ctdb/server/ctdbd.c

   1 /*
   2    standalone ctdb daemon
   3
   4    Copyright (C) Andrew Tridgell  2006
   5
   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.
  10
  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.
  15
  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/>.
  18 */
  19
  20 #include "replace.h"
  21 #include "system/filesys.h"
  22 #include "system/time.h"
  23 #include "system/wait.h"
  24 #include "system/network.h"
  25 #include "system/syslog.h"
  26
  27 #include <popt.h>
  28 #include <talloc.h>
  29 /* Allow use of deprecated function tevent_loop_allow_nesting() */
  30 #define TEVENT_DEPRECATED
  31 #include <tevent.h>
  32
  33 #include "lib/util/debug.h"
  34 #include "lib/util/samba_util.h"
  35
  36 #include "ctdb_private.h"
  37
  38 #include "common/reqid.h"
  39 #include "common/system.h"
  40 #include "common/common.h"
  41 #include "common/path.h"
  42 #include "common/logging.h"
  43 #include "common/logging_conf.h"
  44
  45 #include "ctdb_config.h"
  46
  47 int script_log_level;
  48 bool fast_start;
  49
  50 /*
  51   called by the transport layer when a packet comes in
  52 */
  53 static void ctdb_recv_pkt(struct ctdb_context *ctdb, uint8_t *data, uint32_t length)
  54 {
  55         struct ctdb_req_header *hdr = (struct ctdb_req_header *)data;
  56
  57         CTDB_INCREMENT_STAT(ctdb, node_packets_recv);
  58
  59         /* up the counter for this source node, so we know its alive */
  60         if (ctdb_validate_pnn(ctdb, hdr->srcnode)) {
  61                 /* as a special case, redirected calls don't increment the rx_cnt */
  62                 if (hdr->operation != CTDB_REQ_CALL ||
  63                     ((struct ctdb_req_call_old *)hdr)->hopcount == 0) {
  64                         ctdb->nodes[hdr->srcnode]->rx_cnt++;
  65                 }
  66         }
  67
  68         ctdb_input_pkt(ctdb, hdr);
  69 }
  70
  71 static const struct ctdb_upcalls ctdb_upcalls = {
  72         .recv_pkt       = ctdb_recv_pkt,
  73         .node_dead      = ctdb_node_dead,
  74         .node_connected = ctdb_node_connected
  75 };
  76
  77 static struct ctdb_context *ctdb_init(struct tevent_context *ev)
  78 {
  79         int ret;
  80         struct ctdb_context *ctdb;
  81
  82         ctdb = talloc_zero(ev, struct ctdb_context);
  83         if (ctdb == NULL) {
  84                 DBG_ERR("Memory error\n");
  85                 return NULL;
  86         }
  87         ctdb->ev  = ev;
  88
  89         /* Wrap early to exercise code. */
  90         ret = reqid_init(ctdb, INT_MAX-200, &ctdb->idr);
  91         if (ret != 0) {
  92                 D_ERR("reqid_init failed (%s)\n", strerror(ret));
  93                 talloc_free(ctdb);
  94                 return NULL;
  95         }
  96
  97         ret = srvid_init(ctdb, &ctdb->srv);
  98         if (ret != 0) {
  99                 D_ERR("srvid_init failed (%s)\n", strerror(ret));
 100                 talloc_free(ctdb);
 101                 return NULL;
 102         }
 103
 104         ctdb->daemon.name = path_socket(ctdb, "ctdbd");
 105         if (ctdb->daemon.name == NULL) {
 106                 DBG_ERR("Memory allocation error\n");
 107                 talloc_free(ctdb);
 108                 return NULL;
 109         }
 110
 111         ctdbd_pidfile = path_pidfile(ctdb, "ctdbd");
 112         if (ctdbd_pidfile == NULL) {
 113                 DBG_ERR("Memory allocation error\n");
 114                 talloc_free(ctdb);
 115                 return NULL;
 116         }
 117
 118         gettimeofday(&ctdb->ctdbd_start_time, NULL);
 119
 120         gettimeofday(&ctdb->last_recovery_started, NULL);
 121         gettimeofday(&ctdb->last_recovery_finished, NULL);
 122
 123         ctdb->recovery_mode    = CTDB_RECOVERY_NORMAL;
 124
 125         ctdb->upcalls = &ctdb_upcalls;
 126
 127         ctdb->statistics.statistics_start_time = timeval_current();
 128
 129         ctdb->capabilities = CTDB_CAP_DEFAULT;
 130
 131         /*
 132          * Initialise this node's PNN to the unknown value.  This will
 133          * be set to the correct value by either ctdb_add_node() as
 134          * part of loading the nodes file or by
 135          * ctdb_tcp_listen_automatic() when the transport is
 136          * initialised.  At some point we should de-optimise this and
 137          * pull it out into ctdb_start_daemon() so it is done clearly
 138          * and only in one place.
