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 "includes.h"
  21 #include "system/filesys.h"
  22 #include "popt.h"
  23 #include "system/time.h"
  24 #include "system/wait.h"
  25 #include "system/network.h"
  26 #include "cmdline.h"
  27 #include "../include/ctdb_private.h"
  28
  29 static struct {
  30         const char *nlist;
  31         const char *transport;
  32         const char *myaddress;
  33         const char *public_address_list;
  34         const char *event_script_dir;
  35         const char *notification_script;
  36         const char *logfile;
  37         const char *recovery_lock_file;
  38         const char *db_dir;
  39         const char *db_dir_persistent;
  40         const char *db_dir_state;
  41         const char *public_interface;
  42         const char *single_public_ip;
  43         int         valgrinding;
  44         int         nosetsched;
  45         int         use_syslog;
  46         int         start_as_disabled;
  47         int         start_as_stopped;
  48         int         no_lmaster;
  49         int         no_recmaster;
  50         int         lvs;
  51         int         script_log_level;
  52         int         no_publicipcheck;
  53         int         max_persistent_check_errors;
  54 } options = {
  55         .nlist = NULL,
  56         .public_address_list = NULL,
  57         .transport = "tcp",
  58         .event_script_dir = NULL,
  59         .logfile = LOGDIR "/log.ctdb",
  60         .db_dir = CTDB_VARDIR,
  61         .db_dir_persistent = CTDB_VARDIR "/persistent",
  62         .db_dir_state = CTDB_VARDIR "/state",
  63         .script_log_level = DEBUG_ERR,
  64 };
  65
  66 int script_log_level;
  67 bool fast_start;
  68
  69 /*
  70   called by the transport layer when a packet comes in
  71 */
  72 static void ctdb_recv_pkt(struct ctdb_context *ctdb, uint8_t *data, uint32_t length)
  73 {
  74         struct ctdb_req_header *hdr = (struct ctdb_req_header *)data;
  75
  76         CTDB_INCREMENT_STAT(ctdb, node_packets_recv);
  77
  78         /* up the counter for this source node, so we know its alive */
  79         if (ctdb_validate_pnn(ctdb, hdr->srcnode)) {
  80                 /* as a special case, redirected calls don't increment the rx_cnt */
  81                 if (hdr->operation != CTDB_REQ_CALL ||
  82                     ((struct ctdb_req_call *)hdr)->hopcount == 0) {
  83                         ctdb->nodes[hdr->srcnode]->rx_cnt++;
  84                 }
  85         }
  86
  87         ctdb_input_pkt(ctdb, hdr);
  88 }
  89
  90 static const struct ctdb_upcalls ctdb_upcalls = {
  91         .recv_pkt       = ctdb_recv_pkt,
  92         .node_dead      = ctdb_node_dead,
  93         .node_connected = ctdb_node_connected
  94 };
  95
  96
  97
  98 /*
  99   main program
 100 */
 101 int main(int argc, const char *argv[])
 102 {
 103         struct ctdb_context *ctdb;
 104         int interactive = 0;
 105
 106         struct poptOption popt_options[] = {
 107                 POPT_AUTOHELP
 108                 POPT_CTDB_CMDLINE
 109                 { "interactive", 'i', POPT_ARG_NONE, &interactive, 0, "don't fork", NULL },
 110                 { "public-addresses", 0, POPT_ARG_STRING, &options.public_address_list, 0, "public address list file", "filename" },
 111                 { "public-interface", 0, POPT_ARG_STRING, &options.public_interface, 0, "public interface", "interface"},
 112                 { "single-public-ip", 0, POPT_ARG_STRING, &options.single_public_ip, 0, "single public ip", "ip-address"},
 113                 { "event-script-dir", 0, POPT_ARG_STRING, &options.event_script_dir, 0, "event script directory", "dirname" },
 114                 { "logfile", 0, POPT_ARG_STRING, &options.logfile, 0, "log file location", "filename" },
 115                 { "nlist", 0, POPT_ARG_STRING, &options.nlist, 0, "node list file", "filename" },
 116                 { "notification-script", 0, POPT_ARG_STRING, &options.notification_script, 0, "notification script", "filename" },
 117                 { "listen", 0, POPT_ARG_STRING, &options.myaddress, 0, "address to listen on", "address" },
 118                 { "transport", 0, POPT_ARG_STRING, &options.transport, 0, "protocol transport", NULL },
 119                 { "dbdir", 0, POPT_ARG_STRING, &options.db_dir, 0, "directory for the tdb files", NULL },
 120                 { "dbdir-persistent", 0, POPT_ARG_STRING, &options.db_dir_persistent, 0, "directory for persistent tdb files", NULL },
 121                 { "dbdir-state", 0, POPT_ARG_STRING, &options.db_dir_state, 0, "directory for internal state tdb files", NULL },
 122                 { "reclock", 0, POPT_ARG_STRING, &options.recovery_lock_file, 0, "location of recovery lock file", "filename" },
 123                 { "pidfile", 0, POPT_ARG_STRING, &ctdbd_pidfile, 0, "location of PID file", "filename" },
 124                 { "valgrinding", 0, POPT_ARG_NONE, &options.valgrinding, 0, "disable setscheduler SCHED_FIFO call, use mmap for tdbs", NULL },
 125                 { "nosetsched", 0, POPT_ARG_NONE, &options.nosetsched, 0, "disable setscheduler SCHED_FIFO call, use mmap for tdbs", NULL },
 126                 { "syslog", 0, POPT_ARG_NONE, &options.use_syslog, 0, "log messages to syslog", NULL },
 127                 { "start-as-disabled", 0, POPT_ARG_NONE, &options.start_as_disabled, 0, "Node starts in disabled state", NULL },
