3 * Bacula Director -- migrate.c -- responsible for doing
6 * Kern Sibbald, September MMIV
8 * Basic tasks done here:
9 * Open DB and create records for this job.
10 * Open Message Channel with Storage daemon to tell him a job will be starting.
11 * Open connection with Storage daemon and pass him commands
13 * When the Storage daemon finishes the job, update the DB.
18 Bacula® - The Network Backup Solution
20 Copyright (C) 2004-2006 Free Software Foundation Europe e.V.
22 The main author of Bacula is Kern Sibbald, with contributions from
23 many others, a complete list can be found in the file AUTHORS.
24 This program is Free Software; you can redistribute it and/or
25 modify it under the terms of version two of the GNU General Public
26 License as published by the Free Software Foundation plus additions
27 that are listed in the file LICENSE.
29 This program is distributed in the hope that it will be useful, but
30 WITHOUT ANY WARRANTY; without even the implied warranty of
31 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
32 General Public License for more details.
34 You should have received a copy of the GNU General Public License
35 along with this program; if not, write to the Free Software
36 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
39 Bacula® is a registered trademark of John Walker.
40 The licensor of Bacula is the Free Software Foundation Europe
41 (FSFE), Fiduciary Program, Sumatrastrasse 25, 8006 Zürich,
42 Switzerland, email:ftf@fsfeurope.org.
49 #include "lib/bregex.h"
54 static const int dbglevel = 10;
56 static char OKbootstrap[] = "3000 OK bootstrap\n";
57 static int get_job_to_migrate(JCR *jcr);
59 static bool regex_find_jobids(JCR *jcr, idpkt *ids, const char *query1,
60 const char *query2, const char *type);
61 static bool find_mediaid_then_jobids(JCR *jcr, idpkt *ids, const char *query1,
63 static bool find_jobids_from_mediaid_list(JCR *jcr, idpkt *ids, const char *type);
64 static void start_migration_job(JCR *jcr);
65 static int get_next_dbid_from_list(char **p, DBId_t *DBId);
68 * Called here before the job is run to do the job
69 * specific setup. Note, one of the important things to
70 * complete in this init code is to make the definitive
71 * choice of input and output storage devices. This is
72 * because immediately after the init, the job is queued
73 * in the jobq.c code, and it checks that all the resources
74 * (storage resources in particular) are available, so these
75 * must all be properly defined.
77 * previous_jr refers to the job DB record of the Job that is
78 * going to be migrated.
79 * prev_job refers to the job resource of the Job that is
80 * going to be migrated.
81 * jcr is the jcr for the current "migration" job. It is a
82 * control job that is put in the DB as a migration job, which
83 * means that this job migrated a previous job to a new job.
84 * No Volume or File data is associated with this control
86 * mig_jcr refers to the newly migrated job that is run by
87 * the current jcr. It is a backup job that moves (migrates) the
88 * data written for the previous_jr into the new pool. This
89 * job (mig_jcr) becomes the new backup job that replaces
90 * the original backup job. Note, this jcr is not really run. It
91 * is simply attached to the current jcr. It will show up in
92 * the Director's status output, but not in the SD or FD, both of
93 * which deal only with the current migration job (i.e. jcr).
95 bool do_migration_init(JCR *jcr)
101 JCR *mig_jcr; /* newly migrated job */
105 apply_pool_overrides(jcr);
107 jcr->jr.PoolId = get_or_create_pool_record(jcr, jcr->pool->name());
108 if (jcr->jr.PoolId == 0) {
109 Dmsg1(dbglevel, "JobId=%d no PoolId\n", (int)jcr->JobId);
110 Jmsg(jcr, M_FATAL, 0, _("Could not get or create a Pool record.\n"));
114 * Note, at this point, pool is the pool for this job. We
115 * transfer it to rpool (read pool), and a bit later,
116 * pool will be changed to point to the write pool,
117 * which comes from pool->NextPool.
