2 Bacula® - The Network Backup Solution
4 Copyright (C) 2004-2008 Free Software Foundation Europe e.V.
6 The main author of Bacula is Kern Sibbald, with contributions from
7 many others, a complete list can be found in the file AUTHORS.
8 This program is Free Software; you can redistribute it and/or
9 modify it under the terms of version two of the GNU General Public
10 License as published by the Free Software Foundation and included
13 This program is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; if not, write to the Free Software
20 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
23 Bacula® is a registered trademark of Kern Sibbald.
24 The licensor of Bacula is the Free Software Foundation Europe
25 (FSFE), Fiduciary Program, Sumatrastrasse 25, 8006 Zürich,
26 Switzerland, email:ftf@fsfeurope.org.
30 * Bacula Director -- migrate.c -- responsible for doing
31 * migration and copy jobs.
33 * Also handles Copy jobs (March MMVIII)
35 * Kern Sibbald, September MMIV
37 * Basic tasks done here:
38 * Open DB and create records for this job.
39 * Open Message Channel with Storage daemon to tell him a job will be starting.
40 * Open connection with Storage daemon and pass him commands
42 * When the Storage daemon finishes the job, update the DB.
51 #include "lib/bregex.h"
56 static const int dbglevel = 10;
58 static char OKbootstrap[] = "3000 OK bootstrap\n";
59 static int get_job_to_migrate(JCR *jcr);
61 static bool regex_find_jobids(JCR *jcr, idpkt *ids, const char *query1,
62 const char *query2, const char *type);
63 static bool find_mediaid_then_jobids(JCR *jcr, idpkt *ids, const char *query1,
65 static bool find_jobids_from_mediaid_list(JCR *jcr, idpkt *ids, const char *type);
66 static void start_migration_job(JCR *jcr);
67 static int get_next_dbid_from_list(char **p, DBId_t *DBId);
70 * Called here before the job is run to do the job
71 * specific setup. Note, one of the important things to
72 * complete in this init code is to make the definitive
73 * choice of input and output storage devices. This is
74 * because immediately after the init, the job is queued
75 * in the jobq.c code, and it checks that all the resources
76 * (storage resources in particular) are available, so these
77 * must all be properly defined.
79 * previous_jr refers to the job DB record of the Job that is
80 * going to be migrated.
81 * prev_job refers to the job resource of the Job that is
82 * going to be migrated.
83 * jcr is the jcr for the current "migration" job. It is a
84 * control job that is put in the DB as a migration job, which
85 * means that this job migrated a previous job to a new job.
86 * No Volume or File data is associated with this control
88 * mig_jcr refers to the newly migrated job that is run by
89 * the current jcr. It is a backup job that moves (migrates) the
90 * data written for the previous_jr into the new pool. This
91 * job (mig_jcr) becomes the new backup job that replaces
92 * the original backup job. Note, this jcr is not really run. It
93 * is simply attached to the current jcr. It will show up in
94 * the Director's status output, but not in the SD or FD, both of
95 * which deal only with the current migration job (i.e. jcr).
97 bool do_migration_init(JCR *jcr)
103 JCR *mig_jcr; /* newly migrated job */
107 apply_pool_overrides(jcr);
109 if (!allow_duplicate_job(jcr)) {
113 jcr->jr.PoolId = get_or_create_pool_record(jcr, jcr->pool->name());
114 if (jcr->jr.PoolId == 0) {
115 Dmsg1(dbglevel, "JobId=%d no PoolId\n", (int)jcr->JobId);
116 Jmsg(jcr, M_FATAL, 0, _("Could not get or create a Pool record.\n"));
120 * Note, at this point, pool is the pool for this job. We
121 * transfer it to rpool (read pool), and a bit later,
122 * pool will be changed to point to the write pool,
123 * which comes from pool->NextPool.
125 jcr->rpool = jcr->pool; /* save read pool */
126 pm_strcpy(jcr->rpool_source, jcr->pool_source);
129 Dmsg2(dbglevel, "Read pool=%s (From %s)\n", jcr->rpool->name(), jcr->rpool_source);
131 /* If we find a job or jobs to migrate it is previous_jr.JobId */
132 count = get_job_to_migrate(jcr);
140 Dmsg1(dbglevel, "Back from get_job_to_migrate JobId=%d\n", (int)jcr->JobId);
142 if (jcr->previous_jr.JobId == 0) {
143 Dmsg1(dbglevel, "JobId=%d no previous JobId\n", (int)jcr->JobId);
144 Jmsg(jcr, M_INFO, 0, _("No previous Job found to migrate.\n"));
145 return true; /* no work */
148 if (!get_or_create_fileset_record(jcr)) {
149 Dmsg1(dbglevel, "JobId=%d no FileSet\n", (int)jcr->JobId);
150 Jmsg(jcr, M_FATAL, 0, _("Could not get or create the FileSet record.\n"));
154 create_restore_bootstrap_file(jcr);
156 if (jcr->previous_jr.JobId == 0 || jcr->ExpectedFiles == 0) {
157 set_jcr_job_status(jcr, JS_Terminated);
158 Dmsg1(dbglevel, "JobId=%d expected files == 0\n", (int)jcr->JobId);
159 if (jcr->previous_jr.JobId == 0) {
160 Jmsg(jcr, M_INFO, 0, _("No previous Job found to migrate.\n"));
162 Jmsg(jcr, M_INFO, 0, _("Previous Job has no data to migrate.\n"));
164 return true; /* no work */
168 Dmsg5(dbglevel, "JobId=%d: Current: Name=%s JobId=%d Type=%c Level=%c\n",
170 jcr->jr.Name, (int)jcr->jr.JobId,
171 jcr->jr.JobType, jcr->jr.JobLevel);
174 job = (JOB *)GetResWithName(R_JOB, jcr->jr.Name);
175 prev_job = (JOB *)GetResWithName(R_JOB, jcr->previous_jr.Name);
178 Jmsg(jcr, M_FATAL, 0, _("Job resource not found for \"%s\".\n"), jcr->jr.Name);
182 Jmsg(jcr, M_FATAL, 0, _("Previous Job resource not found for \"%s\".\n"),
183 jcr->previous_jr.Name);
187 jcr->spool_data = job->spool_data; /* turn on spooling if requested in job */
189 /* Create a migation jcr */
190 mig_jcr = jcr->mig_jcr = new_jcr(sizeof(JCR), dird_free_jcr);
191 memcpy(&mig_jcr->previous_jr, &jcr->previous_jr, sizeof(mig_jcr->previous_jr));
194 * Turn the mig_jcr into a "real" job that takes on the aspects of
195 * the previous backup job "prev_job".
