* Version $Id$
*/
/*
- Copyright (C) 2000-2003 Kern Sibbald and John Walker
+ Copyright (C) 2000-2004 Kern Sibbald and John Walker
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
/* Forward referenced subroutines */
static void terminate_dird(int sig);
static int check_resources();
-static void reload_config(int sig);
/* Exported subroutines */
+extern "C" void reload_config(int sig);
+
/* Imported subroutines */
-extern JCR *wait_for_next_job(char *runjob);
-extern void term_scheduler();
-extern void term_ua_server();
-extern int do_backup(JCR *jcr);
-extern void backup_cleanup(void);
-extern void start_UA_server(char *addr, int port);
-extern void run_job(JCR *jcr);
-extern void init_job_server(int max_workers);
+JCR *wait_for_next_job(char *runjob);
+void term_scheduler();
+void term_ua_server();
+int do_backup(JCR *jcr);
+void backup_cleanup(void);
+void start_UA_server(dlist *addrs);
+void init_job_server(int max_workers);
+void term_job_server();
+void store_jobtype(LEX *lc, RES_ITEM *item, int index, int pass);
+void store_level(LEX *lc, RES_ITEM *item, int index, int pass);
+void store_replace(LEX *lc, RES_ITEM *item, int index, int pass);
static char *configfile = NULL;
static char *runjob = NULL;
static int background = 1;
+static void init_reload(void);
/* Globals Exported */
DIRRES *director; /* Director resource */
int FDConnectTimeout;
int SDConnectTimeout;
+/* Globals Imported */
+extern int r_first, r_last; /* first and last resources */
+extern RES_TABLE resources[];
+extern RES **res_head;
+extern RES_ITEM job_items[];
+extern URES res_all;
+
+
#define CONFIG_FILE "./bacula-dir.conf" /* default configuration file */
static void usage()
{
fprintf(stderr, _(
+"Copyright (C) 2000-2004 Kern Sibbald and John Walker\n"
"\nVersion: " VERSION " (" BDATE ")\n\n"
"Usage: dird [-f -s] [-c config_file] [-d debug_level] [config_file]\n"
" -c <file> set configuration file to file\n"
init_stack_dump();
my_name_is(argc, argv, "bacula-dir");
- textdomain("bacula-dir");
+ textdomain("bacula");
init_msg(NULL, NULL); /* initialize message handler */
+ init_reload();
daemon_start_time = time(NULL);
- memset(&last_job, 0, sizeof(last_job));
while ((ch = getopt(argc, argv, "c:d:fg:r:stu:v?")) != -1) {
switch (ch) {
- case 'c': /* specify config file */
- if (configfile != NULL) {
- free(configfile);
- }
- configfile = bstrdup(optarg);
- break;
+ case 'c': /* specify config file */
+ if (configfile != NULL) {
+ free(configfile);
+ }
+ configfile = bstrdup(optarg);
+ break;
- case 'd': /* set debug level */
- debug_level = atoi(optarg);
- if (debug_level <= 0) {
- debug_level = 1;
- }
- Dmsg1(0, "Debug level = %d\n", debug_level);
- break;
+ case 'd': /* set debug level */
+ debug_level = atoi(optarg);
+ if (debug_level <= 0) {
+ debug_level = 1;
+ }
+ Dmsg1(0, "Debug level = %d\n", debug_level);
+ break;
- case 'f': /* run in foreground */
- background = FALSE;
- break;
+ case 'f': /* run in foreground */
+ background = FALSE;
+ break;
- case 'g': /* set group id */
- gid = optarg;
- break;
+ case 'g': /* set group id */
+ gid = optarg;
+ break;
- case 'r': /* run job */
- if (runjob != NULL) {
- free(runjob);
- }
- if (optarg) {
- runjob = bstrdup(optarg);
- }
- break;
+ case 'r': /* run job */
+ if (runjob != NULL) {
+ free(runjob);
+ }
+ if (optarg) {
+ runjob = bstrdup(optarg);
+ }
+ break;
- case 's': /* turn off signals */
- no_signals = TRUE;
- break;
+ case 's': /* turn off signals */