119 jcr->rpool = jcr->pool; /* save read pool */
120 pm_strcpy(jcr->rpool_source, jcr->pool_source);
123 Dmsg2(dbglevel, "Read pool=%s (From %s)\n", jcr->rpool->name(), jcr->rpool_source);
125 /* If we find a job or jobs to migrate it is previous_jr.JobId */
126 count = get_job_to_migrate(jcr);
134 Dmsg1(dbglevel, "Back from get_job_to_migrate JobId=%d\n", (int)jcr->JobId);
136 if (jcr->previous_jr.JobId == 0) {
137 Dmsg1(dbglevel, "JobId=%d no previous JobId\n", (int)jcr->JobId);
138 Jmsg(jcr, M_INFO, 0, _("No previous Job found to migrate.\n"));
139 return true; /* no work */
142 if (!get_or_create_fileset_record(jcr)) {
143 Dmsg1(dbglevel, "JobId=%d no FileSet\n", (int)jcr->JobId);
144 Jmsg(jcr, M_FATAL, 0, _("Could not get or create the FileSet record.\n"));
148 create_restore_bootstrap_file(jcr);
150 if (jcr->previous_jr.JobId == 0 || jcr->ExpectedFiles == 0) {
151 set_jcr_job_status(jcr, JS_Terminated);
152 Dmsg1(dbglevel, "JobId=%d expected files == 0\n", (int)jcr->JobId);
153 if (jcr->previous_jr.JobId == 0) {
154 Jmsg(jcr, M_INFO, 0, _("No previous Job found to migrate.\n"));
156 Jmsg(jcr, M_INFO, 0, _("Previous Job has no data to migrate.\n"));
158 return true; /* no work */
161 Dmsg5(dbglevel, "JobId=%d: Previous: Name=%s JobId=%d Type=%c Level=%c\n",
163 jcr->previous_jr.Name, (int)jcr->previous_jr.JobId,
164 jcr->previous_jr.JobType, jcr->previous_jr.JobLevel);
166 Dmsg5(dbglevel, "JobId=%d: Current: Name=%s JobId=%d Type=%c Level=%c\n",
168 jcr->jr.Name, (int)jcr->jr.JobId,
169 jcr->jr.JobType, jcr->jr.JobLevel);
172 job = (JOB *)GetResWithName(R_JOB, jcr->jr.Name);
173 prev_job = (JOB *)GetResWithName(R_JOB, jcr->previous_jr.Name);
176 Jmsg(jcr, M_FATAL, 0, _("Job resource not found for \"%s\".\n"), jcr->jr.Name);
180 Jmsg(jcr, M_FATAL, 0, _("Previous Job resource not found for \"%s\".\n"),
181 jcr->previous_jr.Name);
185 /* Create a migation jcr */
186 mig_jcr = jcr->mig_jcr = new_jcr(sizeof(JCR), dird_free_jcr);
187 memcpy(&mig_jcr->previous_jr, &jcr->previous_jr, sizeof(mig_jcr->previous_jr));
190 * Turn the mig_jcr into a "real" job that takes on the aspects of
191 * the previous backup job "prev_job".
193 set_jcr_defaults(mig_jcr, prev_job);
194 if (!setup_job(mig_jcr)) {
195 Jmsg(jcr, M_FATAL, 0, _("setup job failed.\n"));
199 /* Now reset the job record from the previous job */
200 memcpy(&mig_jcr->jr, &jcr->previous_jr, sizeof(mig_jcr->jr));
201 /* Update the jr to reflect the new values of PoolId, FileSetId, and JobId. */
202 mig_jcr->jr.PoolId = jcr->jr.PoolId;
203 mig_jcr->jr.FileSetId = jcr->jr.FileSetId;
204 mig_jcr->jr.JobId = mig_jcr->JobId;
206 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
207 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
208 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
211 * Get the PoolId used with the original job. Then
212 * find the pool name from the database record.