197 set_jcr_defaults(mig_jcr, prev_job);
198 if (!setup_job(mig_jcr)) {
199 Jmsg(jcr, M_FATAL, 0, _("setup job failed.\n"));
203 /* Now reset the job record from the previous job */
204 memcpy(&mig_jcr->jr, &jcr->previous_jr, sizeof(mig_jcr->jr));
205 /* Update the jr to reflect the new values of PoolId and JobId. */
206 mig_jcr->jr.PoolId = jcr->jr.PoolId;
207 mig_jcr->jr.JobId = mig_jcr->JobId;
209 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
210 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
211 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
214 * Get the PoolId used with the original job. Then
215 * find the pool name from the database record.
217 memset(&pr, 0, sizeof(pr));
218 pr.PoolId = mig_jcr->previous_jr.PoolId;
219 if (!db_get_pool_record(jcr, jcr->db, &pr)) {
220 Jmsg(jcr, M_FATAL, 0, _("Pool for JobId %s not in database. ERR=%s\n"),
221 edit_int64(pr.PoolId, ed1), db_strerror(jcr->db));
224 /* Get the pool resource corresponding to the original job */
225 pool = (POOL *)GetResWithName(R_POOL, pr.Name);
227 Jmsg(jcr, M_FATAL, 0, _("Pool resource \"%s\" not found.\n"), pr.Name);
231 /* If pool storage specified, use it for restore */
232 copy_rstorage(mig_jcr, pool->storage, _("Pool resource"));
233 copy_rstorage(jcr, pool->storage, _("Pool resource"));
236 * If the original backup pool has a NextPool, make sure a
237 * record exists in the database. Note, in this case, we
238 * will be migrating from pool to pool->NextPool.
240 if (pool->NextPool) {
241 jcr->jr.PoolId = get_or_create_pool_record(jcr, pool->NextPool->name());
242 if (jcr->jr.PoolId == 0) {
246 if (!set_migration_wstorage(jcr, pool)) {
249 mig_jcr->pool = jcr->pool = pool->NextPool;
250 pm_strcpy(jcr->pool_source, _("Job Pool's NextPool resource"));
251 mig_jcr->jr.PoolId = jcr->jr.PoolId;
253 Dmsg2(dbglevel, "Write pool=%s read rpool=%s\n", jcr->pool->name(), jcr->rpool->name());
258 * Do a Migration of a previous job
260 * Returns: false on failure
263 bool do_migration(JCR *jcr)
267 JCR *mig_jcr = jcr->mig_jcr; /* newly migrated job */
270 * If mig_jcr is NULL, there is nothing to do for this job,
271 * so set a normal status, cleanup and return OK.
274 set_jcr_job_status(jcr, JS_Terminated);
275 migration_cleanup(jcr, jcr->JobStatus);
279 if (!db_get_job_record(jcr, jcr->db, &jcr->previous_jr)) {
280 Jmsg(jcr, M_FATAL, 0, _("Could not get job record for JobId %s to migrate. ERR=%s"),
281 edit_int64(jcr->previous_jr.JobId, ed1),
282 db_strerror(jcr->db));
283 set_jcr_job_status(jcr, JS_Terminated);
284 migration_cleanup(jcr, jcr->JobStatus);
287 /* Make sure this job was not already migrated */
288 if (jcr->previous_jr.JobType != JT_BACKUP) {
289 Jmsg(jcr, M_INFO, 0, _("JobId %s already migrated probably by another Job. Migration stopped.\n"),
290 edit_int64(jcr->previous_jr.JobId, ed1));
291 set_jcr_job_status(jcr, JS_Terminated);
292 migration_cleanup(jcr, jcr->JobStatus);
296 /* Print Job Start message */
297 Jmsg(jcr, M_INFO, 0, _("Start %s JobId %s, Job=%s\n"),
298 jcr->get_JobType() == JT_MIGRATE ? "Migration" : "Copy",
299 edit_uint64(jcr->JobId, ed1), jcr->Job);
302 * Open a message channel connection with the Storage
303 * daemon. This is to let him know that our client
304 * will be contacting him for a backup session.