 128                 { "start-as-stopped", 0, POPT_ARG_NONE, &options.start_as_stopped, 0, "Node starts in stopped state", NULL },
 129                 { "no-lmaster", 0, POPT_ARG_NONE, &options.no_lmaster, 0, "disable lmaster role on this node", NULL },
 130                 { "no-recmaster", 0, POPT_ARG_NONE, &options.no_recmaster, 0, "disable recmaster role on this node", NULL },
 131                 { "lvs", 0, POPT_ARG_NONE, &options.lvs, 0, "lvs is enabled on this node", NULL },
 132                 { "script-log-level", 0, POPT_ARG_INT, &options.script_log_level, 0, "log level of event script output", NULL },
 133                 { "nopublicipcheck", 0, POPT_ARG_NONE, &options.no_publicipcheck, 0, "don't check we have/don't have the correct public ip addresses", NULL },
 134                 { "max-persistent-check-errors", 0, POPT_ARG_INT,
 135                   &options.max_persistent_check_errors, 0,
 136                   "max allowed persistent check errors (default 0)", NULL },
 137                 { "log-ringbuf-size", 0, POPT_ARG_INT, &log_ringbuf_size, 0, "Number of log messages we can store in the memory ringbuffer", NULL },
 138                 { "sloppy-start", 0, POPT_ARG_NONE, &fast_start, 0, "Do not perform full recovery on start", NULL },
 139                 POPT_TABLEEND
 140         };
 141         int opt, ret;
 142         const char **extra_argv;
 143         int extra_argc = 0;
 144         poptContext pc;
 145         struct event_context *ev;
 146
 147         pc = poptGetContext(argv[0], argc, argv, popt_options, POPT_CONTEXT_KEEP_FIRST);
 148
 149         while ((opt = poptGetNextOpt(pc)) != -1) {
 150                 switch (opt) {
 151                 default:
 152                         fprintf(stderr, "Invalid option %s: %s\n",
 153                                 poptBadOption(pc, 0), poptStrerror(opt));
 154                         exit(1);
 155                 }
 156         }
 157
 158         /* setup the remaining options for the main program to use */
 159         extra_argv = poptGetArgs(pc);
 160         if (extra_argv) {
 161                 extra_argv++;
 162                 while (extra_argv[extra_argc]) extra_argc++;
 163         }
 164
 165         talloc_enable_null_tracking();
 166
 167         ctdb_block_signal(SIGPIPE);
 168         fault_setup("ctdbd");
 169
 170         ev = event_context_init(NULL);
 171         tevent_loop_allow_nesting(ev);
 172
 173         ctdb = ctdb_cmdline_init(ev);
 174
 175         ctdb->start_as_disabled = options.start_as_disabled;
 176         ctdb->start_as_stopped  = options.start_as_stopped;
 177
 178         script_log_level = options.script_log_level;
 179
 180         ret = ctdb_set_logfile(ctdb, options.logfile, options.use_syslog);
 181         if (ret == -1) {
 182                 printf("ctdb_set_logfile to %s failed - %s\n",
 183                        options.use_syslog?"syslog":options.logfile, ctdb_errstr(ctdb));
 184                 exit(1);
 185         }
 186
 187         DEBUG(DEBUG_NOTICE,("CTDB starting on node\n"));
 188
 189         gettimeofday(&ctdb->ctdbd_start_time, NULL);
 190         gettimeofday(&ctdb->last_recovery_started, NULL);
 191         gettimeofday(&ctdb->last_recovery_finished, NULL);
 192         ctdb->recovery_mode    = CTDB_RECOVERY_NORMAL;
 193         ctdb->recovery_master  = (uint32_t)-1;
 194         ctdb->upcalls          = &ctdb_upcalls;
 195         ctdb->idr              = idr_init(ctdb);
 196         ctdb->recovery_lock_fd = -1;
 197
 198         ctdb_tunables_set_defaults(ctdb);
 199
 200         ret = ctdb_set_recovery_lock_file(ctdb, options.recovery_lock_file);
 201         if (ret == -1) {
 202                 DEBUG(DEBUG_ALERT,("ctdb_set_recovery_lock_file failed - %s\n", ctdb_errstr(ctdb)));
 203                 exit(1);
 204         }
 205
 206         ret = ctdb_set_transport(ctdb, options.transport);
 207         if (ret == -1) {
 208                 DEBUG(DEBUG_ALERT,("ctdb_set_transport failed - %s\n", ctdb_errstr(ctdb)));
 209                 exit(1);
 210         }
 211
 212         /* tell ctdb what address to listen on */
 213         if (options.myaddress) {
 214                 ret = ctdb_set_address(ctdb, options.myaddress);
 215                 if (ret == -1) {
 216                         DEBUG(DEBUG_ALERT,("ctdb_set_address failed - %s\n", ctdb_errstr(ctdb)));
 217                         exit(1);
 218                 }
 219         }
 220
 221         /* set ctdbd capabilities */
 222         ctdb->capabilities = 0;
 223         if (options.no_lmaster == 0) {
 224                 ctdb->capabilities |= CTDB_CAP_LMASTER;
 225         }
 226         if (options.no_recmaster == 0) {
 227                 ctdb->capabilities |= CTDB_CAP_RECMASTER;
 228         }
 229         if (options.lvs != 0) {
 230                 ctdb->capabilities |= CTDB_CAP_LVS;
 231         }
 232
 233         /* Initialise this node's PNN to the unknown value.  This will
 234          * be set to the correct value by either ctdb_add_node() as
 235          * part of loading the nodes file or by
 236          * ctdb_tcp_listen_automatic() when the transport is
 237          * initialised.  At some point we should de-optimise this and
 238          * pull it out into ctdb_start_daemon() so it is done clearly
 239          * and only in one place.