+ no_signals = TRUE;
+ break;
- case 't': /* test config */
- test_config = TRUE;
- break;
+ case 't': /* test config */
+ test_config = TRUE;
+ break;
- case 'u': /* set uid */
- uid = optarg;
- break;
+ case 'u': /* set uid */
+ uid = optarg;
+ break;
- case 'v': /* verbose */
- verbose++;
- break;
+ case 'v': /* verbose */
+ verbose++;
+ break;
- case '?':
- default:
- usage();
+ case '?':
+ default:
+ usage();
}
}
if (!no_signals) {
init_signals(terminate_dird);
}
- signal(SIGCHLD, SIG_IGN);
if (argc) {
if (configfile != NULL) {
init_stack_dump(); /* grab new pid */
}
- drop(uid, gid); /* reduce priveleges if requested */
-
/* Create pid must come after we are a daemon -- so we have our final pid */
- create_pid_file(director->pid_directory, "bacula-dir", director->DIRport);
+ create_pid_file(director->pid_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
+ read_state_file(director->working_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
-/* signal(SIGHUP, reload_config); */
+ drop(uid, gid); /* reduce priveleges if requested */
+
+ signal(SIGHUP, reload_config);
init_console_msg(working_directory);
4 /* UA */ + 4 /* sched+watchdog+jobsvr+misc */);
Dmsg0(200, "Start UA server\n");
- start_UA_server(director->DIRaddr, director->DIRport);
+ start_UA_server(director->DIRaddrs);
start_watchdog(); /* start network watchdog thread */
+ init_jcr_subsystem(); /* start JCR watchdogs etc. */
+
init_job_server(director->MaxConcurrentJobs);
Dmsg0(200, "wait for next job\n");
/* Main loop -- call scheduler to get next job to run */
while ((jcr = wait_for_next_job(runjob))) {
run_job(jcr); /* run job */
+ free_jcr(jcr); /* release jcr */
if (runjob) { /* command line, run a single job? */
break; /* yes, terminate */
}
exit(1);
}
already_here = TRUE;
- delete_pid_file(director->pid_directory, "bacula-dir",
- director->DIRport);
- stop_watchdog();
- signal(SIGCHLD, SIG_IGN); /* don't worry about children now */
+ write_state_file(director->working_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
+ delete_pid_file(director->pid_directory, "bacula-dir", get_first_port_host_order(director->DIRaddrs));
+// signal(SIGCHLD, SIG_IGN); /* don't worry about children now */
term_scheduler();
+ term_job_server();
if (runjob) {
free(runjob);
}
free_config_resources();
term_ua_server();
term_msg(); /* terminate message handler */
+ stop_watchdog();
close_memory_pool(); /* release free memory in pool */
- sm_dump(False);
- exit(sig != 0);
+ sm_dump(false);
+ exit(sig);
+}
+
+struct RELOAD_TABLE {
+ int job_count;
+ RES **res_table;
+};
+
+static const int max_reloads = 10;
+static RELOAD_TABLE reload_table[max_reloads];
+
+static void init_reload(void)
+{
+ for (int i=0; i < max_reloads; i++) {
+ reload_table[i].job_count = 0;
+ reload_table[i].res_table = NULL;
+ }
+}
+
+static void free_saved_resources(int table)
+{
+ int num = r_last - r_first + 1;
+ RES **res_tab = reload_table[table].res_table;
+ if (!res_tab) {
+ Dmsg1(100, "res_tab for table %d already released.\n", table);
+ return;
+ }
+ Dmsg1(100, "Freeing resources for table %d\n", table);
+ for (int j=0; j<num; j++) {
+ free_resource(res_tab[j], r_first + j);
+ }
+ free(res_tab);
+ reload_table[table].job_count = 0;
+ reload_table[table].res_table = NULL;
+}
+
+/*
+ * Called here at the end of every job that was
+ * hooked decrementing the active job_count. When
+ * it goes to zero, no one is using the associated
+ * resource table, so free it.
