3 * Bacula Director daemon -- this is the main program
5 * Kern Sibbald, March MM
10 Copyright (C) 2000-2004 Kern Sibbald and John Walker
12 This program is free software; you can redistribute it and/or
13 modify it under the terms of the GNU General Public License as
14 published by the Free Software Foundation; either version 2 of
15 the License, or (at your option) any later version.
17 This program is distributed in the hope that it will be useful,
18 but WITHOUT ANY WARRANTY; without even the implied warranty of
19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 General Public License for more details.
22 You should have received a copy of the GNU General Public
23 License along with this program; if not, write to the Free
24 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
32 /* Forward referenced subroutines */
33 static void terminate_dird(int sig);
34 static int check_resources();
36 /* Exported subroutines */
38 extern "C" void reload_config(int sig);
41 /* Imported subroutines */
42 JCR *wait_for_next_job(char *runjob);
43 void term_scheduler();
44 void term_ua_server();
45 int do_backup(JCR *jcr);
46 void backup_cleanup(void);
47 void start_UA_server(char *addr, int port);
48 void init_job_server(int max_workers);
49 void store_jobtype(LEX *lc, RES_ITEM *item, int index, int pass);
50 void store_level(LEX *lc, RES_ITEM *item, int index, int pass);
51 void store_replace(LEX *lc, RES_ITEM *item, int index, int pass);
53 static char *configfile = NULL;
54 static char *runjob = NULL;
55 static int background = 1;
56 static void init_reload(void);
58 /* Globals Exported */
59 DIRRES *director; /* Director resource */
63 /* Globals Imported */
64 extern int r_first, r_last; /* first and last resources */
65 extern RES_TABLE resources[];
66 extern RES **res_head;
67 extern RES_ITEM job_items[];
71 #define CONFIG_FILE "./bacula-dir.conf" /* default configuration file */
76 "\nVersion: " VERSION " (" BDATE ")\n\n"
77 "Usage: dird [-f -s] [-c config_file] [-d debug_level] [config_file]\n"
78 " -c <file> set configuration file to file\n"
79 " -dnn set debug level to nn\n"
80 " -f run in foreground (for debugging)\n"
82 " -r <job> run <job> now\n"
84 " -t test - read configuration and exit\n"
86 " -v verbose user messages\n"
87 " -? print this message.\n"
94 /*********************************************************************
96 * Main Bacula Server program
99 int main (int argc, char *argv[])
103 int no_signals = FALSE;
104 int test_config = FALSE;
109 my_name_is(argc, argv, "bacula-dir");
110 textdomain("bacula");
111 init_msg(NULL, NULL); /* initialize message handler */
113 daemon_start_time = time(NULL);
115 while ((ch = getopt(argc, argv, "c:d:fg:r:stu:v?")) != -1) {
117 case 'c': /* specify config file */
118 if (configfile != NULL) {
121 configfile = bstrdup(optarg);
124 case 'd': /* set debug level */
125 debug_level = atoi(optarg);
126 if (debug_level <= 0) {
129 Dmsg1(0, "Debug level = %d\n", debug_level);
132 case 'f': /* run in foreground */
136 case 'g': /* set group id */
140 case 'r': /* run job */
141 if (runjob != NULL) {
145 runjob = bstrdup(optarg);
149 case 's': /* turn off signals */
153 case 't': /* test config */
157 case 'u': /* set uid */
161 case 'v': /* verbose */
175 init_signals(terminate_dird);
179 if (configfile != NULL) {
182 configfile = bstrdup(*argv);
190 if (configfile == NULL) {
191 configfile = bstrdup(CONFIG_FILE);
194 parse_config(configfile);
196 if (!check_resources()) {
197 Jmsg((JCR *)NULL, M_ERROR_TERM, 0, _("Please correct configuration file: %s\n"), configfile);
204 my_name_is(0, NULL, director->hdr.name); /* set user defined name */
206 FDConnectTimeout = (int)director->FDConnectTimeout;
207 SDConnectTimeout = (int)director->SDConnectTimeout;
211 init_stack_dump(); /* grab new pid */
214 /* Create pid must come after we are a daemon -- so we have our final pid */
