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 John Walker.
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
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 */
167 Dmsg5(dbglevel, "JobId=%d: Previous: Name=%s JobId=%d Type=%c Level=%c\n",
169 jcr->previous_jr.Name, (int)jcr->previous_jr.JobId,
170 jcr->previous_jr.JobType, jcr->previous_jr.JobLevel);
172 Dmsg5(dbglevel, "JobId=%d: Current: Name=%s JobId=%d Type=%c Level=%c\n",
174 jcr->jr.Name, (int)jcr->jr.JobId,
175 jcr->jr.JobType, jcr->jr.JobLevel);
178 job = (JOB *)GetResWithName(R_JOB, jcr->jr.Name);
179 prev_job = (JOB *)GetResWithName(R_JOB, jcr->previous_jr.Name);
182 Jmsg(jcr, M_FATAL, 0, _("Job resource not found for \"%s\".\n"), jcr->jr.Name);
186 Jmsg(jcr, M_FATAL, 0, _("Previous Job resource not found for \"%s\".\n"),
187 jcr->previous_jr.Name);
191 jcr->spool_data = job->spool_data; /* turn on spooling if requested in job */
193 /* Create a migation jcr */
194 mig_jcr = jcr->mig_jcr = new_jcr(sizeof(JCR), dird_free_jcr);
195 memcpy(&mig_jcr->previous_jr, &jcr->previous_jr, sizeof(mig_jcr->previous_jr));
198 * Turn the mig_jcr into a "real" job that takes on the aspects of
199 * the previous backup job "prev_job".
201 set_jcr_defaults(mig_jcr, prev_job);
202 if (!setup_job(mig_jcr)) {
203 Jmsg(jcr, M_FATAL, 0, _("setup job failed.\n"));
207 /* Now reset the job record from the previous job */
208 memcpy(&mig_jcr->jr, &jcr->previous_jr, sizeof(mig_jcr->jr));
209 /* Update the jr to reflect the new values of PoolId and JobId. */
210 mig_jcr->jr.PoolId = jcr->jr.PoolId;
211 mig_jcr->jr.JobId = mig_jcr->JobId;
213 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
214 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
215 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
218 * Get the PoolId used with the original job. Then
219 * find the pool name from the database record.
221 memset(&pr, 0, sizeof(pr));
222 pr.PoolId = mig_jcr->previous_jr.PoolId;
223 if (!db_get_pool_record(jcr, jcr->db, &pr)) {
224 Jmsg(jcr, M_FATAL, 0, _("Pool for JobId %s not in database. ERR=%s\n"),
225 edit_int64(pr.PoolId, ed1), db_strerror(jcr->db));
228 /* Get the pool resource corresponding to the original job */
229 pool = (POOL *)GetResWithName(R_POOL, pr.Name);
231 Jmsg(jcr, M_FATAL, 0, _("Pool resource \"%s\" not found.\n"), pr.Name);
235 /* If pool storage specified, use it for restore */
236 copy_rstorage(mig_jcr, pool->storage, _("Pool resource"));
237 copy_rstorage(jcr, pool->storage, _("Pool resource"));
240 * If the original backup pool has a NextPool, make sure a
241 * record exists in the database. Note, in this case, we
242 * will be migrating from pool to pool->NextPool.
244 if (pool->NextPool) {
245 jcr->jr.PoolId = get_or_create_pool_record(jcr, pool->NextPool->name());
246 if (jcr->jr.PoolId == 0) {
250 if (!set_migration_wstorage(jcr, pool)) {
253 mig_jcr->pool = jcr->pool = pool->NextPool;
254 pm_strcpy(jcr->pool_source, _("Job Pool's NextPool resource"));
255 mig_jcr->jr.PoolId = jcr->jr.PoolId;
257 Dmsg2(dbglevel, "Write pool=%s read rpool=%s\n", jcr->pool->name(), jcr->rpool->name());
262 * Do a Migration of a previous job
264 * Returns: false on failure
267 bool do_migration(JCR *jcr)
271 JCR *mig_jcr = jcr->mig_jcr; /* newly migrated job */
274 * If mig_jcr is NULL, there is nothing to do for this job,
275 * so set a normal status, cleanup and return OK.
278 set_jcr_job_status(jcr, JS_Terminated);
279 migration_cleanup(jcr, jcr->JobStatus);
283 /* Print Job Start message */
284 Jmsg(jcr, M_INFO, 0, _("Start %s JobId %s, Job=%s\n"),
285 jcr->JobType == JT_MIGRATE ? "Migration" : "Copy",
286 edit_uint64(jcr->JobId, ed1), jcr->Job);
291 * Open a message channel connection with the Storage
292 * daemon. This is to let him know that our client
293 * will be contacting him for a backup session.