307 Dmsg0(110, "Open connection with storage daemon\n");
308 set_jcr_job_status(jcr, JS_WaitSD);
309 set_jcr_job_status(mig_jcr, JS_WaitSD);
311 * Start conversation with Storage daemon
313 if (!connect_to_storage_daemon(jcr, 10, SDConnectTimeout, 1)) {
316 sd = jcr->store_bsock;
318 * Now start a job with the Storage daemon
320 Dmsg2(dbglevel, "Read store=%s, write store=%s\n",
321 ((STORE *)jcr->rstorage->first())->name(),
322 ((STORE *)jcr->wstorage->first())->name());
323 if (((STORE *)jcr->rstorage->first())->name() == ((STORE *)jcr->wstorage->first())->name()) {
324 Jmsg(jcr, M_FATAL, 0, _("Read storage \"%s\" same as write storage.\n"),
325 ((STORE *)jcr->rstorage->first())->name());
328 if (!start_storage_daemon_job(jcr, jcr->rstorage, jcr->wstorage)) {
331 Dmsg0(150, "Storage daemon connection OK\n");
333 if (!send_bootstrap_file(jcr, sd) ||
334 !response(jcr, sd, OKbootstrap, "Bootstrap", DISPLAY_ERROR)) {
339 * We re-update the job start record so that the start
340 * time is set after the run before job. This avoids
341 * that any files created by the run before job will
342 * be saved twice. They will be backed up in the current
343 * job, but not in the next one unless they are changed.
344 * Without this, they will be backed up in this job and
345 * in the next job run because in that case, their date
346 * is after the start of this run.
348 jcr->start_time = time(NULL);
349 jcr->jr.StartTime = jcr->start_time;
350 jcr->jr.JobTDate = jcr->start_time;
351 set_jcr_job_status(jcr, JS_Running);
353 /* Update job start record for this migration control job */
354 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
355 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
360 mig_jcr->start_time = time(NULL);
361 mig_jcr->jr.StartTime = mig_jcr->start_time;
362 mig_jcr->jr.JobTDate = mig_jcr->start_time;
363 set_jcr_job_status(mig_jcr, JS_Running);
365 /* Update job start record for the real migration backup job */
366 if (!db_update_job_start_record(mig_jcr, mig_jcr->db, &mig_jcr->jr)) {
367 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(mig_jcr->db));
371 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
372 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
373 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
377 * Start the job prior to starting the message thread below
378 * to avoid two threads from using the BSOCK structure at
381 if (!sd->fsend("run")) {
386 * Now start a Storage daemon message thread
388 if (!start_storage_daemon_message_thread(jcr)) {
393 set_jcr_job_status(jcr, JS_Running);
394 set_jcr_job_status(mig_jcr, JS_Running);
396 /* Pickup Job termination data */
397 /* Note, the SD stores in jcr->JobFiles/ReadBytes/JobBytes/Errors */
398 wait_for_storage_daemon_termination(jcr);
399 set_jcr_job_status(jcr, jcr->SDJobStatus);
400 db_write_batch_file_records(jcr); /* used by bulk batch file insert */
401 if (jcr->JobStatus != JS_Terminated) {
405 migration_cleanup(jcr, jcr->JobStatus);
406 if (jcr->get_JobType() == JT_MIGRATE && mig_jcr) {
408 UAContext *ua = new_ua_context(jcr);
409 edit_uint64(jcr->previous_jr.JobId, jobid);
410 /* Purge all old file records, but leave Job record */
411 purge_files_from_jobs(ua, jobid);
422 /* Add an item to the list if it is unique */
423 static void add_unique_id(idpkt *ids, char *item)
425 const int maxlen = 30;
429 /* Walk through current list to see if each item is the same as item */
432 for (int i=0; i<maxlen; i++) {
435 } else if (*q == ',') {
442 if (strcmp(item, id) == 0) {
446 /* Did not find item, so add it to list */
447 if (ids->count == 0) {
450 pm_strcat(ids->list, ",");
452 pm_strcat(ids->list, item);
454 // Dmsg3(0, "add_uniq count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
459 * Callback handler make list of DB Ids
461 static int unique_dbid_handler(void *ctx, int num_fields, char **row)
463 idpkt *ids = (idpkt *)ctx;
465 add_unique_id(ids, row[0]);
466 Dmsg3(dbglevel, "dbid_hdlr count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
476 static int item_compare(void *item1, void *item2)
478 uitem *i1 = (uitem *)item1;
479 uitem *i2 = (uitem *)item2;
480 return strcmp(i1->item, i2->item);
483 static int unique_name_handler(void *ctx, int num_fields, char **row)
485 dlist *list = (dlist *)ctx;
487 uitem *new_item = (uitem *)malloc(sizeof(uitem));
490 memset(new_item, 0, sizeof(uitem));
491 new_item->item = bstrdup(row[0]);
492 Dmsg1(dbglevel, "Unique_name_hdlr Item=%s\n", row[0]);