215 create_pid_file(director->pid_directory, "bacula-dir", director->DIRport);
216 read_state_file(director->working_directory, "bacula-dir", director->DIRport);
218 drop(uid, gid); /* reduce priveleges if requested */
220 signal(SIGHUP, reload_config);
222 init_console_msg(working_directory);
224 set_thread_concurrency(director->MaxConcurrentJobs * 2 +
225 4 /* UA */ + 4 /* sched+watchdog+jobsvr+misc */);
227 Dmsg0(200, "Start UA server\n");
228 start_UA_server(director->DIRaddr, director->DIRport);
230 start_watchdog(); /* start network watchdog thread */
232 init_jcr_subsystem(); /* start JCR watchdogs etc. */
234 init_job_server(director->MaxConcurrentJobs);
236 Dmsg0(200, "wait for next job\n");
237 /* Main loop -- call scheduler to get next job to run */
238 while ((jcr = wait_for_next_job(runjob))) {
239 run_job(jcr); /* run job */
240 free_jcr(jcr); /* release jcr */
241 if (runjob) { /* command line, run a single job? */
242 break; /* yes, terminate */
249 /* Cleanup and then exit */
250 static void terminate_dird(int sig)
252 static int already_here = FALSE;
254 if (already_here) { /* avoid recursive temination problems */
258 write_state_file(director->working_directory, "bacula-dir", director->DIRport);
259 delete_pid_file(director->pid_directory, "bacula-dir", director->DIRport);
260 // signal(SIGCHLD, SIG_IGN); /* don't worry about children now */
265 if (configfile != NULL) {
268 if (debug_level > 5) {
269 print_memory_pool_stats();
271 free_config_resources();
273 term_msg(); /* terminate message handler */
275 close_memory_pool(); /* release free memory in pool */
280 struct RELOAD_TABLE {
285 static const int max_reloads = 10;
286 static RELOAD_TABLE reload_table[max_reloads];
288 static void init_reload(void)
290 for (int i=0; i < max_reloads; i++) {
291 reload_table[i].job_count = 0;
292 reload_table[i].res_table = NULL;
296 static void free_saved_resources(int table)
298 int num = r_last - r_first + 1;
299 RES **res_tab = reload_table[table].res_table;
301 Dmsg1(100, "res_tab for table %d already released.\n", table);
304 Dmsg1(100, "Freeing resources for table %d\n", table);
305 for (int j=0; j<num; j++) {
306 free_resource(res_tab[j], r_first + j);
309 reload_table[table].job_count = 0;
310 reload_table[table].res_table = NULL;
314 * Called here at the end of every job that was
315 * hooked decrementing the active job_count. When
316 * it goes to zero, no one is using the associated
317 * resource table, so free it.
319 static void reload_job_end_cb(JCR *jcr, void *ctx)
321 int reload_id = (int)ctx;
322 Dmsg3(100, "reload job_end JobId=%d table=%d cnt=%d\n", jcr->JobId,
323 reload_id, reload_table[reload_id].job_count);
326 if (--reload_table[reload_id].job_count <= 0) {
327 free_saved_resources(reload_id);
333 static int find_free_reload_table_entry()
336 for (int i=0; i < max_reloads; i++) {
337 if (reload_table[i].res_table == NULL) {
346 * If we get here, we have received a SIGHUP, which means to
347 * reread our configuration file.
349 * The algorithm used is as follows: we count how many jobs are
350 * running and mark the running jobs to make a callback on
351 * exiting. The old config is saved with the reload table
352 * id in a reload table. The new config file is read. Now, as
353 * each job exits, it calls back to the reload_job_end_cb(), which
354 * decrements the count of open jobs for the given reload table.
355 * When the count goes to zero, we release those resources.
356 * This allows us to have pointers into the resource table (from
357 * jobs), and once they exit and all the pointers are released, we
358 * release the old table. Note, if no new jobs are running since the
359 * last reload, then the old resources will be immediately release.
360 * A console is considered a job because it may have pointers to
361 * resources, but a SYSTEM job is not since it *should* not have any
362 * permanent pointers to jobs.