296 Dmsg0(110, "Open connection with storage daemon\n");
297 set_jcr_job_status(jcr, JS_WaitSD);
298 set_jcr_job_status(mig_jcr, JS_WaitSD);
300 * Start conversation with Storage daemon
302 if (!connect_to_storage_daemon(jcr, 10, SDConnectTimeout, 1)) {
305 sd = jcr->store_bsock;
307 * Now start a job with the Storage daemon
309 Dmsg2(dbglevel, "Read store=%s, write store=%s\n",
310 ((STORE *)jcr->rstorage->first())->name(),
311 ((STORE *)jcr->wstorage->first())->name());
312 if (((STORE *)jcr->rstorage->first())->name() == ((STORE *)jcr->wstorage->first())->name()) {
313 Jmsg(jcr, M_FATAL, 0, _("Read storage \"%s\" same as write storage.\n"),
314 ((STORE *)jcr->rstorage->first())->name());
317 if (!start_storage_daemon_job(jcr, jcr->rstorage, jcr->wstorage)) {
320 Dmsg0(150, "Storage daemon connection OK\n");
322 if (!send_bootstrap_file(jcr, sd) ||
323 !response(jcr, sd, OKbootstrap, "Bootstrap", DISPLAY_ERROR)) {
328 * We re-update the job start record so that the start
329 * time is set after the run before job. This avoids
330 * that any files created by the run before job will
331 * be saved twice. They will be backed up in the current
332 * job, but not in the next one unless they are changed.
333 * Without this, they will be backed up in this job and
334 * in the next job run because in that case, their date
335 * is after the start of this run.
337 jcr->start_time = time(NULL);
338 jcr->jr.StartTime = jcr->start_time;
339 jcr->jr.JobTDate = jcr->start_time;
340 set_jcr_job_status(jcr, JS_Running);
342 /* Update job start record for this migration control job */
343 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
344 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
349 mig_jcr->start_time = time(NULL);
350 mig_jcr->jr.StartTime = mig_jcr->start_time;
351 mig_jcr->jr.JobTDate = mig_jcr->start_time;
352 set_jcr_job_status(mig_jcr, JS_Running);
354 /* Update job start record for the real migration backup job */
355 if (!db_update_job_start_record(mig_jcr, mig_jcr->db, &mig_jcr->jr)) {
356 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(mig_jcr->db));
360 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
361 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
362 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
366 * Start the job prior to starting the message thread below
367 * to avoid two threads from using the BSOCK structure at
370 if (!sd->fsend("run")) {
375 * Now start a Storage daemon message thread
377 if (!start_storage_daemon_message_thread(jcr)) {
382 set_jcr_job_status(jcr, JS_Running);
383 set_jcr_job_status(mig_jcr, JS_Running);
385 /* Pickup Job termination data */
386 /* Note, the SD stores in jcr->JobFiles/ReadBytes/JobBytes/Errors */
387 wait_for_storage_daemon_termination(jcr);
388 set_jcr_job_status(jcr, jcr->SDJobStatus);
389 db_write_batch_file_records(jcr); /* used by bulk batch file insert */
390 if (jcr->JobStatus != JS_Terminated) {
394 migration_cleanup(jcr, jcr->JobStatus);
395 if (jcr->JobType == JT_MIGRATE && mig_jcr) {
397 UAContext *ua = new_ua_context(jcr);
398 edit_uint64(jcr->previous_jr.JobId, jobid);
399 /* Purge all old file records, but leave Job record */
400 purge_files_from_jobs(ua, jobid);
411 /* Add an item to the list if it is unique */
412 static void add_unique_id(idpkt *ids, char *item)
417 /* Walk through current list to see if each item is the same as item */
420 for (int i=0; i<(int)sizeof(id); i++) {
423 } else if (*q == ',') {
430 if (strcmp(item, id) == 0) {
434 /* Did not find item, so add it to list */
435 if (ids->count == 0) {
438 pm_strcat(ids->list, ",");
440 pm_strcat(ids->list, item);
442 // Dmsg3(0, "add_uniq count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
447 * Callback handler make list of DB Ids
449 static int unique_dbid_handler(void *ctx, int num_fields, char **row)
451 idpkt *ids = (idpkt *)ctx;
453 add_unique_id(ids, row[0]);
454 Dmsg3(dbglevel, "dbid_hdlr count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
464 static int item_compare(void *item1, void *item2)
466 uitem *i1 = (uitem *)item1;
467 uitem *i2 = (uitem *)item2;
468 return strcmp(i1->item, i2->item);
471 static int unique_name_handler(void *ctx, int num_fields, char **row)
473 dlist *list = (dlist *)ctx;
475 uitem *new_item = (uitem *)malloc(sizeof(uitem));
478 memset(new_item, 0, sizeof(uitem));
