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
  26 #include <popt.h>
  27 #include <talloc.h>
  28 /* Allow use of deprecated function tevent_loop_allow_nesting() */
  29 #define TEVENT_DEPRECATED
  30 #include <tevent.h>
  31
  32 #include "lib/util/debug.h"
  33 #include "lib/util/samba_util.h"
  34
  35 #include "ctdb_private.h"
  36
  37 #include "common/reqid.h"
  38 #include "common/system.h"
  39 #include "common/common.h"
  40 #include "common/logging.h"
  41
  42 static struct {
  43         const char *socket;
  44         const char *debuglevel;
  45         const char *nlist;
  46         const char *transport;
  47         const char *myaddress;
  48         const char *public_address_list;
  49         const char *event_script_dir;
  50         const char *notification_script;
  51         const char *logging;
  52         const char *recovery_lock;
  53         const char *db_dir;
  54         const char *db_dir_persistent;
  55         const char *db_dir_state;
  56         const char *public_interface;
  57         int         valgrinding;
  58         int         nosetsched;
  59         int         start_as_disabled;
  60         int         start_as_stopped;
  61         int         no_lmaster;
  62         int         no_recmaster;
  63         int         script_log_level;
  64         int         no_publicipcheck;
  65         int         max_persistent_check_errors;
  66         int         torture;
  67 } options = {
  68         .socket = CTDB_RUNDIR "/ctdbd.socket",
  69         .debuglevel = "NOTICE",
  70         .nlist = NULL,
  71         .public_address_list = NULL,
  72         .transport = "tcp",
  73         .event_script_dir = NULL,
  74         .logging = "file:" LOGDIR "/log.ctdb",
  75         .db_dir = CTDB_VARDIR,
  76         .db_dir_persistent = CTDB_VARDIR "/persistent",
  77         .db_dir_state = CTDB_VARDIR "/state",
  78         .script_log_level = DEBUG_ERR,
  79 };
  80
  81 int script_log_level;
  82 bool fast_start;
  83
  84 /*
  85   called by the transport layer when a packet comes in
  86 */
  87 static void ctdb_recv_pkt(struct ctdb_context *ctdb, uint8_t *data, uint32_t length)
  88 {
  89         struct ctdb_req_header *hdr = (struct ctdb_req_header *)data;
  90
  91         CTDB_INCREMENT_STAT(ctdb, node_packets_recv);
  92
  93         /* up the counter for this source node, so we know its alive */
  94         if (ctdb_validate_pnn(ctdb, hdr->srcnode)) {
  95                 /* as a special case, redirected calls don't increment the rx_cnt */
  96                 if (hdr->operation != CTDB_REQ_CALL ||
  97                     ((struct ctdb_req_call_old *)hdr)->hopcount == 0) {
  98                         ctdb->nodes[hdr->srcnode]->rx_cnt++;
  99                 }
 100         }
 101
 102         ctdb_input_pkt(ctdb, hdr);
 103 }
 104
 105 static const struct ctdb_upcalls ctdb_upcalls = {
 106         .recv_pkt       = ctdb_recv_pkt,
 107         .node_dead      = ctdb_node_dead,
 108         .node_connected = ctdb_node_connected
 109 };
 110
 111
 112
 113 /*
 114   main program
 115 */
 116 int main(int argc, const char *argv[])
 117 {
 118         struct ctdb_context *ctdb;
 119         int interactive = 0;
 120
 121         struct poptOption popt_options[] = {
 122                 POPT_AUTOHELP
 123                 { "socket", 0, POPT_ARG_STRING, &options.socket, 0, "local socket name", "filename" },
 124                 { "debug", 'd', POPT_ARG_STRING, &options.debuglevel, 0, "debug level", NULL },
 125                 { "interactive", 'i', POPT_ARG_NONE, &interactive, 0, "don't fork", NULL },
 126                 { "public-addresses", 0, POPT_ARG_STRING, &options.public_address_list, 0, "public address list file", "filename" },
 127                 { "public-interface", 0, POPT_ARG_STRING, &options.public_interface, 0, "public interface", "interface"},