+ */
+static void reload_job_end_cb(JCR *jcr, void *ctx)
+{
+ int reload_id = (int)((long int)ctx);
+ Dmsg3(100, "reload job_end JobId=%d table=%d cnt=%d\n", jcr->JobId,
+ reload_id, reload_table[reload_id].job_count);
+ lock_jcr_chain();
+ LockRes();
+ if (--reload_table[reload_id].job_count <= 0) {
+ free_saved_resources(reload_id);
+ }
+ UnlockRes();
+ unlock_jcr_chain();
+}
+
+static int find_free_reload_table_entry()
+{
+ int table = -1;
+ for (int i=0; i < max_reloads; i++) {
+ if (reload_table[i].res_table == NULL) {
+ table = i;
+ break;
+ }
+ }
+ return table;
}
/*
* If we get here, we have received a SIGHUP, which means to
- * reread our configuration file.
+ * reread our configuration file.
*
- * ***FIXME*** Check that there are no jobs running before
- * doing this.
+ * The algorithm used is as follows: we count how many jobs are
+ * running and mark the running jobs to make a callback on
+ * exiting. The old config is saved with the reload table
+ * id in a reload table. The new config file is read. Now, as
+ * each job exits, it calls back to the reload_job_end_cb(), which
+ * decrements the count of open jobs for the given reload table.
+ * When the count goes to zero, we release those resources.
+ * This allows us to have pointers into the resource table (from
+ * jobs), and once they exit and all the pointers are released, we
+ * release the old table. Note, if no new jobs are running since the
+ * last reload, then the old resources will be immediately release.
+ * A console is considered a job because it may have pointers to
+ * resources, but a SYSTEM job is not since it *should* not have any
+ * permanent pointers to jobs.
*/
-static void reload_config(int sig)
+extern "C"
+void reload_config(int sig)
{
- static int already_here = FALSE;
+ static bool already_here = false;
sigset_t set;
+ JCR *jcr;
+ int njobs = 0; /* number of running jobs */
+ int table, rtable;
if (already_here) {
abort(); /* Oops, recursion -> die */
}
- already_here = TRUE;
- sigfillset(&set);
+ already_here = true;
+ sigemptyset(&set);
+ sigaddset(&set, SIGHUP);
sigprocmask(SIG_BLOCK, &set, NULL);
- free_config_resources();
+// Jmsg(NULL, M_INFO, 0, "Entering experimental reload config code. Bug reports will not be accepted.\n");
+
+ lock_jcr_chain();
+ LockRes();
+
+ table = find_free_reload_table_entry();
+ if (table < 0) {
+ Jmsg(NULL, M_ERROR, 0, _("Too many open reload requests. Request ignored.\n"));
+ goto bail_out;
+ }
+
+ Dmsg1(100, "Reload_config njobs=%d\n", njobs);
+ reload_table[table].res_table = save_config_resources();
+ Dmsg1(100, "Saved old config in table %d\n", table);
parse_config(configfile);
- Dmsg0(200, "check_resources()\n");
+ Dmsg0(100, "Reloaded config file\n");
if (!check_resources()) {
- Jmsg(NULL, M_ERROR_TERM, 0, _("Please correct configuration file: %s\n"), configfile);
+ rtable = find_free_reload_table_entry(); /* save new, bad table */
+ if (rtable < 0) {
+ Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
+ Jmsg(NULL, M_ERROR_TERM, 0, _("Out of reload table entries. Giving up.\n"));
+ } else {
+ Jmsg(NULL, M_ERROR, 0, _("Please correct configuration file: %s\n"), configfile);
+ }
+ reload_table[rtable].res_table = save_config_resources();
+ /* Now restore old resoure values */
+ int num = r_last - r_first + 1;
+ RES **res_tab = reload_table[table].res_table;