479 new_item->item = bstrdup(row[0]);
480 Dmsg1(dbglevel, "Unique_name_hdlr Item=%s\n", row[0]);
481 item = (uitem *)list->binary_insert((void *)new_item, item_compare);
482 if (item != new_item) { /* already in list */
483 free(new_item->item);
484 free((char *)new_item);
490 /* Get Job names in Pool */
491 const char *sql_job =
492 "SELECT DISTINCT Job.Name from Job,Pool"
493 " WHERE Pool.Name='%s' AND Job.PoolId=Pool.PoolId";
495 /* Get JobIds from regex'ed Job names */
496 const char *sql_jobids_from_job =
497 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool"
498 " WHERE Job.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
499 " ORDER by Job.StartTime";
501 /* Get Client names in Pool */
502 const char *sql_client =
503 "SELECT DISTINCT Client.Name from Client,Pool,Job"
504 " WHERE Pool.Name='%s' AND Job.ClientId=Client.ClientId AND"
505 " Job.PoolId=Pool.PoolId";
507 /* Get JobIds from regex'ed Client names */
508 const char *sql_jobids_from_client =
509 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool,Client"
510 " WHERE Client.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
511 " AND Job.ClientId=Client.ClientId AND Job.Type='B'"
512 " ORDER by Job.StartTime";
514 /* Get Volume names in Pool */
515 const char *sql_vol =
516 "SELECT DISTINCT VolumeName FROM Media,Pool WHERE"
517 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
518 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
520 /* Get JobIds from regex'ed Volume names */
521 const char *sql_jobids_from_vol =
522 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Media,JobMedia,Job"
523 " WHERE Media.VolumeName='%s' AND Media.MediaId=JobMedia.MediaId"
524 " AND JobMedia.JobId=Job.JobId AND Job.Type='B'"
525 " ORDER by Job.StartTime";
528 const char *sql_smallest_vol =
529 "SELECT Media.MediaId FROM Media,Pool,JobMedia WHERE"
530 " Media.MediaId in (SELECT DISTINCT MediaId from JobMedia) AND"
531 " Media.VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
532 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
533 " ORDER BY VolBytes ASC LIMIT 1";
535 const char *sql_oldest_vol =
536 "SELECT Media.MediaId FROM Media,Pool,JobMedia WHERE"
537 " Media.MediaId in (SELECT DISTINCT MediaId from JobMedia) AND"
538 " Media.VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
539 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
540 " ORDER BY LastWritten ASC LIMIT 1";
542 /* Get JobIds when we have selected MediaId */
543 const char *sql_jobids_from_mediaid =
544 "SELECT DISTINCT Job.JobId,Job.StartTime FROM JobMedia,Job"
545 " WHERE JobMedia.JobId=Job.JobId AND JobMedia.MediaId=%s"
547 " ORDER by Job.StartTime";
549 /* Get tne number of bytes in the pool */
550 const char *sql_pool_bytes =
551 "SELECT SUM(VolBytes) FROM Media,Pool WHERE"
552 " VolStatus in ('Full','Used','Error','Append') AND Media.Enabled=1 AND"
553 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
555 /* Get tne number of bytes in the Jobs */
556 const char *sql_job_bytes =
557 "SELECT SUM(JobBytes) FROM Job WHERE JobId IN (%s)";
560 /* Get Media Ids in Pool */
561 const char *sql_mediaids =
562 "SELECT MediaId FROM Media,Pool WHERE"
563 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
564 " Media.PoolId=Pool.PoolId AND Pool.Name='%s' ORDER BY LastWritten ASC";
566 /* Get JobIds in Pool longer than specified time */
567 const char *sql_pool_time =
568 "SELECT DISTINCT Job.JobId from Pool,Job,Media,JobMedia WHERE"
569 " Pool.Name='%s' AND Media.PoolId=Pool.PoolId AND"
570 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
572 " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId"
573 " AND Job.RealEndTime<='%s'";
576 * const char *sql_ujobid =
577 * "SELECT DISTINCT Job.Job from Client,Pool,Media,Job,JobMedia "
578 * " WHERE Media.PoolId=Pool.PoolId AND Pool.Name='%s' AND"
579 * " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId";
586 * This is the central piece of code that finds a job or jobs
587 * actually JobIds to migrate. It first looks to see if one
588 * has been "manually" specified in jcr->MigrateJobId, and if
589 * so, it returns that JobId to be run. Otherwise, it
590 * examines the Selection Type to see what kind of migration
591 * we are doing (Volume, Job, Client, ...) and applies any
592 * Selection Pattern if appropriate to obtain a list of JobIds.
