3 * Bacula Director daemon -- this is the main program
5 * Kern Sibbald, March MM
10 Copyright (C) 2000-2005 Kern Sibbald
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(dlist *addrs);
48 void init_job_server(int max_workers);
49 void term_job_server();
50 void store_jobtype(LEX *lc, RES_ITEM *item, int index, int pass);
51 void store_level(LEX *lc, RES_ITEM *item, int index, int pass);
52 void store_replace(LEX *lc, RES_ITEM *item, int index, int pass);
53 void init_device_resources();
55 static char *configfile = NULL;
56 static char *runjob = NULL;
57 static int background = 1;
58 static void init_reload(void);
60 /* Globals Exported */
61 DIRRES *director; /* Director resource */
65 /* Globals Imported */
66 extern int r_first, r_last; /* first and last resources */
67 extern RES_TABLE resources[];
68 extern RES **res_head;
69 extern RES_ITEM job_items[];
73 #define CONFIG_FILE "./bacula-dir.conf" /* default configuration file */
78 "Copyright (C) 2000-2005 Kern Sibbald.\n"
79 "\nVersion: " VERSION " (" BDATE ")\n\n"
80 "Usage: dird [-f -s] [-c config_file] [-d debug_level] [config_file]\n"
81 " -c <file> set configuration file to file\n"
82 " -dnn set debug level to nn\n"
83 " -f run in foreground (for debugging)\n"
85 " -r <job> run <job> now\n"
87 " -t test - read configuration and exit\n"
89 " -v verbose user messages\n"
90 " -? print this message.\n"
97 /*********************************************************************
99 * Main Bacula Server program
102 int main (int argc, char *argv[])
106 int no_signals = FALSE;
107 int test_config = FALSE;
112 my_name_is(argc, argv, "bacula-dir");
113 textdomain("bacula");
114 init_msg(NULL, NULL); /* initialize message handler */
116 daemon_start_time = time(NULL);
118 while ((ch = getopt(argc, argv, "c:d:fg:r:stu:v?")) != -1) {
120 case 'c': /* specify config file */
121 if (configfile != NULL) {
124 configfile = bstrdup(optarg);
127 case 'd': /* set debug level */
128 debug_level = atoi(optarg);
129 if (debug_level <= 0) {
132 Dmsg1(0, "Debug level = %d\n", debug_level);
135 case 'f': /* run in foreground */
139 case 'g': /* set group id */
143 case 'r': /* run job */
144 if (runjob != NULL) {
148 runjob = bstrdup(optarg);
152 case 's': /* turn off signals */
156 case 't': /* test config */
160 case 'u': /* set uid */
164 case 'v': /* verbose */
178 init_signals(terminate_dird);
182 if (configfile != NULL) {
185 configfile = bstrdup(*argv);
193 if (configfile == NULL) {
194 configfile = bstrdup(CONFIG_FILE);
197 parse_config(configfile);
199 if (!check_resources()) {
200 Jmsg((JCR *)NULL, M_ERROR_TERM, 0, _("Please correct configuration file: %s\n"), configfile);
207 my_name_is(0, NULL, director->hdr.name); /* set user defined name */
209 FDConnectTimeout = (int)director->FDConnectTimeout;
210 SDConnectTimeout = (int)director->SDConnectTimeout;
214 init_stack_dump(); /* grab new pid */
217 /* Create pid must come after we are a daemon -- so we have our final pid */
218 create_pid_file(director->pid_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
219 read_state_file(director->working_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
221 drop(uid, gid); /* reduce priveleges if requested */
223 signal(SIGHUP, reload_config);
225 init_console_msg(working_directory);
227 init_python_interpreter(director->hdr.name, director->scripts_directory ?