214 memset(&pr, 0, sizeof(pr));
215 pr.PoolId = mig_jcr->previous_jr.PoolId;
216 if (!db_get_pool_record(jcr, jcr->db, &pr)) {
217 Jmsg(jcr, M_FATAL, 0, _("Pool for JobId %s not in database. ERR=%s\n"),
218 edit_int64(pr.PoolId, ed1), db_strerror(jcr->db));
221 /* Get the pool resource corresponding to the original job */
222 pool = (POOL *)GetResWithName(R_POOL, pr.Name);
224 Jmsg(jcr, M_FATAL, 0, _("Pool resource \"%s\" not found.\n"), pr.Name);
228 /* If pool storage specified, use it for restore */
229 copy_rstorage(mig_jcr, pool->storage, _("Pool resource"));
230 copy_rstorage(jcr, pool->storage, _("Pool resource"));
233 * If the original backup pool has a NextPool, make sure a
234 * record exists in the database. Note, in this case, we
235 * will be migrating from pool to pool->NextPool.
237 if (pool->NextPool) {
238 jcr->jr.PoolId = get_or_create_pool_record(jcr, pool->NextPool->name());
239 if (jcr->jr.PoolId == 0) {
243 if (!set_migration_wstorage(jcr, pool)) {
246 mig_jcr->pool = jcr->pool = pool->NextPool;
247 pm_strcpy(jcr->pool_source, _("Job Pool's NextPool resource"));
248 mig_jcr->jr.PoolId = jcr->jr.PoolId;
250 Dmsg2(dbglevel, "Write pool=%s read rpool=%s\n", jcr->pool->name(), jcr->rpool->name());
255 * Do a Migration of a previous job
257 * Returns: false on failure
260 bool do_migration(JCR *jcr)
264 JCR *mig_jcr = jcr->mig_jcr; /* newly migrated job */
267 * If mig_jcr is NULL, there is nothing to do for this job,
268 * so set a normal status, cleanup and return OK.
271 set_jcr_job_status(jcr, JS_Terminated);
272 migration_cleanup(jcr, jcr->JobStatus);
276 /* Print Job Start message */
277 Jmsg(jcr, M_INFO, 0, _("Start Migration JobId %s, Job=%s\n"),
278 edit_uint64(jcr->JobId, ed1), jcr->Job);
283 * Open a message channel connection with the Storage
284 * daemon. This is to let him know that our client
285 * will be contacting him for a backup session.
288 Dmsg0(110, "Open connection with storage daemon\n");
289 set_jcr_job_status(jcr, JS_WaitSD);
290 set_jcr_job_status(mig_jcr, JS_WaitSD);
292 * Start conversation with Storage daemon
294 if (!connect_to_storage_daemon(jcr, 10, SDConnectTimeout, 1)) {
297 sd = jcr->store_bsock;
299 * Now start a job with the Storage daemon
301 Dmsg2(dbglevel, "Read store=%s, write store=%s\n",
302 ((STORE *)jcr->rstorage->first())->name(),
303 ((STORE *)jcr->wstorage->first())->name());
304 if (((STORE *)jcr->rstorage->first())->name() == ((STORE *)jcr->wstorage->first())->name()) {
305 Jmsg(jcr, M_FATAL, 0, _("Read storage \"%s\" same as write storage.\n"),
306 ((STORE *)jcr->rstorage->first())->name());
309 if (!start_storage_daemon_job(jcr, jcr->rstorage, jcr->wstorage)) {
312 Dmsg0(150, "Storage daemon connection OK\n");
314 if (!send_bootstrap_file(jcr, sd) ||
315 !response(jcr, sd, OKbootstrap, "Bootstrap", DISPLAY_ERROR)) {
320 * We re-update the job start record so that the start
321 * time is set after the run before job. This avoids
322 * that any files created by the run before job will
323 * be saved twice. They will be backed up in the current
324 * job, but not in the next one unless they are changed.
325 * Without this, they will be backed up in this job and
326 * in the next job run because in that case, their date
327 * is after the start of this run.