493 item = (uitem *)list->binary_insert((void *)new_item, item_compare);
494 if (item != new_item) { /* already in list */
495 free(new_item->item);
496 free((char *)new_item);
502 /* Get Job names in Pool */
503 const char *sql_job =
504 "SELECT DISTINCT Job.Name from Job,Pool"
505 " WHERE Pool.Name='%s' AND Job.PoolId=Pool.PoolId";
507 /* Get JobIds from regex'ed Job names */
508 const char *sql_jobids_from_job =
509 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool"
510 " WHERE Job.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
511 " ORDER by Job.StartTime";
513 /* Get Client names in Pool */
514 const char *sql_client =
515 "SELECT DISTINCT Client.Name from Client,Pool,Job"
516 " WHERE Pool.Name='%s' AND Job.ClientId=Client.ClientId AND"
517 " Job.PoolId=Pool.PoolId";
519 /* Get JobIds from regex'ed Client names */
520 const char *sql_jobids_from_client =
521 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool,Client"
522 " WHERE Client.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
523 " AND Job.ClientId=Client.ClientId AND Job.Type='B'"
524 " ORDER by Job.StartTime";
526 /* Get Volume names in Pool */
527 const char *sql_vol =
528 "SELECT DISTINCT VolumeName FROM Media,Pool WHERE"
529 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
530 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
532 /* Get JobIds from regex'ed Volume names */
533 const char *sql_jobids_from_vol =
534 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Media,JobMedia,Job"
535 " WHERE Media.VolumeName='%s' AND Media.MediaId=JobMedia.MediaId"
536 " AND JobMedia.JobId=Job.JobId AND Job.Type='B'"
537 " ORDER by Job.StartTime";
540 const char *sql_smallest_vol =
541 "SELECT Media.MediaId FROM Media,Pool,JobMedia WHERE"
542 " Media.MediaId in (SELECT DISTINCT MediaId from JobMedia) AND"
543 " Media.VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
544 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
545 " ORDER BY VolBytes ASC LIMIT 1";
547 const char *sql_oldest_vol =
548 "SELECT Media.MediaId FROM Media,Pool,JobMedia WHERE"
549 " Media.MediaId in (SELECT DISTINCT MediaId from JobMedia) AND"
550 " Media.VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
551 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
552 " ORDER BY LastWritten ASC LIMIT 1";
554 /* Get JobIds when we have selected MediaId */
555 const char *sql_jobids_from_mediaid =
556 "SELECT DISTINCT Job.JobId,Job.StartTime FROM JobMedia,Job"
557 " WHERE JobMedia.JobId=Job.JobId AND JobMedia.MediaId IN (%s)"
559 " ORDER by Job.StartTime";
561 /* Get the number of bytes in the pool */
562 const char *sql_pool_bytes =
563 "SELECT SUM(VolBytes) FROM Media,Pool WHERE"
564 " VolStatus in ('Full','Used','Error','Append') AND Media.Enabled=1 AND"
565 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
567 /* Get the number of bytes in the Jobs */
568 const char *sql_job_bytes =
569 "SELECT SUM(JobBytes) FROM Job WHERE JobId IN (%s)";
572 /* Get Media Ids in Pool */
573 const char *sql_mediaids =
574 "SELECT MediaId FROM Media,Pool WHERE"
575 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
576 " Media.PoolId=Pool.PoolId AND Pool.Name='%s' ORDER BY LastWritten ASC";
578 /* Get JobIds in Pool longer than specified time */
579 const char *sql_pool_time =
580 "SELECT DISTINCT Job.JobId from Pool,Job,Media,JobMedia WHERE"
581 " Pool.Name='%s' AND Media.PoolId=Pool.PoolId AND"
582 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
584 " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId"
585 " AND Job.RealEndTime<='%s'";
588 * const char *sql_ujobid =
589 * "SELECT DISTINCT Job.Job from Client,Pool,Media,Job,JobMedia "
590 * " WHERE Media.PoolId=Pool.PoolId AND Pool.Name='%s' AND"
591 * " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId";
598 * This is the central piece of code that finds a job or jobs
599 * actually JobIds to migrate. It first looks to see if one
600 * has been "manually" specified in jcr->MigrateJobId, and if
601 * so, it returns that JobId to be run. Otherwise, it
602 * examines the Selection Type to see what kind of migration
603 * we are doing (Volume, Job, Client, ...) and applies any
604 * Selection Pattern if appropriate to obtain a list of JobIds.
605 * Finally, it will loop over all the JobIds found, except the last
606 * one starting a new job with MigrationJobId set to that JobId, and
607 * finally, it returns the last JobId to the caller.