228 director->scripts_directory : ".");
230 set_thread_concurrency(director->MaxConcurrentJobs * 2 +
231 4 /* UA */ + 4 /* sched+watchdog+jobsvr+misc */);
233 Dmsg0(200, "Start UA server\n");
234 start_UA_server(director->DIRaddrs);
236 start_watchdog(); /* start network watchdog thread */
238 init_jcr_subsystem(); /* start JCR watchdogs etc. */
240 init_job_server(director->MaxConcurrentJobs);
242 init_device_resources();
244 Dmsg0(200, "wait for next job\n");
245 /* Main loop -- call scheduler to get next job to run */
246 while ((jcr = wait_for_next_job(runjob))) {
247 run_job(jcr); /* run job */
248 free_jcr(jcr); /* release jcr */
249 if (runjob) { /* command line, run a single job? */
250 break; /* yes, terminate */
257 /* Cleanup and then exit */
258 static void terminate_dird(int sig)
260 static int already_here = FALSE;
262 if (already_here) { /* avoid recursive temination problems */
266 write_state_file(director->working_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
267 delete_pid_file(director->pid_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
268 // signal(SIGCHLD, SIG_IGN); /* don't worry about children now */
274 if (configfile != NULL) {
277 if (debug_level > 5) {
278 print_memory_pool_stats();
280 free_config_resources();
282 term_msg(); /* terminate message handler */
284 close_memory_pool(); /* release free memory in pool */
289 struct RELOAD_TABLE {
294 static const int max_reloads = 10;
295 static RELOAD_TABLE reload_table[max_reloads];
297 static void init_reload(void)
299 for (int i=0; i < max_reloads; i++) {
300 reload_table[i].job_count = 0;
301 reload_table[i].res_table = NULL;
305 static void free_saved_resources(int table)
307 int num = r_last - r_first + 1;
308 RES **res_tab = reload_table[table].res_table;
310 Dmsg1(100, "res_tab for table %d already released.\n", table);
313 Dmsg1(100, "Freeing resources for table %d\n", table);
314 for (int j=0; j<num; j++) {
315 free_resource(res_tab[j], r_first + j);
318 reload_table[table].job_count = 0;
319 reload_table[table].res_table = NULL;
323 * Called here at the end of every job that was
324 * hooked decrementing the active job_count. When
325 * it goes to zero, no one is using the associated
326 * resource table, so free it.
328 static void reload_job_end_cb(JCR *jcr, void *ctx)
330 int reload_id = (int)((long int)ctx);
331 Dmsg3(100, "reload job_end JobId=%d table=%d cnt=%d\n", jcr->JobId,
332 reload_id, reload_table[reload_id].job_count);
335 if (--reload_table[reload_id].job_count <= 0) {
336 free_saved_resources(reload_id);
342 static int find_free_reload_table_entry()
345 for (int i=0; i < max_reloads; i++) {
346 if (reload_table[i].res_table == NULL) {
355 * If we get here, we have received a SIGHUP, which means to
356 * reread our configuration file.
358 * The algorithm used is as follows: we count how many jobs are
359 * running and mark the running jobs to make a callback on
360 * exiting. The old config is saved with the reload table
361 * id in a reload table. The new config file is read. Now, as
362 * each job exits, it calls back to the reload_job_end_cb(), which
363 * decrements the count of open jobs for the given reload table.
364 * When the count goes to zero, we release those resources.
365 * This allows us to have pointers into the resource table (from
366 * jobs), and once they exit and all the pointers are released, we
367 * release the old table. Note, if no new jobs are running since the
368 * last reload, then the old resources will be immediately release.
369 * A console is considered a job because it may have pointers to
370 * resources, but a SYSTEM job is not since it *should* not have any
371 * permanent pointers to jobs.
374 void reload_config(int sig)
376 static bool already_here = false;
379 int njobs = 0; /* number of running jobs */
384 abort(); /* Oops, recursion -> die */
388 sigaddset(&set, SIGHUP);
389 sigprocmask(SIG_BLOCK, &set, NULL);
391 // Jmsg(NULL, M_INFO, 0, "Entering experimental reload config code. Bug reports will not be accepted.\n");
396 table = find_free_reload_table_entry();
398 Jmsg(NULL, M_ERROR, 0, _("Too many open reload requests. Request ignored.\n"));
402 Dmsg1(100, "Reload_config njobs=%d\n", njobs);
403 reload_table[table].res_table = save_config_resources();
404 Dmsg1(100, "Saved old config in table %d\n", table);
406 ok = parse_config(configfile, 0); /* no exit on error */
408 Dmsg0(100, "Reloaded config file\n");
409 if (!ok || !check_resources()) {
410 rtable = find_free_reload_table_entry(); /* save new, bad table */
412 Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
413 Jmsg(NULL, M_ERROR_TERM, 0, _("Out of reload table entries. Giving up.\n"));
415 Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
416 Jmsg(NULL, M_ERROR, 0, _("Resetting previous configuration.\n"));
418 reload_table[rtable].res_table = save_config_resources();
419 /* Now restore old resoure values */
420 int num = r_last - r_first + 1;
421 RES **res_tab = reload_table[table].res_table;
422 for (int i=0; i<num; i++) {
423 res_head[i] = res_tab[i];
425 table = rtable; /* release new, bad, saved table below */
428 * Hook all active jobs so that they release this table
431 if (jcr->JobType != JT_SYSTEM) {
432 reload_table[table].job_count++;
433 job_end_push(jcr, reload_job_end_cb, (void *)((long int)table));
436 free_locked_jcr(jcr);
441 set_working_directory(director->working_directory);
442 FDConnectTimeout = director->FDConnectTimeout;
443 SDConnectTimeout = director->SDConnectTimeout;
444 Dmsg0(0, "Director's configuration file reread.\n");
446 init_device_resources(); /* Update Device resources */
448 /* Now release saved resources, if no jobs using the resources */
450 free_saved_resources(table);
456 sigprocmask(SIG_UNBLOCK, &set, NULL);
457 signal(SIGHUP, reload_config);
458 already_here = false;
462 * Make a quick check to see that we have all the
465 * **** FIXME **** this routine could be a lot more
466 * intelligent and comprehensive.