329 jcr->start_time = time(NULL);
330 jcr->jr.StartTime = jcr->start_time;
331 jcr->jr.JobTDate = jcr->start_time;
332 set_jcr_job_status(jcr, JS_Running);
334 /* Update job start record for this migration control job */
335 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
336 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
341 mig_jcr->start_time = time(NULL);
342 mig_jcr->jr.StartTime = mig_jcr->start_time;
343 mig_jcr->jr.JobTDate = mig_jcr->start_time;
344 set_jcr_job_status(mig_jcr, JS_Running);
346 /* Update job start record for the real migration backup job */
347 if (!db_update_job_start_record(mig_jcr, mig_jcr->db, &mig_jcr->jr)) {
348 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(mig_jcr->db));
352 Dmsg4(dbglevel, "mig_jcr: Name=%s JobId=%d Type=%c Level=%c\n",
353 mig_jcr->jr.Name, (int)mig_jcr->jr.JobId,
354 mig_jcr->jr.JobType, mig_jcr->jr.JobLevel);
358 * Start the job prior to starting the message thread below
359 * to avoid two threads from using the BSOCK structure at
362 if (!bnet_fsend(sd, "run")) {
367 * Now start a Storage daemon message thread
369 if (!start_storage_daemon_message_thread(jcr)) {
374 set_jcr_job_status(jcr, JS_Running);
375 set_jcr_job_status(mig_jcr, JS_Running);
377 /* Pickup Job termination data */
378 /* Note, the SD stores in jcr->JobFiles/ReadBytes/JobBytes/Errors */
379 wait_for_storage_daemon_termination(jcr);
381 set_jcr_job_status(jcr, jcr->SDJobStatus);
382 if (jcr->JobStatus != JS_Terminated) {
386 migration_cleanup(jcr, jcr->JobStatus);
388 UAContext *ua = new_ua_context(jcr);
389 purge_job_records_from_catalog(ua, jcr->previous_jr.JobId);
400 /* Add an item to the list if it is unique */
401 static void add_unique_id(idpkt *ids, char *item)
406 /* Walk through current list to see if each item is the same as item */
409 for (int i=0; i<(int)sizeof(id); i++) {
412 } else if (*q == ',') {
419 if (strcmp(item, id) == 0) {
423 /* Did not find item, so add it to list */
424 if (ids->count == 0) {
427 pm_strcat(ids->list, ",");
429 pm_strcat(ids->list, item);
431 // Dmsg3(0, "add_uniq count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
436 * Callback handler make list of DB Ids
438 static int unique_dbid_handler(void *ctx, int num_fields, char **row)
440 idpkt *ids = (idpkt *)ctx;
442 add_unique_id(ids, row[0]);
443 Dmsg3(dbglevel, "dbid_hdlr count=%d Ids=%p %s\n", ids->count, ids->list, ids->list);
453 static int item_compare(void *item1, void *item2)
455 uitem *i1 = (uitem *)item1;
456 uitem *i2 = (uitem *)item2;
457 return strcmp(i1->item, i2->item);
460 static int unique_name_handler(void *ctx, int num_fields, char **row)
462 dlist *list = (dlist *)ctx;
464 uitem *new_item = (uitem *)malloc(sizeof(uitem));
467 memset(new_item, 0, sizeof(uitem));
468 new_item->item = bstrdup(row[0]);
469 Dmsg1(dbglevel, "Unique_name_hdlr Item=%s\n", row[0]);
470 item = (uitem *)list->binary_insert((void *)new_item, item_compare);
471 if (item != new_item) { /* already in list */
472 free(new_item->item);
473 free((char *)new_item);
479 /* Get Job names in Pool */
480 const char *sql_job =
481 "SELECT DISTINCT Job.Name from Job,Pool"
482 " WHERE Pool.Name='%s' AND Job.PoolId=Pool.PoolId";
484 /* Get JobIds from regex'ed Job names */
485 const char *sql_jobids_from_job =
486 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool"
487 " WHERE Job.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
488 " ORDER by Job.StartTime";
490 /* Get Client names in Pool */
491 const char *sql_client =
492 "SELECT DISTINCT Client.Name from Client,Pool,Job"
