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