609 * Returns: -1 on error
610 * 0 if no jobs to migrate
611 * 1 if OK and jcr->previous_jr filled in
613 static int get_job_to_migrate(JCR *jcr)
615 char ed1[30], ed2[30];
616 POOL_MEM query(PM_MESSAGE);
621 idpkt ids, mid, jids;
626 char dt[MAX_TIME_LENGTH];
629 ids.list = get_pool_memory(PM_MESSAGE);
632 mid.list = get_pool_memory(PM_MESSAGE);
635 jids.list = get_pool_memory(PM_MESSAGE);
641 * If MigrateJobId is set, then we migrate only that Job,
642 * otherwise, we go through the full selection of jobs to
645 if (jcr->MigrateJobId != 0) {
646 Dmsg1(dbglevel, "At Job start previous jobid=%u\n", jcr->MigrateJobId);
647 edit_uint64(jcr->MigrateJobId, ids.list);
650 switch (jcr->job->selection_type) {
652 if (!regex_find_jobids(jcr, &ids, sql_job, sql_jobids_from_job, "Job")) {
657 if (!regex_find_jobids(jcr, &ids, sql_client, sql_jobids_from_client, "Client")) {
662 if (!regex_find_jobids(jcr, &ids, sql_vol, sql_jobids_from_vol, "Volume")) {
667 if (!jcr->job->selection_pattern) {
668 Jmsg(jcr, M_FATAL, 0, _("No Migration SQL selection pattern specified.\n"));
671 Dmsg1(dbglevel, "SQL=%s\n", jcr->job->selection_pattern);
672 if (!db_sql_query(jcr->db, jcr->job->selection_pattern,
673 unique_dbid_handler, (void *)&ids)) {
674 Jmsg(jcr, M_FATAL, 0,
675 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
679 case MT_SMALLEST_VOL:
680 if (!find_mediaid_then_jobids(jcr, &ids, sql_smallest_vol, "Smallest Volume")) {
685 if (!find_mediaid_then_jobids(jcr, &ids, sql_oldest_vol, "Oldest Volume")) {
690 case MT_POOL_OCCUPANCY:
692 /* Find count of bytes in pool */
693 Mmsg(query, sql_pool_bytes, jcr->rpool->name());
694 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
695 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
698 if (ctx.count == 0) {
699 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
702 pool_bytes = ctx.value;
703 Dmsg2(dbglevel, "highbytes=%lld pool=%lld\n", jcr->rpool->MigrationHighBytes,
705 if (pool_bytes < (int64_t)jcr->rpool->MigrationHighBytes) {
706 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
709 Dmsg0(dbglevel, "We should do Occupation migration.\n");
712 /* Find a list of MediaIds that could be migrated */
713 Mmsg(query, sql_mediaids, jcr->rpool->name());
714 Dmsg1(dbglevel, "query=%s\n", query.c_str());
715 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
716 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
719 if (ids.count == 0) {
720 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
723 Dmsg2(dbglevel, "Pool Occupancy ids=%d MediaIds=%s\n", ids.count, ids.list);
725 if (!find_jobids_from_mediaid_list(jcr, &ids, "Volumes")) {
728 /* ids == list of jobs */
730 for (int i=0; i < (int)ids.count; i++) {
731 stat = get_next_dbid_from_list(&p, &DBId);
732 Dmsg2(dbglevel, "get_next_dbid stat=%d JobId=%u\n", stat, (uint32_t)DBId);
734 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
736 } else if (stat == 0) {
741 Mmsg(mid.list, "%s", edit_int64(DBId, ed1));
742 if (jids.count > 0) {
743 pm_strcat(jids.list, ",");
745 pm_strcat(jids.list, mid.list);
746 jids.count += mid.count;
748 /* Find count of bytes from Jobs */
749 Mmsg(query, sql_job_bytes, mid.list);
750 Dmsg1(dbglevel, "Jobbytes query: %s\n", query.c_str());
751 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
752 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
755 pool_bytes -= ctx.value;
756 Dmsg1(dbglevel, "Total migrate Job bytes=%s\n", edit_int64_with_commas(ctx.value, ed1));
757 Dmsg2(dbglevel, "lowbytes=%s poolafter=%s\n",
758 edit_int64_with_commas(jcr->rpool->MigrationLowBytes, ed1),
759 edit_int64_with_commas(pool_bytes, ed2));
760 if (pool_bytes <= (int64_t)jcr->rpool->MigrationLowBytes) {
761 Dmsg0(dbglevel, "We should be done.\n");
765 /* Transfer jids to ids, where the jobs list is expected */
766 ids.count = jids.count;
767 pm_strcpy(ids.list, jids.list);
768 Dmsg2(dbglevel, "Pool Occupancy ids=%d JobIds=%s\n", ids.count, ids.list);
772 ttime = time(NULL) - (time_t)jcr->rpool->MigrationTime;
773 (void)localtime_r(&ttime, &tm);
774 strftime(dt, sizeof(dt), "%Y-%m-%d %H:%M:%S", &tm);
777 Mmsg(query, sql_pool_time, jcr->rpool->name(), dt);
778 Dmsg1(dbglevel, "query=%s\n", query.c_str());
779 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
780 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
783 if (ids.count == 0) {
784 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
787 Dmsg2(dbglevel, "PoolTime ids=%d JobIds=%s\n", ids.count, ids.list);
791 Jmsg(jcr, M_FATAL, 0, _("Unknown Migration Selection Type.\n"));
797 * Loop over all jobids except the last one, sending
798 * them to start_migration_job(), which will start a job
799 * for each of them. For the last JobId, we handle it below.