365 void reload_config(int sig)
367 static bool already_here = false;
370 int njobs = 0; /* number of running jobs */
374 abort(); /* Oops, recursion -> die */
378 sigaddset(&set, SIGHUP);
379 sigprocmask(SIG_BLOCK, &set, NULL);
381 // Jmsg(NULL, M_INFO, 0, "Entering experimental reload config code. Bug reports will not be accepted.\n");
386 table = find_free_reload_table_entry();
388 Jmsg(NULL, M_ERROR, 0, _("Too many open reload requests. Request ignored.\n"));
392 Dmsg1(100, "Reload_config njobs=%d\n", njobs);
393 reload_table[table].res_table = save_config_resources();
394 Dmsg1(100, "Saved old config in table %d\n", table);
396 parse_config(configfile);
398 Dmsg0(100, "Reloaded config file\n");
399 if (!check_resources()) {
400 rtable = find_free_reload_table_entry(); /* save new, bad table */
402 Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
403 Jmsg(NULL, M_ERROR_TERM, 0, _("Out of reload table entries. Giving up.\n"));
405 Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
407 reload_table[rtable].res_table = save_config_resources();
408 /* Now restore old resoure values */
409 int num = r_last - r_first + 1;
410 RES **res_tab = reload_table[table].res_table;
411 for (int i=0; i<num; i++) {
412 res_head[i] = res_tab[i];
414 table = rtable; /* release new, bad, saved table below */
417 * Hook all active jobs so that they release this table
420 if (jcr->JobType != JT_SYSTEM) {
421 reload_table[table].job_count++;
422 job_end_push(jcr, reload_job_end_cb, (void *)table);
425 free_locked_jcr(jcr);
430 set_working_directory(director->working_directory);
431 FDConnectTimeout = director->FDConnectTimeout;
432 SDConnectTimeout = director->SDConnectTimeout;
433 Dmsg0(0, "Director's configuration file reread.\n");
435 /* Now release saved resources, if no jobs using the resources */
437 free_saved_resources(table);
443 sigprocmask(SIG_UNBLOCK, &set, NULL);
444 signal(SIGHUP, reload_config);
445 already_here = false;
449 * Make a quick check to see that we have all the
452 * **** FIXME **** this routine could be a lot more
453 * intelligent and comprehensive.
455 static int check_resources()
462 job = (JOB *)GetNextRes(R_JOB, NULL);
463 director = (DIRRES *)GetNextRes(R_DIRECTOR, NULL);
465 Jmsg(NULL, M_FATAL, 0, _("No Director resource defined in %s\n\
466 Without that I don't know who I am :-(\n"), configfile);
469 set_working_directory(director->working_directory);
470 if (!director->messages) { /* If message resource not specified */
471 director->messages = (MSGS *)GetNextRes(R_MSGS, NULL);
472 if (!director->messages) {
473 Jmsg(NULL, M_FATAL, 0, _("No Messages resource defined in %s\n"), configfile);
477 if (GetNextRes(R_DIRECTOR, (RES *)director) != NULL) {
478 Jmsg(NULL, M_FATAL, 0, _("Only one Director resource permitted in %s\n"),
485 Jmsg(NULL, M_FATAL, 0, _("No Job records defined in %s\n"), configfile);
488 foreach_res(job, R_JOB) {
492 /* Transfer default items from JobDefs Resource */
493 for (i=0; job_items[i].name; i++) {
494 char **def_svalue, **svalue; /* string value */
495 int *def_ivalue, *ivalue; /* integer value */
496 int64_t *def_lvalue, *lvalue; /* 64 bit values */
499 Dmsg4(400, "Job \"%s\", field \"%s\" bit=%d def=%d\n",
500 job->hdr.name, job_items[i].name,
501 bit_is_set(i, job->hdr.item_present),
502 bit_is_set(i, job->jobdefs->hdr.item_present));
504 if (!bit_is_set(i, job->hdr.item_present) &&
505 bit_is_set(i, job->jobdefs->hdr.item_present)) {
506 Dmsg2(400, "Job \"%s\", field \"%s\": getting default.\n",
507 job->hdr.name, job_items[i].name);
508 offset = (char *)(job_items[i].value) - (char *)&res_all;
510 * Handle strings and directory strings
512 if (job_items[i].handler == store_str ||
513 job_items[i].handler == store_dir) {
514 def_svalue = (char **)((char *)(job->jobdefs) + offset);
515 Dmsg5(400, "Job \"%s\", field \"%s\" def_svalue=%s item %d offset=%u\n",
516 job->hdr.name, job_items[i].name, *def_svalue, i, offset);