593 * Finally, it will loop over all the JobIds found, except the last
594 * one starting a new job with MigrationJobId set to that JobId, and
595 * finally, it returns the last JobId to the caller.
597 * Returns: -1 on error
598 * 0 if no jobs to migrate
599 * 1 if OK and jcr->previous_jr filled in
601 static int get_job_to_migrate(JCR *jcr)
603 char ed1[30], ed2[30];
604 POOL_MEM query(PM_MESSAGE);
609 idpkt ids, mid, jids;
614 char dt[MAX_TIME_LENGTH];
617 ids.list = get_pool_memory(PM_MESSAGE);
620 mid.list = get_pool_memory(PM_MESSAGE);
623 jids.list = get_pool_memory(PM_MESSAGE);
629 * If MigrateJobId is set, then we migrate only that Job,
630 * otherwise, we go through the full selection of jobs to
633 if (jcr->MigrateJobId != 0) {
634 Dmsg1(dbglevel, "At Job start previous jobid=%u\n", jcr->MigrateJobId);
635 edit_uint64(jcr->MigrateJobId, ids.list);
638 switch (jcr->job->selection_type) {
640 if (!regex_find_jobids(jcr, &ids, sql_job, sql_jobids_from_job, "Job")) {
645 if (!regex_find_jobids(jcr, &ids, sql_client, sql_jobids_from_client, "Client")) {
650 if (!regex_find_jobids(jcr, &ids, sql_vol, sql_jobids_from_vol, "Volume")) {
655 if (!jcr->job->selection_pattern) {
656 Jmsg(jcr, M_FATAL, 0, _("No Migration SQL selection pattern specified.\n"));
659 Dmsg1(dbglevel, "SQL=%s\n", jcr->job->selection_pattern);
660 if (!db_sql_query(jcr->db, jcr->job->selection_pattern,
661 unique_dbid_handler, (void *)&ids)) {
662 Jmsg(jcr, M_FATAL, 0,
663 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
667 case MT_SMALLEST_VOL:
668 if (!find_mediaid_then_jobids(jcr, &ids, sql_smallest_vol, "Smallest Volume")) {
673 if (!find_mediaid_then_jobids(jcr, &ids, sql_oldest_vol, "Oldest Volume")) {
678 case MT_POOL_OCCUPANCY:
680 /* Find count of bytes in pool */
681 Mmsg(query, sql_pool_bytes, jcr->rpool->name());
682 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
683 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
686 if (ctx.count == 0) {
687 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
690 pool_bytes = ctx.value;
691 Dmsg2(dbglevel, "highbytes=%lld pool=%lld\n", jcr->rpool->MigrationHighBytes,
693 if (pool_bytes < (int64_t)jcr->rpool->MigrationHighBytes) {
694 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
697 Dmsg0(dbglevel, "We should do Occupation migration.\n");
700 /* Find a list of MediaIds that could be migrated */
701 Mmsg(query, sql_mediaids, jcr->rpool->name());
702 Dmsg1(dbglevel, "query=%s\n", query.c_str());
703 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
704 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
707 if (ids.count == 0) {
708 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
711 Dmsg2(dbglevel, "Pool Occupancy ids=%d MediaIds=%s\n", ids.count, ids.list);
713 if (!find_jobids_from_mediaid_list(jcr, &ids, "Volumes")) {
716 /* ids == list of jobs */
718 for (int i=0; i < (int)ids.count; i++) {
719 stat = get_next_dbid_from_list(&p, &DBId);
720 Dmsg2(dbglevel, "get_next_dbid stat=%d JobId=%u\n", stat, (uint32_t)DBId);
722 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
724 } else if (stat == 0) {
729 Mmsg(mid.list, "%s", edit_int64(DBId, ed1));
730 if (jids.count > 0) {
731 pm_strcat(jids.list, ",");
733 pm_strcat(jids.list, mid.list);
734 jids.count += mid.count;
736 /* Find count of bytes from Jobs */
737 Mmsg(query, sql_job_bytes, mid.list);
738 Dmsg1(dbglevel, "Jobbytes query: %s\n", query.c_str());
739 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
740 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
743 pool_bytes -= ctx.value;
744 Dmsg1(dbglevel, "Total migrate Job bytes=%s\n", edit_int64(ctx.value, ed1));