802 if (ids.count == 0) {
803 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
807 Jmsg(jcr, M_INFO, 0, _("The following %u JobId%s were chosen to be migrated: %s\n"),
808 ids.count, ids.count==0?"":"s", ids.list);
810 Dmsg2(dbglevel, "Before loop count=%d ids=%s\n", ids.count, ids.list);
811 for (int i=1; i < (int)ids.count; i++) {
813 stat = get_next_jobid_from_list(&p, &JobId);
814 Dmsg3(dbglevel, "get_jobid_no=%d stat=%d JobId=%u\n", i, stat, JobId);
815 jcr->MigrateJobId = JobId;
816 start_migration_job(jcr);
817 Dmsg0(dbglevel, "Back from start_migration_job\n");
819 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
821 } else if (stat == 0) {
822 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
827 /* Now get the last JobId and handle it in the current job */
829 stat = get_next_jobid_from_list(&p, &JobId);
830 Dmsg2(dbglevel, "Last get_next_jobid stat=%d JobId=%u\n", stat, (int)JobId);
832 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
834 } else if (stat == 0) {
835 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
839 jcr->previous_jr.JobId = JobId;
840 Dmsg1(dbglevel, "Previous jobid=%d\n", (int)jcr->previous_jr.JobId);
842 if (!db_get_job_record(jcr, jcr->db, &jcr->previous_jr)) {
843 Jmsg(jcr, M_FATAL, 0, _("Could not get job record for JobId %s to migrate. ERR=%s"),
844 edit_int64(jcr->previous_jr.JobId, ed1),
845 db_strerror(jcr->db));
848 Jmsg(jcr, M_INFO, 0, _("Migration using JobId=%s Job=%s\n"),
849 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
850 Dmsg3(dbglevel, "Migration JobId=%d using JobId=%s Job=%s\n",
852 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
862 free_pool_memory(ids.list);
863 free_pool_memory(mid.list);
864 free_pool_memory(jids.list);
868 static void start_migration_job(JCR *jcr)
870 UAContext *ua = new_ua_context(jcr);
873 Mmsg(ua->cmd, "run %s jobid=%s", jcr->job->hdr.name,
874 edit_uint64(jcr->MigrateJobId, ed1));
875 Dmsg1(dbglevel, "=============== Migration cmd=%s\n", ua->cmd);
876 parse_ua_args(ua); /* parse command */
877 JobId_t jobid = run_cmd(ua, ua->cmd);
879 Jmsg(jcr, M_ERROR, 0, _("Could not start migration job.\n"));
881 Jmsg(jcr, M_INFO, 0, _("Migration JobId %d started.\n"), (int)jobid);
886 static bool find_mediaid_then_jobids(JCR *jcr, idpkt *ids, const char *query1,
890 POOL_MEM query(PM_MESSAGE);
893 /* Basic query for MediaId */
894 Mmsg(query, query1, jcr->rpool->name());
895 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
896 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
899 if (ids->count == 0) {
900 Jmsg(jcr, M_INFO, 0, _("No %s found to migrate.\n"), type);
901 ok = true; /* Not an error */
903 } else if (ids->count != 1) {
904 Jmsg(jcr, M_FATAL, 0, _("SQL error. Expected 1 MediaId got %d\n"), ids->count);
907 Dmsg2(dbglevel, "%s MediaIds=%s\n", type, ids->list);
909 ok = find_jobids_from_mediaid_list(jcr, ids, type);
916 * This routine returns:
917 * false if an error occurred
919 * ids.count number of jobids found (may be zero)
921 static bool find_jobids_from_mediaid_list(JCR *jcr, idpkt *ids, const char *type)
924 POOL_MEM query(PM_MESSAGE);
926 Mmsg(query, sql_jobids_from_mediaid, ids->list);
928 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
929 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
932 if (ids->count == 0) {
933 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
941 static bool regex_find_jobids(JCR *jcr, idpkt *ids, const char *query1,
942 const char *query2, const char *type)
946 uitem *last_item = NULL;
951 POOL_MEM query(PM_MESSAGE);
953 item_chain = New(dlist(item, &item->link));
954 if (!jcr->job->selection_pattern) {
955 Jmsg(jcr, M_FATAL, 0, _("No Migration %s selection pattern specified.\n"),
959 Dmsg1(dbglevel, "regex-sel-pattern=%s\n", jcr->job->selection_pattern);
960 /* Basic query for names */
961 Mmsg(query, query1, jcr->rpool->name());
962 Dmsg1(dbglevel, "get name query1=%s\n", query.c_str());
963 if (!db_sql_query(jcr->db, query.c_str(), unique_name_handler,
964 (void *)item_chain)) {
965 Jmsg(jcr, M_FATAL, 0,
966 _("SQL to get %s failed. ERR=%s\n"), type, db_strerror(jcr->db));
969 Dmsg1(dbglevel, "query1 returned %d names\n", item_chain->size());
970 if (item_chain->size() == 0) {
971 Jmsg(jcr, M_INFO, 0, _("Query of Pool \"%s\" returned no Jobs to migrate.\n"),
974 goto bail_out; /* skip regex match */
976 /* Compile regex expression */
977 rc = regcomp(&preg, jcr->job->selection_pattern, REG_EXTENDED);
979 regerror(rc, &preg, prbuf, sizeof(prbuf));
980 Jmsg(jcr, M_FATAL, 0, _("Could not compile regex pattern \"%s\" ERR=%s\n"),
981 jcr->job->selection_pattern, prbuf);
984 /* Now apply the regex to the names and remove any item not matched */
985 foreach_dlist(item, item_chain) {
986 const int nmatch = 30;
987 regmatch_t pmatch[nmatch];
989 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
990 free(last_item->item);
991 item_chain->remove(last_item);
993 Dmsg1(dbglevel, "get name Item=%s\n", item->item);
994 rc = regexec(&preg, item->item, nmatch, pmatch, 0);
996 last_item = NULL; /* keep this one */
1002 free(last_item->item);
1003 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
1004 item_chain->remove(last_item);
1008 if (item_chain->size() == 0) {
1009 Jmsg(jcr, M_INFO, 0, _("Regex pattern matched no Jobs to migrate.\n"));
1011 goto bail_out; /* skip regex match */
1015 * At this point, we have a list of items in item_chain
1016 * that have been matched by the regex, so now we need
1017 * to look up their jobids.
