn]->flags & NODE_FLAGS_BANNED) {
233 DEBUG(DEBUG_DEBUG,(__location__ " ctdb operation %u"
235 " length %u from node %u to %u while node"
237 hdr->operation, hdr->reqid,
239 hdr->srcnode, hdr->destnode));
243 /* for ctdb_call inter-node operations verify that the
244 remote node that sent us the call is running in the
245 same generation instance as this node
247 if (ctdb->vnn_map->generation != hdr->generation) {
248 DEBUG(DEBUG_DEBUG,(__location__ " ctdb operation %u"
250 " length %u from node %u to %u had an"
251 " invalid generation id:%u while our"
252 " generation id is:%u\n",
253 hdr->operation, hdr->reqid,
255 hdr->srcnode, hdr->destnode,
256 hdr->generation, ctdb->vnn_map->generation));
261 switch (hdr->operation) {
263 ctdb->statistics.node.req_call++;
264 ctdb_request_call(ctdb, hdr);
267 case CTDB_REPLY_CALL:
268 ctdb->statistics.node.reply_call++;
269 ctdb_reply_call(ctdb, hdr);
272 case CTDB_REPLY_ERROR:
273 ctdb->statistics.node.reply_error++;
274 ctdb_reply_error(ctdb, hdr);
277 case CTDB_REQ_DMASTER:
278 ctdb->statistics.node.req_dmaster++;
279 ctdb_request_dmaster(ctdb, hdr);
282 case CTDB_REPLY_DMASTER:
283 ctdb->statistics.node.reply_dmaster++;
284 ctdb_reply_dmaster(ctdb, hdr);
287 case CTDB_REQ_MESSAGE:
288 ctdb->statistics.node.req_message++;
289 ctdb_request_message(ctdb, hdr);
292 case CTDB_REQ_CONTROL:
293 ctdb->statistics.node.req_control++;
294 ctdb_request_control(ctdb, hdr);
297 case CTDB_REPLY_CONTROL:
298 ctdb->statistics.node.reply_control++;
299 ctdb_reply_control(ctdb, hdr);
302 case CTDB_REQ_KEEPALIVE:
303 ctdb->statistics.keepalive_packets_recv++;
307 DEBUG(DEBUG_CRIT,("%s: Packet with unknown operation %u\n",
308 __location__, hdr->operation));
313 talloc_free(tmp_ctx);
318 called by the transport layer when a node is dead
320 void ctdb_node_dead(struct ctdb_node *node)
322 if (node->flags & NODE_FLAGS_DISCONNECTED) {
323 DEBUG(DEBUG_INFO,("%s: node %s is already marked disconnected: %u connected\n",
324 node->ctdb->name, node->name,
325 node->ctdb->num_connected));
328 node->ctdb->num_connected--;
329 node->flags |= NODE_FLAGS_DISCONNECTED | NODE_FLAGS_UNHEALTHY;
331 node->dead_count = 0;
333 DEBUG(DEBUG_NOTICE,("%s: node %s is dead: %u connected\n",
334 node->ctdb->name, node->name, node->ctdb->num_connected));
335 ctdb_daemon_cancel_controls(node->ctdb, node);
337 node->ctdb->methods->restart(node);
341 called by the transport layer when a node is connected
343 void ctdb_node_connected(struct ctdb_node *node)
345 if (!(node->flags & NODE_FLAGS_DISCONNECTED)) {
346 DEBUG(DEBUG_INFO,("%s: node %s is already marked connected: %u connected\n",
347 node->ctdb->name, node->name,
348 node->ctdb->num_connected));
351 node->ctdb->num_connected++;
352 node->dead_count = 0;
353 node->flags &= ~NODE_FLAGS_DISCONNECTED;
354 node->flags |= NODE_FLAGS_UNHEALTHY;
355 DEBUG(DEBUG_INFO,("%s: connected to %s - %u connected\n",
356 node->ctdb->name, node->name, node->ctdb->num_connected));
360 struct ctdb_context *ctdb;
361 struct ctdb_req_header *hdr;
366 trigered when a deferred packet is due
368 static void queue_next_trigger(struct event_context *ev, struct timed_event *te,
369 struct timeval t, void *private_data)
371 struct queue_next *q = talloc_get_type(private_data, struct queue_next);
372 ctdb_input_pkt(q->ctdb, q->hdr);
377 defer a packet, so it is processed on the next event loop
378 this is used for sending packets to ourselves
380 static void ctdb_defer_packet(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
382 struct queue_next *q;
383 q = talloc(ctdb, struct queue_next);
385 DEBUG(DEBUG_ERR,(__location__ " Failed to allocate deferred packet\n"));
389 q->hdr = talloc_memdup(ctdb, hdr, hdr->length);
390 if (q->hdr == NULL) {
391 DEBUG(DEBUG_ERR,("Error copying deferred packet to self\n"));
395 /* use this to put packets directly into our recv function */
396 ctdb_input_pkt(q->ctdb, q->hdr);
398 event_add_timed(ctdb->ev, q, timeval_zero(), queue_next_trigger, q);
404 broadcast a packet to all nodes
406 static void ctdb_broadcast_packet_all(struct ctdb_context *ctdb,
407 struct ctdb_req_header *hdr)
410 for (i=0;i<ctdb->num_nodes;i++) {
411 hdr->destnode = ctdb->nodes[i]->pnn;
412 ctdb_queue_packet(ctdb, hdr);
417 broadcast a packet to all nodes in the current vnnmap
419 static void ctdb_broadcast_packet_vnnmap(struct ctdb_context *ctdb,
420 struct ctdb_req_header *hdr)
423 for (i=0;i<ctdb->vnn_map->size;i++) {
424 hdr->destnode = ctdb->vnn_map->map[i];
425 ctdb_queue_packet(ctdb, hdr);
430 broadcast a packet to all connected nodes
432 static void ctdb_broadcast_packet_connected(struct ctdb_context *ctdb,
433 struct ctdb_req_header *hdr)
436 for (i=0;i<ctdb->num_nodes;i++) {
437 if (!(ctdb->nodes[i]->flags & NODE_FLAGS_DISCONNECTED)) {
438 hdr->destnode = ctdb->nodes[i]->pnn;
439 ctdb_queue_packet(ctdb, hdr);
445 queue a packet or die
447 void ctdb_queue_packet(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
449 struct ctdb_node *node;
451 switch (hdr->destnode) {
452 case CTDB_BROADCAST_ALL:
453 ctdb_broadcast_packet_all(ctdb, hdr);
455 case CTDB_BROADCAST_VNNMAP:
456 ctdb_broadcast_packet_vnnmap(ctdb, hdr);
458 case CTDB_BROADCAST_CONNECTED:
459 ctdb_broadcast_packet_connected(ctdb, hdr);
463 ctdb->statistics.node_packets_sent++;
465 if (!ctdb_validate_pnn(ctdb, hdr->destnode)) {
466 DEBUG(DEBUG_CRIT,(__location__ " cant send to node %u that does not exist\n",
471 node = ctdb->nodes[hdr->destnode];
473 if (hdr->destnode == ctdb->pnn) {
474 ctdb_defer_packet(ctdb, hdr);
477 if (ctdb->methods->queue_pkt(node, (uint8_t *)hdr, hdr->length) != 0) {
478 ctdb_fatal(ctdb, "Unable to queue packet\n");