 128                 { "event-script-dir", 0, POPT_ARG_STRING, &options.event_script_dir, 0, "event script directory", "dirname" },
 129                 { "logging", 0, POPT_ARG_STRING, &options.logging, 0, "logging method to be used", NULL },
 130                 { "nlist", 0, POPT_ARG_STRING, &options.nlist, 0, "node list file", "filename" },
 131                 { "notification-script", 0, POPT_ARG_STRING, &options.notification_script, 0, "notification script", "filename" },
 132                 { "listen", 0, POPT_ARG_STRING, &options.myaddress, 0, "address to listen on", "address" },
 133                 { "transport", 0, POPT_ARG_STRING, &options.transport, 0, "protocol transport", NULL },
 134                 { "dbdir", 0, POPT_ARG_STRING, &options.db_dir, 0, "directory for the tdb files", NULL },
 135                 { "dbdir-persistent", 0, POPT_ARG_STRING, &options.db_dir_persistent, 0, "directory for persistent tdb files", NULL },
 136                 { "dbdir-state", 0, POPT_ARG_STRING, &options.db_dir_state, 0, "directory for internal state tdb files", NULL },
 137                 { "reclock", 0, POPT_ARG_STRING, &options.recovery_lock, 0, "recovery lock", "lock" },
 138                 { "pidfile", 0, POPT_ARG_STRING, &ctdbd_pidfile, 0, "location of PID file", "filename" },
 139                 { "valgrinding", 0, POPT_ARG_NONE, &options.valgrinding, 0, "disable setscheduler SCHED_FIFO call, use mmap for tdbs", NULL },
 140                 { "nosetsched", 0, POPT_ARG_NONE, &options.nosetsched, 0, "disable setscheduler SCHED_FIFO call, use mmap for tdbs", NULL },
 141                 { "start-as-disabled", 0, POPT_ARG_NONE, &options.start_as_disabled, 0, "Node starts in disabled state", NULL },
 142                 { "start-as-stopped", 0, POPT_ARG_NONE, &options.start_as_stopped, 0, "Node starts in stopped state", NULL },
 143                 { "no-lmaster", 0, POPT_ARG_NONE, &options.no_lmaster, 0, "disable lmaster role on this node", NULL },
 144                 { "no-recmaster", 0, POPT_ARG_NONE, &options.no_recmaster, 0, "disable recmaster role on this node", NULL },
 145                 { "script-log-level", 0, POPT_ARG_INT, &options.script_log_level, 0, "log level of event script output", NULL },
 146                 { "nopublicipcheck", 0, POPT_ARG_NONE, &options.no_publicipcheck, 0, "don't check we have/don't have the correct public ip addresses", NULL },
 147                 { "max-persistent-check-errors", 0, POPT_ARG_INT,
 148                   &options.max_persistent_check_errors, 0,
 149                   "max allowed persistent check errors (default 0)", NULL },
 150                 { "sloppy-start", 0, POPT_ARG_NONE, &fast_start, 0, "Do not perform full recovery on start", NULL },
 151                 { "torture", 0, POPT_ARG_NONE, &options.torture, 0, "enable nastiness in library", NULL },
 152                 POPT_TABLEEND
 153         };
 154         int opt, ret;
 155         const char **extra_argv;
 156         poptContext pc;
 157         struct tevent_context *ev;
 158
 159         pc = poptGetContext(argv[0], argc, argv, popt_options, POPT_CONTEXT_KEEP_FIRST);
 160
 161         while ((opt = poptGetNextOpt(pc)) != -1) {
 162                 switch (opt) {
 163                 default:
 164                         fprintf(stderr, "Invalid option %s: %s\n",
 165                                 poptBadOption(pc, 0), poptStrerror(opt));
 166                         exit(1);
 167                 }
 168         }
 169
 170         /* If there are extra arguments then exit with usage message */
 171         extra_argv = poptGetArgs(pc);
 172         if (extra_argv) {
 173                 extra_argv++;
 174                 if (extra_argv[0])  {
 175                         poptPrintHelp(pc, stdout, 0);
 176                         exit(1);
 177                 }
 178         }
 179
 180         talloc_enable_null_tracking();
 181
 182         fault_setup();
 183