745 Dmsg2(dbglevel, "lowbytes=%s poolafter=%s\n",
746 edit_int64(jcr->rpool->MigrationLowBytes, ed1),
747 edit_int64(pool_bytes, ed2));
748 if (pool_bytes <= (int64_t)jcr->rpool->MigrationLowBytes) {
749 Dmsg0(dbglevel, "We should be done.\n");
753 /* Transfer jids to ids, where the jobs list is expected */
754 ids.count = jids.count;
755 pm_strcpy(ids.list, jids.list);
756 Dmsg2(dbglevel, "Pool Occupancy ids=%d JobIds=%s\n", ids.count, ids.list);
760 ttime = time(NULL) - (time_t)jcr->rpool->MigrationTime;
761 (void)localtime_r(&ttime, &tm);
762 strftime(dt, sizeof(dt), "%Y-%m-%d %H:%M:%S", &tm);
765 Mmsg(query, sql_pool_time, jcr->rpool->name(), dt);
766 Dmsg1(dbglevel, "query=%s\n", query.c_str());
767 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
768 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
771 if (ids.count == 0) {
772 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
775 Dmsg2(dbglevel, "PoolTime ids=%d JobIds=%s\n", ids.count, ids.list);
779 Jmsg(jcr, M_FATAL, 0, _("Unknown Migration Selection Type.\n"));
785 * Loop over all jobids except the last one, sending
786 * them to start_migration_job(), which will start a job
787 * for each of them. For the last JobId, we handle it below.
790 if (ids.count == 0) {
791 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
795 Jmsg(jcr, M_INFO, 0, _("The following %u JobId%s were chosen to be migrated: %s\n"),
796 ids.count, ids.count==0?"":"s", ids.list);
798 Dmsg2(dbglevel, "Before loop count=%d ids=%s\n", ids.count, ids.list);
799 for (int i=1; i < (int)ids.count; i++) {
801 stat = get_next_jobid_from_list(&p, &JobId);
802 Dmsg3(dbglevel, "get_jobid_no=%d stat=%d JobId=%u\n", i, stat, JobId);
803 jcr->MigrateJobId = JobId;
804 start_migration_job(jcr);
805 Dmsg0(dbglevel, "Back from start_migration_job\n");
807 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
809 } else if (stat == 0) {
810 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
815 /* Now get the last JobId and handle it in the current job */
817 stat = get_next_jobid_from_list(&p, &JobId);
818 Dmsg2(dbglevel, "Last get_next_jobid stat=%d JobId=%u\n", stat, (int)JobId);
820 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
822 } else if (stat == 0) {
823 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
827 jcr->previous_jr.JobId = JobId;
828 Dmsg1(dbglevel, "Previous jobid=%d\n", (int)jcr->previous_jr.JobId);
830 if (!db_get_job_record(jcr, jcr->db, &jcr->previous_jr)) {
831 Jmsg(jcr, M_FATAL, 0, _("Could not get job record for JobId %s to migrate. ERR=%s"),
832 edit_int64(jcr->previous_jr.JobId, ed1),
833 db_strerror(jcr->db));
836 Jmsg(jcr, M_INFO, 0, _("Migration using JobId=%s Job=%s\n"),
837 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
838 Dmsg3(dbglevel, "Migration JobId=%d using JobId=%s Job=%s\n",
840 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
850 free_pool_memory(ids.list);
851 free_pool_memory(mid.list);
852 free_pool_memory(jids.list);
856 static void start_migration_job(JCR *jcr)
858 UAContext *ua = new_ua_context(jcr);
861 Mmsg(ua->cmd, "run %s jobid=%s", jcr->job->hdr.name,
862 edit_uint64(jcr->MigrateJobId, ed1));
863 Dmsg1(dbglevel, "=============== Migration cmd=%s\n", ua->cmd);
864 parse_ua_args(ua); /* parse command */
865 JobId_t jobid = run_cmd(ua, ua->cmd);
867 Jmsg(jcr, M_ERROR, 0, _("Could not start migration job.\n"));
869 Jmsg(jcr, M_INFO, 0, _("Migration JobId %d started.\n"), (int)jobid);
874 static bool find_mediaid_then_jobids(JCR *jcr, idpkt *ids, const char *query1,
878 POOL_MEM query(PM_MESSAGE);
881 /* Basic query for MediaId */
882 Mmsg(query, query1, jcr->rpool->name());