493 " WHERE Pool.Name='%s' AND Job.ClientId=Client.ClientId AND"
494 " Job.PoolId=Pool.PoolId";
496 /* Get JobIds from regex'ed Client names */
497 const char *sql_jobids_from_client =
498 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Job,Pool,Client"
499 " WHERE Client.Name='%s' AND Pool.Name='%s' AND Job.PoolId=Pool.PoolId"
500 " AND Job.ClientId=Client.ClientId AND Job.Type='B'"
501 " ORDER by Job.StartTime";
503 /* Get Volume names in Pool */
504 const char *sql_vol =
505 "SELECT DISTINCT VolumeName FROM Media,Pool WHERE"
506 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
507 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
509 /* Get JobIds from regex'ed Volume names */
510 const char *sql_jobids_from_vol =
511 "SELECT DISTINCT Job.JobId,Job.StartTime FROM Media,JobMedia,Job"
512 " WHERE Media.VolumeName='%s' AND Media.MediaId=JobMedia.MediaId"
513 " AND JobMedia.JobId=Job.JobId AND Job.Type='B'"
514 " ORDER by Job.StartTime";
517 const char *sql_smallest_vol =
518 "SELECT MediaId FROM Media,Pool WHERE"
519 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
520 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
521 " ORDER BY VolBytes ASC LIMIT 1";
523 const char *sql_oldest_vol =
524 "SELECT MediaId FROM Media,Pool WHERE"
525 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
526 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'"
527 " ORDER BY LastWritten ASC LIMIT 1";
529 /* Get JobIds when we have selected MediaId */
530 const char *sql_jobids_from_mediaid =
531 "SELECT DISTINCT Job.JobId,Job.StartTime FROM JobMedia,Job"
532 " WHERE JobMedia.JobId=Job.JobId AND JobMedia.MediaId=%s"
534 " ORDER by Job.StartTime";
536 /* Get tne number of bytes in the pool */
537 const char *sql_pool_bytes =
538 "SELECT SUM(VolBytes) FROM Media,Pool WHERE"
539 " VolStatus in ('Full','Used','Error','Append') AND Media.Enabled=1 AND"
540 " Media.PoolId=Pool.PoolId AND Pool.Name='%s'";
542 /* Get tne number of bytes in the Jobs */
543 const char *sql_job_bytes =
544 "SELECT SUM(JobBytes) FROM Job WHERE JobId IN (%s)";
547 /* Get Media Ids in Pool */
548 const char *sql_mediaids =
549 "SELECT MediaId FROM Media,Pool WHERE"
550 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
551 " Media.PoolId=Pool.PoolId AND Pool.Name='%s' ORDER BY LastWritten ASC";
553 /* Get JobIds in Pool longer than specified time */
554 const char *sql_pool_time =
555 "SELECT DISTINCT Job.JobId from Pool,Job,Media,JobMedia WHERE"
556 " Pool.Name='%s' AND Media.PoolId=Pool.PoolId AND"
557 " VolStatus in ('Full','Used','Error') AND Media.Enabled=1 AND"
559 " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId"
560 " AND Job.RealEndTime<='%s'";
563 * const char *sql_ujobid =
564 * "SELECT DISTINCT Job.Job from Client,Pool,Media,Job,JobMedia "
565 * " WHERE Media.PoolId=Pool.PoolId AND Pool.Name='%s' AND"
566 * " JobMedia.JobId=Job.JobId AND Job.PoolId=Media.PoolId";
573 * This is the central piece of code that finds a job or jobs
574 * actually JobIds to migrate. It first looks to see if one
575 * has been "manually" specified in jcr->MigrateJobId, and if
576 * so, it returns that JobId to be run. Otherwise, it
577 * examines the Selection Type to see what kind of migration
578 * we are doing (Volume, Job, Client, ...) and applies any
579 * Selection Pattern if appropriate to obtain a list of JobIds.
580 * Finally, it will loop over all the JobIds found, except the last
581 * one starting a new job with MigrationJobId set to that JobId, and
582 * finally, it returns the last JobId to the caller.