1020 foreach_dlist(item, item_chain) {
1021 Dmsg2(dbglevel, "Got %s: %s\n", type, item->item);
1022 Mmsg(query, query2, item->item, jcr->rpool->name());
1023 Dmsg1(dbglevel, "get id from name query2=%s\n", query.c_str());
1024 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
1025 Jmsg(jcr, M_FATAL, 0,
1026 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
1030 if (ids->count == 0) {
1031 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
1036 Dmsg2(dbglevel, "Count=%d Jobids=%s\n", ids->count, ids->list);
1037 foreach_dlist(item, item_chain) {
1046 * Release resources allocated during backup.
1048 void migration_cleanup(JCR *jcr, int TermCode)
1050 char sdt[MAX_TIME_LENGTH], edt[MAX_TIME_LENGTH];
1051 char ec1[30], ec2[30], ec3[30], ec4[30], ec5[30], elapsed[50];
1052 char ec6[50], ec7[50], ec8[50];
1053 char term_code[100], sd_term_msg[100];
1054 const char *term_msg;
1055 int msg_type = M_INFO;
1059 JCR *mig_jcr = jcr->mig_jcr;
1060 POOL_MEM query(PM_MESSAGE);
1062 Dmsg2(100, "Enter migrate_cleanup %d %c\n", TermCode, TermCode);
1063 update_job_end(jcr, TermCode);
1064 memset(&mr, 0, sizeof(mr));
1067 * Check if we actually did something.
1068 * mig_jcr is jcr of the newly migrated job.
1071 mig_jcr->JobFiles = jcr->JobFiles = jcr->SDJobFiles;
1072 mig_jcr->JobBytes = jcr->JobBytes = jcr->SDJobBytes;
1073 mig_jcr->VolSessionId = jcr->VolSessionId;
1074 mig_jcr->VolSessionTime = jcr->VolSessionTime;
1075 mig_jcr->jr.RealEndTime = 0;
1076 mig_jcr->jr.PriorJobId = jcr->previous_jr.JobId;
1078 update_job_end(mig_jcr, TermCode);
1080 /* Update final items to set them to the previous job's values */
1081 Mmsg(query, "UPDATE Job SET StartTime='%s',EndTime='%s',"
1082 "JobTDate=%s WHERE JobId=%s",
1083 jcr->previous_jr.cStartTime, jcr->previous_jr.cEndTime,
1084 edit_uint64(jcr->previous_jr.JobTDate, ec1),
1085 edit_uint64(mig_jcr->jr.JobId, ec2));
1086 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1088 /* Now mark the previous job as migrated if it terminated normally */
1089 if (jcr->get_JobType() == JT_MIGRATE && jcr->JobStatus == JS_Terminated) {
1090 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1091 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1092 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1095 if (!db_get_job_record(jcr, jcr->db, &jcr->jr)) {
1096 Jmsg(jcr, M_WARNING, 0, _("Error getting Job record for Job report: ERR=%s"),
1097 db_strerror(jcr->db));
1098 set_jcr_job_status(jcr, JS_ErrorTerminated);
1101 update_bootstrap_file(mig_jcr);
1103 if (!db_get_job_volume_names(mig_jcr, mig_jcr->db, mig_jcr->jr.JobId, &mig_jcr->VolumeName)) {
1105 * Note, if the job has failed, most likely it did not write any
1106 * tape, so suppress this "error" message since in that case
1107 * it is normal. Or look at it the other way, only for a
1108 * normal exit should we complain about this error.