883 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
884 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
887 if (ids->count == 0) {
888 Jmsg(jcr, M_INFO, 0, _("No %s found to migrate.\n"), type);
889 ok = true; /* Not an error */
891 } else if (ids->count != 1) {
892 Jmsg(jcr, M_FATAL, 0, _("SQL error. Expected 1 MediaId got %d\n"), ids->count);
895 Dmsg2(dbglevel, "%s MediaIds=%s\n", type, ids->list);
897 ok = find_jobids_from_mediaid_list(jcr, ids, type);
904 * This routine returns:
905 * false if an error occurred
907 * ids.count number of jobids found (may be zero)
909 static bool find_jobids_from_mediaid_list(JCR *jcr, idpkt *ids, const char *type)
912 POOL_MEM query(PM_MESSAGE);
914 Mmsg(query, sql_jobids_from_mediaid, ids->list);
916 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
917 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
920 if (ids->count == 0) {
921 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
929 static bool regex_find_jobids(JCR *jcr, idpkt *ids, const char *query1,
930 const char *query2, const char *type)
934 uitem *last_item = NULL;
939 POOL_MEM query(PM_MESSAGE);
941 item_chain = New(dlist(item, &item->link));
942 if (!jcr->job->selection_pattern) {
943 Jmsg(jcr, M_FATAL, 0, _("No Migration %s selection pattern specified.\n"),
947 Dmsg1(dbglevel, "regex-sel-pattern=%s\n", jcr->job->selection_pattern);
948 /* Basic query for names */
949 Mmsg(query, query1, jcr->rpool->name());
950 Dmsg1(dbglevel, "get name query1=%s\n", query.c_str());
951 if (!db_sql_query(jcr->db, query.c_str(), unique_name_handler,
952 (void *)item_chain)) {
953 Jmsg(jcr, M_FATAL, 0,
954 _("SQL to get %s failed. ERR=%s\n"), type, db_strerror(jcr->db));
957 Dmsg1(dbglevel, "query1 returned %d names\n", item_chain->size());
958 if (item_chain->size() == 0) {
959 Jmsg(jcr, M_INFO, 0, _("Query of Pool \"%s\" returned no Jobs to migrate.\n"),
962 goto bail_out; /* skip regex match */
964 /* Compile regex expression */
965 rc = regcomp(&preg, jcr->job->selection_pattern, REG_EXTENDED);
967 regerror(rc, &preg, prbuf, sizeof(prbuf));
968 Jmsg(jcr, M_FATAL, 0, _("Could not compile regex pattern \"%s\" ERR=%s\n"),
969 jcr->job->selection_pattern, prbuf);
972 /* Now apply the regex to the names and remove any item not matched */
973 foreach_dlist(item, item_chain) {
974 const int nmatch = 30;
975 regmatch_t pmatch[nmatch];
977 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
978 free(last_item->item);
979 item_chain->remove(last_item);
981 Dmsg1(dbglevel, "get name Item=%s\n", item->item);
982 rc = regexec(&preg, item->item, nmatch, pmatch, 0);
984 last_item = NULL; /* keep this one */
990 free(last_item->item);
991 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
992 item_chain->remove(last_item);
996 if (item_chain->size() == 0) {
997 Jmsg(jcr, M_INFO, 0, _("Regex pattern matched no Jobs to migrate.\n"));
999 goto bail_out; /* skip regex match */
1003 * At this point, we have a list of items in item_chain
1004 * that have been matched by the regex, so now we need
1005 * to look up their jobids.
1008 foreach_dlist(item, item_chain) {
1009 Dmsg2(dbglevel, "Got %s: %s\n", type, item->item);
1010 Mmsg(query, query2, item->item, jcr->rpool->name());
1011 Dmsg1(dbglevel, "get id from name query2=%s\n", query.c_str());
1012 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
1013 Jmsg(jcr, M_FATAL, 0,
1014 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
1018 if (ids->count == 0) {
1019 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
1024 Dmsg2(dbglevel, "Count=%d Jobids=%s\n", ids->count, ids->list);
1025 foreach_dlist(item, item_chain) {
1034 * Release resources allocated during backup.