 139          */
 140         ctdb->pnn = CTDB_UNKNOWN_PNN;
 141
 142         ctdb->do_checkpublicip = true;
 143
 144         return ctdb;
 145 }
 146
 147
 148 /*
 149   main program
 150 */
 151 int main(int argc, const char *argv[])
 152 {
 153         struct ctdb_context *ctdb = NULL;
 154         int interactive_opt = 0;
 155         bool interactive = false;
 156
 157         struct poptOption popt_options[] = {
 158                 POPT_AUTOHELP
 159                 { "interactive", 'i', POPT_ARG_NONE, &interactive_opt, 0,
 160                   "don't fork, log to stderr", NULL },
 161                 POPT_TABLEEND
 162         };
 163         int opt, ret;
 164         const char **extra_argv;
 165         poptContext pc;
 166         struct tevent_context *ev;
 167         const char *ctdb_base;
 168         struct conf_context *conf;
 169         const char *logging_location;
 170         const char *test_mode;
 171         bool ok;
 172
 173         /*
 174          * Basic setup
 175          */
 176
 177         talloc_enable_null_tracking();
 178
 179         fault_setup();
 180
 181         ev = tevent_context_init(NULL);
 182         if (ev == NULL) {
 183                 fprintf(stderr, "tevent_context_init() failed\n");
 184                 exit(1);
 185         }
 186         tevent_loop_allow_nesting(ev);
 187
 188         ctdb = ctdb_init(ev);
 189         if (ctdb == NULL) {
 190                 fprintf(stderr, "Failed to init ctdb\n");
 191                 exit(1);
 192         }
 193
 194         /* Default value for CTDB_BASE - don't override */
 195         setenv("CTDB_BASE", CTDB_ETCDIR, 0);
 196         ctdb_base = getenv("CTDB_BASE");
 197         if (ctdb_base == NULL) {
 198                 D_ERR("CTDB_BASE not set\n");
 199                 exit(1);
 200         }
 201
 202         /*
 203          * Command-line option handling
 204          */
 205
 206         pc = poptGetContext(argv[0], argc, argv, popt_options, POPT_CONTEXT_KEEP_FIRST);
 207
 208         while ((opt = poptGetNextOpt(pc)) != -1) {
 209                 switch (opt) {
 210                 default:
 211                         fprintf(stderr, "Invalid option %s: %s\n",
 212                                 poptBadOption(pc, 0), poptStrerror(opt));
 213                         goto fail;
 214                 }
 215         }
 216
 217         /* If there are extra arguments then exit with usage message */
 218         extra_argv = poptGetArgs(pc);
 219         if (extra_argv) {
 220                 extra_argv++;
 221                 if (extra_argv[0])  {
 222                         poptPrintHelp(pc, stdout, 0);
 223                         goto fail;
 224                 }
 225         }
 226
 227         interactive = (interactive_opt != 0);
 228
 229         /*
 230          * Configuration file handling
 231          */
 232
 233         ret = ctdbd_config_load(ctdb, &conf);
 234         if (ret != 0) {
 235                 /* ctdbd_config_load() logs the failure */
 236                 goto fail;
 237         }
 238
 239         /*
 240          * Logging setup/options
 241          */
 242
 243         test_mode = getenv("CTDB_TEST_MODE");
 244
 245         /* Log to stderr (ignoring configuration) when running as interactive */
 246         if (interactive) {
 247                 logging_location = "file:";
 248                 setenv("CTDB_INTERACTIVE", "true", 1);
 249         } else {
 250                 logging_location = logging_conf_location(conf);
 251         }
 252
 253         if (strcmp(logging_location, "syslog") != 0 && test_mode == NULL) {
 254                 /* This can help when CTDB logging is misconfigured */
 255                 syslog(LOG_DAEMON|LOG_NOTICE,
 256                        "CTDB logging to location %s",
 257                        logging_location);
 258         }
 259
 260         /* Initialize logging and set the debug level */
 261         ok = ctdb_logging_init(ctdb,
 262                                logging_location,
 263                                logging_conf_log_level(conf));
 264         if (!ok) {
 265                 goto fail;
 266         }
 267         setenv("CTDB_LOGGING", logging_location, 1);
 268         setenv("CTDB_DEBUGLEVEL", debug_level_to_string(DEBUGLEVEL), 1);
 269
 270         script_log_level = debug_level_from_string(
 271                                         ctdb_config.script_log_level);
 272
 273         D_NOTICE("CTDB starting on node\n");
 274
 275         /*
 276          * Cluster setup/options
 277          */
 278
 279         ret = ctdb_set_transport(ctdb, ctdb_config.transport);
 280         if (ret == -1) {