 240          */
 241         ctdb->pnn = -1;
 242
 243         /* Default value for CTDB_BASE - don't override */
 244         setenv("CTDB_BASE", ETCDIR "/ctdb", 0);
 245
 246         /* tell ctdb what nodes are available */
 247         if (options.nlist != NULL) {
 248                 ctdb->nodes_file = options.nlist;
 249         } else {
 250                 ctdb->nodes_file =
 251                         talloc_asprintf(ctdb, "%s/nodes", getenv("CTDB_BASE"));
 252                 if (ctdb->nodes_file == NULL) {
 253                         DEBUG(DEBUG_ALERT,(__location__ " Out of memory\n"));
 254                         exit(1);
 255                 }
 256         }
 257         ctdb_load_nodes_file(ctdb);
 258
 259         ctdb->db_directory = options.db_dir;
 260         ctdb_mkdir_p_or_die(ctdb, ctdb->db_directory, 0700);
 261
 262         ctdb->db_directory_persistent = options.db_dir_persistent;
 263         ctdb_mkdir_p_or_die(ctdb, ctdb->db_directory_persistent, 0700);
 264
 265         ctdb->db_directory_state = options.db_dir_state;
 266         ctdb_mkdir_p_or_die(ctdb, ctdb->db_directory_state, 0700);
 267
 268         if (options.public_interface) {
 269                 ctdb->default_public_interface = talloc_strdup(ctdb, options.public_interface);
 270                 CTDB_NO_MEMORY(ctdb, ctdb->default_public_interface);
 271         }
 272
 273         if (options.single_public_ip) {
 274                 if (options.public_interface == NULL) {
 275                         DEBUG(DEBUG_ALERT,("--single_public_ip used but --public_interface is not specified. You must specify the public interface when using single public ip. Exiting\n"));
 276                         exit(10);
 277                 }
 278
 279                 ret = ctdb_set_single_public_ip(ctdb, options.public_interface,
 280                                                 options.single_public_ip);
 281                 if (ret != 0) {
 282                         DEBUG(DEBUG_ALERT,("Invalid --single-public-ip argument : %s . This is not a valid ip address. Exiting.\n", options.single_public_ip));
 283                         exit(10);
 284                 }
 285         }
 286
 287         if (options.event_script_dir != NULL) {
 288                 ctdb->event_script_dir = options.event_script_dir;
 289         } else {
 290                 ctdb->event_script_dir = talloc_asprintf(ctdb, "%s/events.d",
 291                                                          getenv("CTDB_BASE"));
 292                 if (ctdb->event_script_dir == NULL) {
 293                         DEBUG(DEBUG_ALERT,(__location__ " Out of memory\n"));
 294                         exit(1);
 295                 }
 296         }
 297
 298         if (options.notification_script != NULL) {
 299                 ret = ctdb_set_notification_script(ctdb, options.notification_script);
 300                 if (ret == -1) {
 301                         DEBUG(DEBUG_ALERT,("Unable to setup notification script\n"));
 302                         exit(1);
 303                 }
 304         }
 305
 306         ctdb->valgrinding = options.valgrinding;
 307         if (options.valgrinding || options.nosetsched) {
 308                 ctdb->do_setsched = 0;
 309         } else {
 310                 ctdb->do_setsched = 1;
 311         }
 312
 313         ctdb->public_addresses_file = options.public_address_list;
 314         ctdb->do_checkpublicip = !options.no_publicipcheck;
 315
 316         if (options.max_persistent_check_errors < 0) {
 317                 ctdb->max_persistent_check_errors = 0xFFFFFFFFFFFFFFFFLL;
 318         } else {
 319                 ctdb->max_persistent_check_errors = (uint64_t)options.max_persistent_check_errors;
 320         }
 321
 322         /* start the protocol running (as a child) */
 323         return ctdb_start_daemon(ctdb, interactive?false:true, options.use_syslog);
 324 }