1036 void migration_cleanup(JCR *jcr, int TermCode)
1038 char sdt[MAX_TIME_LENGTH], edt[MAX_TIME_LENGTH];
1039 char ec1[30], ec2[30], ec3[30], ec4[30], ec5[30], elapsed[50];
1040 char ec6[50], ec7[50], ec8[50];
1041 char term_code[100], sd_term_msg[100];
1042 const char *term_msg;
1043 int msg_type = M_INFO;
1047 JCR *mig_jcr = jcr->mig_jcr;
1048 POOL_MEM query(PM_MESSAGE);
1050 Dmsg2(100, "Enter migrate_cleanup %d %c\n", TermCode, TermCode);
1051 update_job_end(jcr, TermCode);
1052 memset(&mr, 0, sizeof(mr));
1055 * Check if we actually did something.
1056 * mig_jcr is jcr of the newly migrated job.
1059 mig_jcr->JobFiles = jcr->JobFiles = jcr->SDJobFiles;
1060 mig_jcr->JobBytes = jcr->JobBytes = jcr->SDJobBytes;
1061 mig_jcr->VolSessionId = jcr->VolSessionId;
1062 mig_jcr->VolSessionTime = jcr->VolSessionTime;
1063 mig_jcr->jr.RealEndTime = 0;
1064 mig_jcr->jr.PriorJobId = jcr->previous_jr.JobId;
1066 update_job_end(mig_jcr, TermCode);
1068 /* Update final items to set them to the previous job's values */
1069 Mmsg(query, "UPDATE Job SET StartTime='%s',EndTime='%s',"
1070 "JobTDate=%s WHERE JobId=%s",
1071 jcr->previous_jr.cStartTime, jcr->previous_jr.cEndTime,
1072 edit_uint64(jcr->previous_jr.JobTDate, ec1),
1073 edit_uint64(mig_jcr->jr.JobId, ec2));
1074 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1076 /* Now mark the previous job as migrated if it terminated normally */
1077 if (jcr->JobType == JT_MIGRATE && jcr->JobStatus == JS_Terminated) {
1078 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1079 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1080 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1083 if (!db_get_job_record(jcr, jcr->db, &jcr->jr)) {
1084 Jmsg(jcr, M_WARNING, 0, _("Error getting Job record for Job report: ERR=%s"),
1085 db_strerror(jcr->db));
1086 set_jcr_job_status(jcr, JS_ErrorTerminated);
1089 update_bootstrap_file(mig_jcr);
1091 if (!db_get_job_volume_names(mig_jcr, mig_jcr->db, mig_jcr->jr.JobId, &mig_jcr->VolumeName)) {
1093 * Note, if the job has failed, most likely it did not write any
1094 * tape, so suppress this "error" message since in that case
1095 * it is normal. Or look at it the other way, only for a
1096 * normal exit should we complain about this error.
1098 if (jcr->JobStatus == JS_Terminated && jcr->jr.JobBytes) {
1099 Jmsg(jcr, M_ERROR, 0, "%s", db_strerror(mig_jcr->db));
1101 mig_jcr->VolumeName[0] = 0; /* none */
1104 if (mig_jcr->VolumeName[0]) {
1105 /* Find last volume name. Multiple vols are separated by | */
1106 char *p = strrchr(mig_jcr->VolumeName, '|');
1110 p = mig_jcr->VolumeName; /* no |, take full name */
1112 bstrncpy(mr.VolumeName, p, sizeof(mr.VolumeName));
1113 if (!db_get_media_record(jcr, jcr->db, &mr)) {
1114 Jmsg(jcr, M_WARNING, 0, _("Error getting Media record for Volume \"%s\": ERR=%s"),
1115 mr.VolumeName, db_strerror(jcr->db));
1119 switch (jcr->JobStatus) {
1121 if (jcr->Errors || jcr->SDErrors) {
1122 term_msg = _("%s OK -- with warnings");
1124 term_msg = _("%s OK");
1128 case JS_ErrorTerminated:
1129 term_msg = _("*** %s Error ***");
1130 msg_type = M_ERROR; /* Generate error message */
1131 if (jcr->store_bsock) {
1132 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1133 if (jcr->SD_msg_chan) {
1134 pthread_cancel(jcr->SD_msg_chan);
1139 term_msg = _("%s Canceled");
1140 if (jcr->store_bsock) {
1141 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1142 if (jcr->SD_msg_chan) {
1143 pthread_cancel(jcr->SD_msg_chan);
1148 term_msg = _("Inappropriate %s term code");