 184         ev = tevent_context_init(NULL);
 185         if (ev == NULL) {
 186                 fprintf(stderr, "tevent_context_init() failed\n");
 187                 exit(1);
 188         }
 189         tevent_loop_allow_nesting(ev);
 190
 191         ctdb = ctdb_init(ev);
 192         if (ctdb == NULL) {
 193                 fprintf(stderr, "Failed to init ctdb\n");
 194                 exit(1);
 195         }
 196
 197         if (options.torture == 1) {
 198                 ctdb_set_flags(ctdb, CTDB_FLAG_TORTURE);
 199         }
 200
 201         /* Log to stderr when running as interactive */
 202         if (interactive) {
 203                 options.logging = "file:";
 204         }
 205
 206         /* Initialize logging and set the debug level */
 207         if (!ctdb_logging_init(ctdb, options.logging, options.debuglevel)) {
 208                 exit(1);
 209         }
 210         setenv("CTDB_LOGGING", options.logging, 1);
 211         setenv("CTDB_DEBUGLEVEL", debug_level_to_string(DEBUGLEVEL), 1);
 212
 213         setenv("CTDB_SOCKET", options.socket, 1);
 214         ret = ctdb_set_socketname(ctdb, options.socket);
 215         if (ret == -1) {
 216                 DEBUG(DEBUG_ERR, ("ctdb_set_socketname() failed\n"));
 217                 exit(1);
 218         }
 219
 220         ctdb->start_as_disabled = options.start_as_disabled;
 221         ctdb->start_as_stopped  = options.start_as_stopped;
 222
 223         script_log_level = options.script_log_level;
 224
 225         DEBUG(DEBUG_NOTICE,("CTDB starting on node\n"));
 226
 227         gettimeofday(&ctdb->ctdbd_start_time, NULL);
 228         gettimeofday(&ctdb->last_recovery_started, NULL);
 229         gettimeofday(&ctdb->last_recovery_finished, NULL);
 230         ctdb->recovery_mode    = CTDB_RECOVERY_NORMAL;
 231         ctdb->recovery_master  = (uint32_t)-1;
 232         ctdb->upcalls          = &ctdb_upcalls;
 233
 234         if (options.recovery_lock == NULL) {
 235                 DEBUG(DEBUG_WARNING, ("Recovery lock not set\n"));
 236         }
 237         ctdb->recovery_lock = options.recovery_lock;
 238
 239         TALLOC_FREE(ctdb->idr);
 240         ret = reqid_init(ctdb, 0, &ctdb->idr);;
 241         if (ret != 0) {
 242                 DEBUG(DEBUG_ERR, ("reqid_init failed (%s)\n", strerror(ret)));
 243                 exit(1);
 244         }
 245
 246         ctdb_tunables_set_defaults(ctdb);
 247
 248         ret = ctdb_set_transport(ctdb, options.transport);
 249         if (ret == -1) {
 250                 DEBUG(DEBUG_ERR,("ctdb_set_transport failed - %s\n",
 251                                  ctdb_errstr(ctdb)));
 252                 exit(1);
 253         }
 254
 255         /* tell ctdb what address to listen on */
 256         if (options.myaddress) {
 257                 ret = ctdb_set_address(ctdb, options.myaddress);
 258                 if (ret == -1) {
 259                         DEBUG(DEBUG_ERR,("ctdb_set_address failed - %s\n",
 260                                          ctdb_errstr(ctdb)));
 261                         exit(1);
 262                 }
 263         }
 264
 265         /* set ctdbd capabilities */
 266         ctdb->capabilities = CTDB_CAP_DEFAULT;
 267         if (options.no_lmaster != 0) {
 268                 ctdb->capabilities &= ~CTDB_CAP_LMASTER;
 269         }
 270         if (options.no_recmaster != 0) {
 271                 ctdb->capabilities &= ~CTDB_CAP_RECMASTER;
 272         }
 273
 274         /* Initialise this node's PNN to the unknown value.  This will
 275          * be set to the correct value by either ctdb_add_node() as
 276          * part of loading the nodes file or by
 277          * ctdb_tcp_listen_automatic() when the transport is
 278          * initialised.  At some point we should de-optimise this and
 279          * pull it out into ctdb_start_daemon() so it is done clearly
 280          * and only in one place.