1110 if (jcr->JobStatus == JS_Terminated && jcr->jr.JobBytes) {
1111 Jmsg(jcr, M_ERROR, 0, "%s", db_strerror(mig_jcr->db));
1113 mig_jcr->VolumeName[0] = 0; /* none */
1116 if (mig_jcr->VolumeName[0]) {
1117 /* Find last volume name. Multiple vols are separated by | */
1118 char *p = strrchr(mig_jcr->VolumeName, '|');
1122 p = mig_jcr->VolumeName; /* no |, take full name */
1124 bstrncpy(mr.VolumeName, p, sizeof(mr.VolumeName));
1125 if (!db_get_media_record(jcr, jcr->db, &mr)) {
1126 Jmsg(jcr, M_WARNING, 0, _("Error getting Media record for Volume \"%s\": ERR=%s"),
1127 mr.VolumeName, db_strerror(jcr->db));
1131 switch (jcr->JobStatus) {
1133 if (jcr->Errors || jcr->SDErrors) {
1134 term_msg = _("%s OK -- with warnings");
1136 term_msg = _("%s OK");
1140 case JS_ErrorTerminated:
1141 term_msg = _("*** %s Error ***");
1142 msg_type = M_ERROR; /* Generate error message */
1143 if (jcr->store_bsock) {
1144 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1145 if (jcr->SD_msg_chan) {
1146 pthread_cancel(jcr->SD_msg_chan);
1151 term_msg = _("%s Canceled");
1152 if (jcr->store_bsock) {
1153 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1154 if (jcr->SD_msg_chan) {
1155 pthread_cancel(jcr->SD_msg_chan);
1160 term_msg = _("Inappropriate %s term code");
1164 if (jcr->get_JobType() == JT_MIGRATE && jcr->previous_jr.JobId != 0) {
1165 /* Mark previous job as migrated */
1166 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1167 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1168 db_sql_query(jcr->db, query.c_str(), NULL, NULL);
1170 term_msg = _("%s -- no files to migrate");
1173 bsnprintf(term_code, sizeof(term_code), term_msg, "Migration");
1174 bstrftimes(sdt, sizeof(sdt), jcr->jr.StartTime);
1175 bstrftimes(edt, sizeof(edt), jcr->jr.EndTime);
1176 RunTime = jcr->jr.EndTime - jcr->jr.StartTime;
1180 kbps = (double)jcr->SDJobBytes / (1000 * RunTime);
1184 jobstatus_to_ascii(jcr->SDJobStatus, sd_term_msg, sizeof(sd_term_msg));
1186 Jmsg(jcr, msg_type, 0, _("Bacula %s %s (%s): %s\n"
1187 " Build OS: %s %s %s\n"
1188 " Prev Backup JobId: %s\n"
1189 " New Backup JobId: %s\n"
1190 " Current JobId: %s\n"
1191 " Current Job: %s\n"
1192 " Backup Level: %s%s\n"
1194 " FileSet: \"%s\" %s\n"
1195 " Read Pool: \"%s\" (From %s)\n"
1196 " Read Storage: \"%s\" (From %s)\n"
1197 " Write Pool: \"%s\" (From %s)\n"
1198 " Write Storage: \"%s\" (From %s)\n"
1199 " Catalog: \"%s\" (From %s)\n"
1202 " Elapsed time: %s\n"
1204 " SD Files Written: %s\n"
1205 " SD Bytes Written: %s (%sB)\n"
1206 " Rate: %.1f KB/s\n"
1207 " Volume name(s): %s\n"
1208 " Volume Session Id: %d\n"
1209 " Volume Session Time: %d\n"
1210 " Last Volume Bytes: %s (%sB)\n"
1212 " SD termination status: %s\n"
1213 " Termination: %s\n\n"),
1214 my_name, VERSION, LSMDATE, edt,
1215 HOST_OS, DISTNAME, DISTVER,
1216 edit_uint64(jcr->previous_jr.JobId, ec6),
1217 mig_jcr ? edit_uint64(mig_jcr->jr.JobId, ec7) : "0",
1218 edit_uint64(jcr->jr.JobId, ec8),
1220 level_to_str(jcr->get_JobLevel()), jcr->since,
1221 jcr->client->name(),
1222 jcr->fileset->name(), jcr->FSCreateTime,
1223 jcr->rpool->name(), jcr->rpool_source,
1224 jcr->rstore?jcr->rstore->name():"*None*",
1225 NPRT(jcr->rstore_source),
1226 jcr->pool->name(), jcr->pool_source,
1227 jcr->wstore?jcr->wstore->name():"*None*",
1228 NPRT(jcr->wstore_source),
1229 jcr->catalog->name(), jcr->catalog_source,
1232 edit_utime(RunTime, elapsed, sizeof(elapsed)),
1234 edit_uint64_with_commas(jcr->SDJobFiles, ec1),
1235 edit_uint64_with_commas(jcr->SDJobBytes, ec2),
1236 edit_uint64_with_suffix(jcr->SDJobBytes, ec3),
1238 mig_jcr ? mig_jcr->VolumeName : "",
1240 jcr->VolSessionTime,
1241 edit_uint64_with_commas(mr.VolBytes, ec4),
1242 edit_uint64_with_suffix(mr.VolBytes, ec5),
1247 Dmsg1(100, "migrate_cleanup() mig_jcr=0x%x\n", jcr->mig_jcr);
1249 free_jcr(jcr->mig_jcr);
1250 jcr->mig_jcr = NULL;
1252 Dmsg0(100, "Leave migrate_cleanup()\n");
1256 * Return next DBId from comma separated list
1259 * 1 if next DBId returned
1260 * 0 if no more DBIds are in list
1261 * -1 there is an error
1263 static int get_next_dbid_from_list(char **p, DBId_t *DBId)
1265 const int maxlen = 30;
1270 for (int i=0; i<maxlen; i++) {
1273 } else if (*q == ',') {
1282 } else if (!is_a_number(id)) {
1283 return -1; /* error */
1286 *DBId = str_to_int64(id);
1290 bool set_migration_wstorage(JCR *jcr, POOL *pool)
1292 POOL *wpool = pool->NextPool;
1295 Jmsg(jcr, M_FATAL, 0, _("No Next Pool specification found in Pool \"%s\".\n"),
1300 if (!wpool->storage || wpool->storage->size() == 0) {
1301 Jmsg(jcr, M_FATAL, 0, _("No Storage specification found in Next Pool \"%s\".\n"),
1306 /* If pool storage specified, use it instead of job storage for backup */
1307 copy_wstorage(jcr, wpool->storage, _("Storage from Pool's NextPool resource"));