1152 if (jcr->JobType == JT_MIGRATE && jcr->previous_jr.JobId != 0) {
1153 /* Mark previous job as migrated */
1154 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1155 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1156 db_sql_query(jcr->db, query.c_str(), NULL, NULL);
1158 term_msg = _("%s -- no files to migrate");
1161 bsnprintf(term_code, sizeof(term_code), term_msg, "Migration");
1162 bstrftimes(sdt, sizeof(sdt), jcr->jr.StartTime);
1163 bstrftimes(edt, sizeof(edt), jcr->jr.EndTime);
1164 RunTime = jcr->jr.EndTime - jcr->jr.StartTime;
1168 kbps = (double)jcr->SDJobBytes / (1000 * RunTime);
1172 jobstatus_to_ascii(jcr->SDJobStatus, sd_term_msg, sizeof(sd_term_msg));
1174 Jmsg(jcr, msg_type, 0, _("Bacula %s %s (%s): %s\n"
1175 " Build OS: %s %s %s\n"
1176 " Prev Backup JobId: %s\n"
1177 " New Backup JobId: %s\n"
1178 " Current JobId: %s\n"
1179 " Current Job: %s\n"
1180 " Backup Level: %s%s\n"
1182 " FileSet: \"%s\" %s\n"
1183 " Read Pool: \"%s\" (From %s)\n"
1184 " Read Storage: \"%s\" (From %s)\n"
1185 " Write Pool: \"%s\" (From %s)\n"
1186 " Write Storage: \"%s\" (From %s)\n"
1187 " Catalog: \"%s\" (From %s)\n"
1190 " Elapsed time: %s\n"
1192 " SD Files Written: %s\n"
1193 " SD Bytes Written: %s (%sB)\n"
1194 " Rate: %.1f KB/s\n"
1195 " Volume name(s): %s\n"
1196 " Volume Session Id: %d\n"
1197 " Volume Session Time: %d\n"
1198 " Last Volume Bytes: %s (%sB)\n"
1200 " SD termination status: %s\n"
1201 " Termination: %s\n\n"),
1202 my_name, VERSION, LSMDATE, edt,
1203 HOST_OS, DISTNAME, DISTVER,
1204 edit_uint64(jcr->previous_jr.JobId, ec6),
1205 mig_jcr ? edit_uint64(mig_jcr->jr.JobId, ec7) : "0",
1206 edit_uint64(jcr->jr.JobId, ec8),
1208 level_to_str(jcr->JobLevel), jcr->since,
1209 jcr->client->name(),
1210 jcr->fileset->name(), jcr->FSCreateTime,
1211 jcr->rpool->name(), jcr->rpool_source,
1212 jcr->rstore?jcr->rstore->name():"*None*",
1213 NPRT(jcr->rstore_source),
1214 jcr->pool->name(), jcr->pool_source,
1215 jcr->wstore?jcr->wstore->name():"*None*",
1216 NPRT(jcr->wstore_source),
1217 jcr->catalog->name(), jcr->catalog_source,
1220 edit_utime(RunTime, elapsed, sizeof(elapsed)),
1222 edit_uint64_with_commas(jcr->SDJobFiles, ec1),
1223 edit_uint64_with_commas(jcr->SDJobBytes, ec2),
1224 edit_uint64_with_suffix(jcr->SDJobBytes, ec3),
1226 mig_jcr ? mig_jcr->VolumeName : "",
1228 jcr->VolSessionTime,
1229 edit_uint64_with_commas(mr.VolBytes, ec4),
1230 edit_uint64_with_suffix(mr.VolBytes, ec5),
1235 Dmsg1(100, "migrate_cleanup() mig_jcr=0x%x\n", jcr->mig_jcr);
1237 free_jcr(jcr->mig_jcr);
1238 jcr->mig_jcr = NULL;
1240 Dmsg0(100, "Leave migrate_cleanup()\n");
1244 * Return next DBId from comma separated list
1247 * 1 if next DBId returned
1248 * 0 if no more DBIds are in list
1249 * -1 there is an error
1251 static int get_next_dbid_from_list(char **p, DBId_t *DBId)
1257 for (int i=0; i<(int)sizeof(id); i++) {
1260 } else if (*q == ',') {
1269 } else if (!is_a_number(id)) {
1270 return -1; /* error */
1273 *DBId = str_to_int64(id);
1277 bool set_migration_wstorage(JCR *jcr, POOL *pool)
1279 POOL *wpool = pool->NextPool;
1282 Jmsg(jcr, M_FATAL, 0, _("No Next Pool specification found in Pool \"%s\".\n"),
1287 if (!wpool->storage || wpool->storage->size() == 0) {
1288 Jmsg(jcr, M_FATAL, 0, _("No Storage specification found in Next Pool \"%s\".\n"),
1293 /* If pool storage specified, use it instead of job storage for backup */
1294 copy_wstorage(jcr, wpool->storage, _("Storage from Pool's NextPool resource"));