468 static int check_resources()
475 job = (JOB *)GetNextRes(R_JOB, NULL);
476 director = (DIRRES *)GetNextRes(R_DIRECTOR, NULL);
478 Jmsg(NULL, M_FATAL, 0, _("No Director resource defined in %s\n"
479 "Without that I don't know who I am :-(\n"), configfile);
482 set_working_directory(director->working_directory);
483 if (!director->messages) { /* If message resource not specified */
484 director->messages = (MSGS *)GetNextRes(R_MSGS, NULL);
485 if (!director->messages) {
486 Jmsg(NULL, M_FATAL, 0, _("No Messages resource defined in %s\n"), configfile);
490 if (GetNextRes(R_DIRECTOR, (RES *)director) != NULL) {
491 Jmsg(NULL, M_FATAL, 0, _("Only one Director resource permitted in %s\n"),
498 Jmsg(NULL, M_FATAL, 0, _("No Job records defined in %s\n"), configfile);
501 foreach_res(job, R_JOB) {
505 /* Handle Storage alists specifically */
506 JOB *jobdefs = job->jobdefs;
507 if (jobdefs->storage && !job->storage) {
509 job->storage = New(alist(10, not_owned_by_alist));
510 foreach_alist(st, jobdefs->storage) {
511 job->storage->append(st);
515 /* Transfer default items from JobDefs Resource */
516 for (i=0; job_items[i].name; i++) {
517 char **def_svalue, **svalue; /* string value */
518 int *def_ivalue, *ivalue; /* integer value */
519 int64_t *def_lvalue, *lvalue; /* 64 bit values */
522 Dmsg4(400, "Job \"%s\", field \"%s\" bit=%d def=%d\n",
523 job->hdr.name, job_items[i].name,
524 bit_is_set(i, job->hdr.item_present),
525 bit_is_set(i, job->jobdefs->hdr.item_present));
527 if (!bit_is_set(i, job->hdr.item_present) &&
528 bit_is_set(i, job->jobdefs->hdr.item_present)) {
529 Dmsg2(400, "Job \"%s\", field \"%s\": getting default.\n",
530 job->hdr.name, job_items[i].name);
531 offset = (char *)(job_items[i].value) - (char *)&res_all;
533 * Handle strings and directory strings
535 if (job_items[i].handler == store_str ||
536 job_items[i].handler == store_dir) {
537 def_svalue = (char **)((char *)(job->jobdefs) + offset);
538 Dmsg5(400, "Job \"%s\", field \"%s\" def_svalue=%s item %d offset=%u\n",
539 job->hdr.name, job_items[i].name, *def_svalue, i, offset);
540 svalue = (char **)((char *)job + offset);
542 Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
544 *svalue = bstrdup(*def_svalue);
545 set_bit(i, job->hdr.item_present);
549 } else if (job_items[i].handler == store_res) {
550 def_svalue = (char **)((char *)(job->jobdefs) + offset);
551 Dmsg4(400, "Job \"%s\", field \"%s\" item %d offset=%u\n",
552 job->hdr.name, job_items[i].name, i, offset);
553 svalue = (char **)((char *)job + offset);
555 Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
557 *svalue = *def_svalue;
558 set_bit(i, job->hdr.item_present);
560 * Handle alist resources
562 } else if (job_items[i].handler == store_alist_res) {
563 if (bit_is_set(i, job->jobdefs->hdr.item_present)) {
564 set_bit(i, job->hdr.item_present);
567 * Handle integer fields
568 * Note, our store_yesno does not handle bitmaped fields
570 } else if (job_items[i].handler == store_yesno ||
571 job_items[i].handler == store_pint ||
572 job_items[i].handler == store_jobtype ||
573 job_items[i].handler == store_level ||
574 job_items[i].handler == store_pint ||
575 job_items[i].handler == store_replace) {
576 def_ivalue = (int *)((char *)(job->jobdefs) + offset);
577 Dmsg5(400, "Job \"%s\", field \"%s\" def_ivalue=%d item %d offset=%u\n",