+ for (int i=0; i<num; i++) {
+ res_head[i] = res_tab[i];
+ }
+ table = rtable; /* release new, bad, saved table below */
+ } else {
+ /*
+ * Hook all active jobs so that they release this table
+ */
+ foreach_jcr(jcr) {
+ if (jcr->JobType != JT_SYSTEM) {
+ reload_table[table].job_count++;
+ job_end_push(jcr, reload_job_end_cb, (void *)((long int)table));
+ njobs++;
+ }
+ free_locked_jcr(jcr);
+ }
}
/* Reset globals */
set_working_directory(director->working_directory);
FDConnectTimeout = director->FDConnectTimeout;
SDConnectTimeout = director->SDConnectTimeout;
-
+ Dmsg0(0, "Director's configuration file reread.\n");
+
+ /* Now release saved resources, if no jobs using the resources */
+ if (njobs == 0) {
+ free_saved_resources(table);
+ }
+
+bail_out:
+ UnlockRes();
+ unlock_jcr_chain();
sigprocmask(SIG_UNBLOCK, &set, NULL);
signal(SIGHUP, reload_config);
- already_here = FALSE;
- Dmsg0(0, "Director's configuration file reread.\n");
+ already_here = false;
}
/*
*/
static int check_resources()
{
- int OK = TRUE;
+ bool OK = true;
JOB *job;
LockRes();
- job = (JOB *)GetNextRes(R_JOB, NULL);
+ job = (JOB *)GetNextRes(R_JOB, NULL);
director = (DIRRES *)GetNextRes(R_DIRECTOR, NULL);
if (!director) {
Jmsg(NULL, M_FATAL, 0, _("No Director resource defined in %s\n\
Without that I don't know who I am :-(\n"), configfile);
- OK = FALSE;
+ OK = false;
} else {
set_working_directory(director->working_directory);
if (!director->messages) { /* If message resource not specified */
director->messages = (MSGS *)GetNextRes(R_MSGS, NULL);
if (!director->messages) {
Jmsg(NULL, M_FATAL, 0, _("No Messages resource defined in %s\n"), configfile);
- OK = FALSE;
+ OK = false;
}
}
if (GetNextRes(R_DIRECTOR, (RES *)director) != NULL) {
Jmsg(NULL, M_FATAL, 0, _("Only one Director resource permitted in %s\n"),
configfile);
- OK = FALSE;
+ OK = false;
}
}
if (!job) {
Jmsg(NULL, M_FATAL, 0, _("No Job records defined in %s\n"), configfile);
- OK = FALSE;
+ OK = false;
}
- for (job=NULL; (job = (JOB *)GetNextRes(R_JOB, (RES *)job)); ) {
- if (!job->client) {
- Jmsg(NULL, M_FATAL, 0, _("No Client record defined for job %s\n"), job->hdr.name);
- OK = FALSE;
- }
- if (!job->fileset) {
- Jmsg(NULL, M_FATAL, 0, _("No FileSet record defined for job %s\n"), job->hdr.name);
- OK = FALSE;
- }
- if (!job->storage && job->JobType != JT_VERIFY) {
- Jmsg(NULL, M_FATAL, 0, _("No Storage resource defined for job %s\n"), job->hdr.name);
- OK = FALSE;
+ foreach_res(job, R_JOB) {
+ int i;
+
+ if (job->jobdefs) {
+ /* Transfer default items from JobDefs Resource */
+ for (i=0; job_items[i].name; i++) {
+ char **def_svalue, **svalue; /* string value */
+ int *def_ivalue, *ivalue; /* integer value */
+ int64_t *def_lvalue, *lvalue; /* 64 bit values */
+ uint32_t offset;
+
+ Dmsg4(400, "Job \"%s\", field \"%s\" bit=%d def=%d\n",
+ job->hdr.name, job_items[i].name,
+ bit_is_set(i, job->hdr.item_present),
+ bit_is_set(i, job->jobdefs->hdr.item_present));
+
+ if (!bit_is_set(i, job->hdr.item_present) &&
+ bit_is_set(i, job->jobdefs->hdr.item_present)) {
+ Dmsg2(400, "Job \"%s\", field \"%s\": getting default.\n",
+ job->hdr.name, job_items[i].name);
+ offset = (char *)(job_items[i].value) - (char *)&res_all;
+ /*
+ * Handle strings and directory strings
+ */
+ if (job_items[i].handler == store_str ||
+ job_items[i].handler == store_dir) {