584 * Returns: -1 on error
585 * 0 if no jobs to migrate
586 * 1 if OK and jcr->previous_jr filled in
588 static int get_job_to_migrate(JCR *jcr)
591 POOL_MEM query(PM_MESSAGE);
596 idpkt ids, mid, jids;
601 char dt[MAX_TIME_LENGTH];
604 ids.list = get_pool_memory(PM_MESSAGE);
607 mid.list = get_pool_memory(PM_MESSAGE);
610 jids.list = get_pool_memory(PM_MESSAGE);
616 * If MigrateJobId is set, then we migrate only that Job,
617 * otherwise, we go through the full selection of jobs to
620 if (jcr->MigrateJobId != 0) {
621 Dmsg1(dbglevel, "At Job start previous jobid=%u\n", jcr->MigrateJobId);
622 edit_uint64(jcr->MigrateJobId, ids.list);
625 switch (jcr->job->selection_type) {
627 if (!regex_find_jobids(jcr, &ids, sql_job, sql_jobids_from_job, "Job")) {
632 if (!regex_find_jobids(jcr, &ids, sql_client, sql_jobids_from_client, "Client")) {
637 if (!regex_find_jobids(jcr, &ids, sql_vol, sql_jobids_from_vol, "Volume")) {
642 if (!jcr->job->selection_pattern) {
643 Jmsg(jcr, M_FATAL, 0, _("No Migration SQL selection pattern specified.\n"));
646 Dmsg1(dbglevel, "SQL=%s\n", jcr->job->selection_pattern);
647 if (!db_sql_query(jcr->db, jcr->job->selection_pattern,
648 unique_dbid_handler, (void *)&ids)) {
649 Jmsg(jcr, M_FATAL, 0,
650 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
654 case MT_SMALLEST_VOL:
655 if (!find_mediaid_then_jobids(jcr, &ids, sql_smallest_vol, "Smallest Volume")) {
660 if (!find_mediaid_then_jobids(jcr, &ids, sql_oldest_vol, "Oldest Volume")) {
665 case MT_POOL_OCCUPANCY:
667 /* Find count of bytes in pool */
668 Mmsg(query, sql_pool_bytes, jcr->rpool->name());
669 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
670 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
673 if (ctx.count == 0) {
674 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
677 pool_bytes = ctx.value;
678 Dmsg2(dbglevel, "highbytes=%d pool=%d\n", (int)jcr->rpool->MigrationHighBytes,
680 if (pool_bytes < (int64_t)jcr->rpool->MigrationHighBytes) {
681 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
684 Dmsg0(dbglevel, "We should do Occupation migration.\n");
687 /* Find a list of MediaIds that could be migrated */
688 Mmsg(query, sql_mediaids, jcr->rpool->name());
689 Dmsg1(dbglevel, "query=%s\n", query.c_str());
690 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
691 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
694 if (ids.count == 0) {
695 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
698 Dmsg2(dbglevel, "Pool Occupancy ids=%d MediaIds=%s\n", ids.count, ids.list);
701 * Now loop over MediaIds getting more JobIds to migrate until
702 * we reduce the pool occupancy below the low water mark.