578 job->hdr.name, job_items[i].name, *def_ivalue, i, offset);
579 ivalue = (int *)((char *)job + offset);
580 *ivalue = *def_ivalue;
581 set_bit(i, job->hdr.item_present);
583 * Handle 64 bit integer fields
585 } else if (job_items[i].handler == store_time ||
586 job_items[i].handler == store_size ||
587 job_items[i].handler == store_int64) {
588 def_lvalue = (int64_t *)((char *)(job->jobdefs) + offset);
589 Dmsg5(400, "Job \"%s\", field \"%s\" def_lvalue=%" lld " item %d offset=%u\n",
590 job->hdr.name, job_items[i].name, *def_lvalue, i, offset);
591 lvalue = (int64_t *)((char *)job + offset);
592 *lvalue = *def_lvalue;
593 set_bit(i, job->hdr.item_present);
599 * Ensure that all required items are present
601 for (i=0; job_items[i].name; i++) {
602 if (job_items[i].flags & ITEM_REQUIRED) {
603 if (!bit_is_set(i, job->hdr.item_present)) {
604 Jmsg(NULL, M_FATAL, 0, "\"%s\" directive in Job \"%s\" resource is required, but not found.\n",
605 job_items[i].name, job->hdr.name);
609 /* If this triggers, take a look at lib/parse_conf.h */
610 if (i >= MAX_RES_ITEMS) {
611 Emsg0(M_ERROR_TERM, 0, "Too many items in Job resource\n");
614 } /* End loop over Job res */
616 /* Loop over databases */
618 foreach_res(catalog, R_CATALOG) {
621 * Make sure we can open catalog, otherwise print a warning
622 * message because the server is probably not running.
624 db = db_init_database(NULL, catalog->db_name, catalog->db_user,
625 catalog->db_password, catalog->db_address,
626 catalog->db_port, catalog->db_socket,
627 catalog->mult_db_connections);
628 if (!db || !db_open_database(NULL, db)) {
629 Jmsg(NULL, M_FATAL, 0, _("Could not open database \"%s\".\n"),
632 Jmsg(NULL, M_FATAL, 0, _("%s"), db_strerror(db));
638 /* Loop over all pools, defining/updating them in each database */
640 foreach_res(pool, R_POOL) {
641 create_pool(NULL, db, pool, POOL_OP_UPDATE); /* update request */
644 /* ***FIXME*** we need to update store and media_type records */
646 foreach_res(store, R_STORAGE) {
649 if (store->media_type) {
650 bstrncpy(mr.MediaType, store->media_type, sizeof(mr.MediaType));
652 db_create_mediatype_record(NULL, db, &mr);
656 sr.MediaTypeId = mr.MediaTypeId;
657 bstrncpy(sr.Name, store->name(), sizeof(sr.Name));
658 sr.AutoChanger = store->autochanger;
659 db_create_storage_record(NULL, db, &sr);
662 /* Loop over all counters, defining them in each database */
663 /* Set default value in all counters */
665 foreach_res(counter, R_COUNTER) {
666 /* Write to catalog? */
667 if (!counter->created && counter->Catalog == catalog) {
669 bstrncpy(cr.Counter, counter->hdr.name, sizeof(cr.Counter));
670 cr.MinValue = counter->MinValue;
671 cr.MaxValue = counter->MaxValue;
672 cr.CurrentValue = counter->MinValue;
673 if (counter->WrapCounter) {
674 bstrncpy(cr.WrapCounter, counter->WrapCounter->hdr.name, sizeof(cr.WrapCounter));
676 cr.WrapCounter[0] = 0; /* empty string */
678 if (db_create_counter_record(NULL, db, &cr)) {
679 counter->CurrentValue = cr.CurrentValue;
680 counter->created = true;
681 Dmsg2(100, "Create counter %s val=%d\n", counter->hdr.name, counter->CurrentValue);
684 if (!counter->created) {
685 counter->CurrentValue = counter->MinValue; /* default value */
688 db_close_database(NULL, db);
693 close_msg(NULL); /* close temp message handler */
694 init_msg(NULL, director->messages); /* open daemon message handler */