517 svalue = (char **)((char *)job + offset);
519 Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
521 *svalue = bstrdup(*def_svalue);
522 set_bit(i, job->hdr.item_present);
523 } else if (job_items[i].handler == store_res) {
524 def_svalue = (char **)((char *)(job->jobdefs) + offset);
525 Dmsg4(400, "Job \"%s\", field \"%s\" item %d offset=%u\n",
526 job->hdr.name, job_items[i].name, i, offset);
527 svalue = (char **)((char *)job + offset);
529 Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
531 *svalue = *def_svalue;
532 set_bit(i, job->hdr.item_present);
534 * Handle integer fields
535 * Note, our store_yesno does not handle bitmaped fields
537 } else if (job_items[i].handler == store_yesno ||
538 job_items[i].handler == store_pint ||
539 job_items[i].handler == store_jobtype ||
540 job_items[i].handler == store_level ||
541 job_items[i].handler == store_pint ||
542 job_items[i].handler == store_replace) {
543 def_ivalue = (int *)((char *)(job->jobdefs) + offset);
544 Dmsg5(400, "Job \"%s\", field \"%s\" def_ivalue=%d item %d offset=%u\n",
545 job->hdr.name, job_items[i].name, *def_ivalue, i, offset);
546 ivalue = (int *)((char *)job + offset);
547 *ivalue = *def_ivalue;
548 set_bit(i, job->hdr.item_present);
550 * Handle 64 bit integer fields
552 } else if (job_items[i].handler == store_time ||
553 job_items[i].handler == store_size ||
554 job_items[i].handler == store_int64) {
555 def_lvalue = (int64_t *)((char *)(job->jobdefs) + offset);
556 Dmsg5(400, "Job \"%s\", field \"%s\" def_lvalue=%" lld " item %d offset=%u\n",
557 job->hdr.name, job_items[i].name, *def_lvalue, i, offset);
558 lvalue = (int64_t *)((char *)job + offset);
559 *lvalue = *def_lvalue;
560 set_bit(i, job->hdr.item_present);
566 * Ensure that all required items are present
568 for (i=0; job_items[i].name; i++) {
569 if (job_items[i].flags & ITEM_REQUIRED) {
570 if (!bit_is_set(i, job->hdr.item_present)) {
571 Jmsg(NULL, M_FATAL, 0, "Field \"%s\" in Job \"%s\" resource is required, but not found.\n",
572 job_items[i].name, job->hdr.name);
576 /* If this triggers, take a look at lib/parse_conf.h */
577 if (i >= MAX_RES_ITEMS) {
578 Emsg0(M_ERROR_TERM, 0, "Too many items in Job resource\n");
581 if (job->client && job->client->catalog) {
582 CAT *catalog = job->client->catalog;
586 * Make sure we can open catalog, otherwise print a warning
587 * message because the server is probably not running.
589 db = db_init_database(NULL, catalog->db_name, catalog->db_user,
590 catalog->db_password, catalog->db_address,
591 catalog->db_port, catalog->db_socket);
592 if (!db || !db_open_database(NULL, db)) {
593 Jmsg(NULL, M_FATAL, 0, _("Could not open database \"%s\".\n"),
596 Jmsg(NULL, M_FATAL, 0, _("%s"), db_strerror(db));
600 /* If a pool is defined for this job, create the pool DB
601 * record if it is not already created.
604 create_pool(NULL, db, job->pool, POOL_OP_UPDATE); /* update request */
606 /* Set default value in all counters */
608 foreach_res(counter, R_COUNTER) {
609 /* Write to catalog? */
610 if (!counter->created && counter->Catalog == catalog) {
612 bstrncpy(cr.Counter, counter->hdr.name, sizeof(cr.Counter));
613 cr.MinValue = counter->MinValue;
614 cr.MaxValue = counter->MaxValue;
615 cr.CurrentValue = counter->MinValue;
616 if (counter->WrapCounter) {
617 bstrncpy(cr.WrapCounter, counter->WrapCounter->hdr.name, sizeof(cr.WrapCounter));
619 cr.WrapCounter[0] = 0; /* empty string */
621 if (db_create_counter_record(NULL, db, &cr)) {
622 counter->CurrentValue = cr.CurrentValue;
623 counter->created = true;
624 Dmsg2(100, "Create counter %s val=%d\n", counter->hdr.name, counter->CurrentValue);
627 if (!counter->created) {
628 counter->CurrentValue = counter->MinValue; /* default value */
632 db_close_database(NULL, db);
638 close_msg(NULL); /* close temp message handler */
639 init_msg(NULL, director->messages); /* open daemon message handler */