+ def_svalue = (char **)((char *)(job->jobdefs) + offset);
+ Dmsg5(400, "Job \"%s\", field \"%s\" def_svalue=%s item %d offset=%u\n",
+ job->hdr.name, job_items[i].name, *def_svalue, i, offset);
+ svalue = (char **)((char *)job + offset);
+ if (*svalue) {
+ Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
+ }
+ *svalue = bstrdup(*def_svalue);
+ set_bit(i, job->hdr.item_present);
+ /*
+ * Handle resources
+ */
+ } else if (job_items[i].handler == store_res) {
+ def_svalue = (char **)((char *)(job->jobdefs) + offset);
+ Dmsg4(400, "Job \"%s\", field \"%s\" item %d offset=%u\n",
+ job->hdr.name, job_items[i].name, i, offset);
+ svalue = (char **)((char *)job + offset);
+ if (*svalue) {
+ Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
+ }
+ *svalue = *def_svalue;
+ set_bit(i, job->hdr.item_present);
+ /*
+ * Handle alist resources
+ */
+ } else if (job_items[i].handler == store_alist_res) {
+ int count = job_items[i].default_value;
+ def_svalue = (char **)((char *)(job->jobdefs) + offset);
+ Dmsg4(400, "Job \"%s\", field \"%s\" item %d offset=%u\n",
+ job->hdr.name, job_items[i].name, i, offset);
+ svalue = (char **)((char *)job + offset);
+ if (*svalue) {
+ Pmsg1(000, "Hey something is wrong. p=0x%lu\n", *svalue);
+ }
+ while (count--) {
+ *svalue++ = *def_svalue++;
+ }
+ set_bit(i, job->hdr.item_present);
+ /*
+ * Handle integer fields
+ * Note, our store_yesno does not handle bitmaped fields
+ */
+ } else if (job_items[i].handler == store_yesno ||
+ job_items[i].handler == store_pint ||
+ job_items[i].handler == store_jobtype ||
+ job_items[i].handler == store_level ||
+ job_items[i].handler == store_pint ||
+ job_items[i].handler == store_replace) {
+ def_ivalue = (int *)((char *)(job->jobdefs) + offset);
+ Dmsg5(400, "Job \"%s\", field \"%s\" def_ivalue=%d item %d offset=%u\n",
+ job->hdr.name, job_items[i].name, *def_ivalue, i, offset);
+ ivalue = (int *)((char *)job + offset);
+ *ivalue = *def_ivalue;
+ set_bit(i, job->hdr.item_present);
+ /*
+ * Handle 64 bit integer fields
+ */
+ } else if (job_items[i].handler == store_time ||
+ job_items[i].handler == store_size ||
+ job_items[i].handler == store_int64) {
+ def_lvalue = (int64_t *)((char *)(job->jobdefs) + offset);
+ Dmsg5(400, "Job \"%s\", field \"%s\" def_lvalue=%" lld " item %d offset=%u\n",
+ job->hdr.name, job_items[i].name, *def_lvalue, i, offset);
+ lvalue = (int64_t *)((char *)job + offset);
+ *lvalue = *def_lvalue;
+ set_bit(i, job->hdr.item_present);
+ }
+ }
+ }
}
- if (!job->pool) {
- Jmsg(NULL, M_FATAL, 0, _("No Pool resource defined for job %s\n"), job->hdr.name);
- OK = FALSE;
+ /*
+ * Ensure that all required items are present
+ */
+ for (i=0; job_items[i].name; i++) {
+ if (job_items[i].flags & ITEM_REQUIRED) {
+ if (!bit_is_set(i, job->hdr.item_present)) {
+ Jmsg(NULL, M_FATAL, 0, "\"%s\" directive in Job \"%s\" resource is required, but not found.\n",
+ job_items[i].name, job->hdr.name);
+ OK = false;
+ }
+ }
+ /* If this triggers, take a look at lib/parse_conf.h */
+ if (i >= MAX_RES_ITEMS) {
+ Emsg0(M_ERROR_TERM, 0, "Too many items in Job resource\n");
+ }
}
- if (job->client && job->client->catalog) {
- CAT *catalog = job->client->catalog;
- B_DB *db;
-
- /*
- * Make sure we can open catalog, otherwise print a warning
- * message because the server is probably not running.