 281          */
 282         ctdb->pnn = -1;
 283
 284         /* Default value for CTDB_BASE - don't override */
 285         setenv("CTDB_BASE", CTDB_ETCDIR, 0);
 286
 287         /* tell ctdb what nodes are available */
 288         if (options.nlist != NULL) {
 289                 ctdb->nodes_file = options.nlist;
 290         } else {
 291                 ctdb->nodes_file =
 292                         talloc_asprintf(ctdb, "%s/nodes", getenv("CTDB_BASE"));
 293                 if (ctdb->nodes_file == NULL) {
 294                         DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 295                         exit(1);
 296                 }
 297         }
 298         ctdb_load_nodes_file(ctdb);
 299
 300         ctdb->db_directory = options.db_dir;
 301         mkdir_p_or_die(ctdb->db_directory, 0700);
 302
 303         ctdb->db_directory_persistent = options.db_dir_persistent;
 304         mkdir_p_or_die(ctdb->db_directory_persistent, 0700);
 305
 306         ctdb->db_directory_state = options.db_dir_state;
 307         mkdir_p_or_die(ctdb->db_directory_state, 0700);
 308
 309         if (options.public_interface) {
 310                 ctdb->default_public_interface = talloc_strdup(ctdb, options.public_interface);
 311                 CTDB_NO_MEMORY(ctdb, ctdb->default_public_interface);
 312         }
 313
 314         if (options.event_script_dir != NULL) {
 315                 ctdb->event_script_dir = options.event_script_dir;
 316         } else {
 317                 ctdb->event_script_dir = talloc_asprintf(ctdb, "%s/events.d",
 318                                                          getenv("CTDB_BASE"));
 319                 if (ctdb->event_script_dir == NULL) {
 320                         DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 321                         exit(1);
 322                 }
 323         }
 324
 325         if (options.notification_script != NULL) {
 326                 ret = ctdb_set_notification_script(ctdb, options.notification_script);
 327                 if (ret == -1) {
 328                         DEBUG(DEBUG_ERR,("Unable to setup notification script\n"));
 329                         exit(1);
 330                 }
 331         }
 332
 333         ctdb->valgrinding = (options.valgrinding == 1);
 334         ctdb->do_setsched = (options.nosetsched != 1);
 335         if (ctdb->valgrinding) {
 336                 ctdb->do_setsched = false;
 337         }
 338
 339         ctdb->public_addresses_file = options.public_address_list;
 340         ctdb->do_checkpublicip = (options.no_publicipcheck == 0);
 341
 342         if (options.max_persistent_check_errors < 0) {
 343                 ctdb->max_persistent_check_errors = 0xFFFFFFFFFFFFFFFFLL;
 344         } else {
 345                 ctdb->max_persistent_check_errors = (uint64_t)options.max_persistent_check_errors;
 346         }
 347
 348         /* start the protocol running (as a child) */
 349         return ctdb_start_daemon(ctdb, interactive?false:true);
 350 }