 281                 D_ERR("ctdb_set_transport failed - %s\n", ctdb_errstr(ctdb));
 282                 goto fail;
 283         }
 284
 285         if (ctdb_config.cluster_lock != NULL) {
 286                 ctdb->recovery_lock = ctdb_config.cluster_lock;
 287         } else if (ctdb_config.recovery_lock != NULL) {
 288                 ctdb->recovery_lock = ctdb_config.recovery_lock;
 289         } else {
 290                 D_WARNING("Cluster lock not set\n");
 291         }
 292
 293         /* tell ctdb what address to listen on */
 294         if (ctdb_config.node_address) {
 295                 ret = ctdb_set_address(ctdb, ctdb_config.node_address);
 296                 if (ret == -1) {
 297                         D_ERR("ctdb_set_address failed - %s\n",
 298                               ctdb_errstr(ctdb));
 299                         goto fail;
 300                 }
 301         }
 302
 303         /* tell ctdb what nodes are available */
 304         ctdb->nodes_file = talloc_asprintf(ctdb, "%s/nodes", ctdb_base);
 305         if (ctdb->nodes_file == NULL) {
 306                 DBG_ERR(" Out of memory\n");
 307                 goto fail;
 308         }
 309         ctdb_load_nodes_file(ctdb);
 310
 311         /*
 312          * Database setup/options
 313          */
 314
 315         ctdb->db_directory = ctdb_config.dbdir_volatile;
 316         ok = directory_exist(ctdb->db_directory);
 317         if (! ok) {
 318                 D_ERR("Volatile database directory %s does not exist\n",
 319                       ctdb->db_directory);
 320                 goto fail;
 321         }
 322
 323         ctdb->db_directory_persistent = ctdb_config.dbdir_persistent;
 324         ok = directory_exist(ctdb->db_directory_persistent);
 325         if (! ok) {
 326                 D_ERR("Persistent database directory %s does not exist\n",
 327                       ctdb->db_directory_persistent);
 328                 goto fail;
 329         }
 330
 331         ctdb->db_directory_state = ctdb_config.dbdir_state;
 332         ok = directory_exist(ctdb->db_directory_state);
 333         if (! ok) {
 334                 D_ERR("State database directory %s does not exist\n",
 335                       ctdb->db_directory_state);
 336                 goto fail;
 337         }
 338
 339         if (ctdb_config.lock_debug_script != NULL) {
 340                 ret = setenv("CTDB_DEBUG_LOCKS",
 341                              ctdb_config.lock_debug_script,
 342                              1);
 343                 if (ret != 0) {
 344                         D_ERR("Failed to set up lock debugging (%s)\n",
 345                               strerror(errno));
 346                         goto fail;
 347                 }
 348         }
 349
 350         /*
 351          * Legacy setup/options
 352          */
 353
 354         ctdb->start_as_disabled = (int)ctdb_config.start_as_disabled;
 355         ctdb->start_as_stopped  = (int)ctdb_config.start_as_stopped;
 356
 357         /* set ctdbd capabilities */
 358         if (!ctdb_config.lmaster_capability) {
 359                 ctdb->capabilities &= ~CTDB_CAP_LMASTER;
 360         }
 361         if (!ctdb_config.leader_capability) {
 362                 ctdb->capabilities &= ~CTDB_CAP_RECMASTER;
 363         }
 364
 365         ctdb->do_setsched = ctdb_config.realtime_scheduling;
 366
 367         /*
 368          * Miscellaneous setup
 369          */
 370
 371         ctdb_tunables_load(ctdb);
 372
 373         ctdb->event_script_dir = talloc_asprintf(ctdb,
 374                                                  "%s/events/legacy",
 375                                                  ctdb_base);
 376         if (ctdb->event_script_dir == NULL) {
 377                 DBG_ERR("Out of memory\n");
 378                 goto fail;
 379         }
 380
 381         ctdb->notification_script = talloc_asprintf(ctdb,
 382                                                     "%s/notify.sh",
 383                                                     ctdb_base);
 384         if (ctdb->notification_script == NULL) {
 385                 D_ERR("Unable to set notification script\n");
 386                 goto fail;
 387         }
 388
 389         /*
 390          * Testing and debug options
 391          */
 392
 393         if (test_mode != NULL) {
 394                 ctdb->do_setsched = false;
 395                 ctdb->do_checkpublicip = false;
 396                 fast_start = true;
 397         }
 398
 399         /* start the protocol running (as a child) */
 400         return ctdb_start_daemon(ctdb, interactive, test_mode != NULL);
 401
 402 fail:
 403         talloc_free(ctdb);
 404         exit(1);
 405 }