- */
- db = db_init_database(NULL, catalog->db_name, catalog->db_user,
- catalog->db_password, catalog->db_address,
- catalog->db_port, catalog->db_socket);
- if (!db_open_database(NULL, db)) {
- OK = FALSE; /* Error message already printed */
- } else {
- /* If a pool is defined for this job, create the pool DB
- * record if it is not already created.
- */
- if (job->pool) {
- create_pool(NULL, db, job->pool, POOL_OP_UPDATE); /* update request */
- }
- db_close_database(NULL, db);
+ } /* End loop over Job res */
+
+ /* Loop over databases */
+ CAT *catalog;
+ foreach_res(catalog, R_CATALOG) {
+ B_DB *db;
+ /*
+ * Make sure we can open catalog, otherwise print a warning
+ * message because the server is probably not running.
+ */
+ db = db_init_database(NULL, catalog->db_name, catalog->db_user,
+ catalog->db_password, catalog->db_address,
+ catalog->db_port, catalog->db_socket,
+ catalog->mult_db_connections);
+ if (!db || !db_open_database(NULL, db)) {
+ Jmsg(NULL, M_FATAL, 0, _("Could not open database \"%s\".\n"),
+ catalog->db_name);
+ if (db) {
+ Jmsg(NULL, M_FATAL, 0, _("%s"), db_strerror(db));
}
+ OK = false;
+ continue;
+ }
- } else {
- if (job->client) {
- Jmsg(NULL, M_FATAL, 0, _("No Catalog resource defined for client %s\n"),
- job->client->hdr.name);
- OK = FALSE;
+ /* Loop over all pools, defining/updating them in each database */
+ POOL *pool;
+ foreach_res(pool, R_POOL) {
+ create_pool(NULL, db, pool, POOL_OP_UPDATE); /* update request */
+ }
+ /* Loop over all counters, defining them in each database */
+
+ /* Set default value in all counters */
+ COUNTER *counter;
+ foreach_res(counter, R_COUNTER) {
+ /* Write to catalog? */
+ if (!counter->created && counter->Catalog == catalog) {
+ COUNTER_DBR cr;
+ bstrncpy(cr.Counter, counter->hdr.name, sizeof(cr.Counter));
+ cr.MinValue = counter->MinValue;
+ cr.MaxValue = counter->MaxValue;
+ cr.CurrentValue = counter->MinValue;
+ if (counter->WrapCounter) {
+ bstrncpy(cr.WrapCounter, counter->WrapCounter->hdr.name, sizeof(cr.WrapCounter));
+ } else {
+ cr.WrapCounter[0] = 0; /* empty string */
+ }
+ if (db_create_counter_record(NULL, db, &cr)) {
+ counter->CurrentValue = cr.CurrentValue;
+ counter->created = true;
+ Dmsg2(100, "Create counter %s val=%d\n", counter->hdr.name, counter->CurrentValue);
+ }
+ }
+ if (!counter->created) {
+ counter->CurrentValue = counter->MinValue; /* default value */
}
}
+ db_close_database(NULL, db);
}
UnlockRes();