705 for (int i=0; i < (int)ids.count; i++) {
706 stat = get_next_dbid_from_list(&p, &MediaId);
707 Dmsg2(dbglevel, "get_next_dbid stat=%d MediaId=%u\n", stat, MediaId);
709 Jmsg(jcr, M_FATAL, 0, _("Invalid MediaId found.\n"));
711 } else if (stat == 0) {
715 Mmsg(mid.list, "%s", edit_int64(MediaId, ed1));
716 if (!find_jobids_from_mediaid_list(jcr, &mid, "Volumes")) {
720 pm_strcat(jids.list, ",");
722 pm_strcat(jids.list, mid.list);
723 jids.count += mid.count;
725 /* Now get the count of bytes added */
727 /* Find count of bytes from Jobs */
728 Mmsg(query, sql_job_bytes, mid.list);
729 if (!db_sql_query(jcr->db, query.c_str(), db_int64_handler, (void *)&ctx)) {
730 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
733 pool_bytes -= ctx.value;
734 Dmsg1(dbglevel, "Job bytes=%d\n", (int)ctx.value);
735 Dmsg2(dbglevel, "lowbytes=%d pool=%d\n", (int)jcr->rpool->MigrationLowBytes,
737 if (pool_bytes <= (int64_t)jcr->rpool->MigrationLowBytes) {
738 Dmsg0(dbglevel, "We should be done.\n");
743 Dmsg2(dbglevel, "Pool Occupancy ids=%d JobIds=%s\n", jids.count, jids.list);
748 ttime = time(NULL) - (time_t)jcr->rpool->MigrationTime;
749 (void)localtime_r(&ttime, &tm);
750 strftime(dt, sizeof(dt), "%Y-%m-%d %H:%M:%S", &tm);
753 Mmsg(query, sql_pool_time, jcr->rpool->name(), dt);
754 Dmsg1(dbglevel, "query=%s\n", query.c_str());
755 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)&ids)) {
756 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
759 if (ids.count == 0) {
760 Jmsg(jcr, M_INFO, 0, _("No Volumes found to migrate.\n"));
763 Dmsg2(dbglevel, "PoolTime ids=%d JobIds=%s\n", ids.count, ids.list);
767 Jmsg(jcr, M_FATAL, 0, _("Unknown Migration Selection Type.\n"));
773 * Loop over all jobids except the last one, sending
774 * them to start_migration_job(), which will start a job
775 * for each of them. For the last JobId, we handle it below.
778 if (ids.count == 0) {
779 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
782 Jmsg(jcr, M_INFO, 0, _("The following %u JobId%s will be migrated: %s\n"),
783 ids.count, ids.count==0?"":"s", ids.list);
784 Dmsg2(dbglevel, "Before loop count=%d ids=%s\n", ids.count, ids.list);
785 for (int i=1; i < (int)ids.count; i++) {
787 stat = get_next_jobid_from_list(&p, &JobId);
788 Dmsg3(dbglevel, "get_jobid_no=%d stat=%d JobId=%u\n", i, stat, JobId);
789 jcr->MigrateJobId = JobId;
790 start_migration_job(jcr);
791 Dmsg0(dbglevel, "Back from start_migration_job\n");
793 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
795 } else if (stat == 0) {
796 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
801 /* Now get the last JobId and handle it in the current job */
803 stat = get_next_jobid_from_list(&p, &JobId);
804 Dmsg2(dbglevel, "Last get_next_jobid stat=%d JobId=%u\n", stat, (int)JobId);
806 Jmsg(jcr, M_FATAL, 0, _("Invalid JobId found.\n"));
808 } else if (stat == 0) {
809 Jmsg(jcr, M_INFO, 0, _("No JobIds found to migrate.\n"));
813 jcr->previous_jr.JobId = JobId;
814 Dmsg1(dbglevel, "Previous jobid=%d\n", (int)jcr->previous_jr.JobId);
816 if (!db_get_job_record(jcr, jcr->db, &jcr->previous_jr)) {
817 Jmsg(jcr, M_FATAL, 0, _("Could not get job record for JobId %s to migrate. ERR=%s"),
818 edit_int64(jcr->previous_jr.JobId, ed1),
819 db_strerror(jcr->db));
822 Jmsg(jcr, M_INFO, 0, _("Migration using JobId=%s Job=%s\n"),
823 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
824 Dmsg3(dbglevel, "Migration JobId=%d using JobId=%s Job=%s\n",
826 edit_int64(jcr->previous_jr.JobId, ed1), jcr->previous_jr.Job);
836 free_pool_memory(ids.list);
837 free_pool_memory(mid.list);
838 free_pool_memory(jids.list);
842 static void start_migration_job(JCR *jcr)
844 UAContext *ua = new_ua_context(jcr);
847 Mmsg(ua->cmd, "run %s jobid=%s", jcr->job->hdr.name,
848 edit_uint64(jcr->MigrateJobId, ed1));
849 Dmsg1(dbglevel, "=============== Migration cmd=%s\n", ua->cmd);
850 parse_ua_args(ua); /* parse command */
851 int stat = run_cmd(ua, ua->cmd);
853 Jmsg(jcr, M_ERROR, 0, _("Could not start migration job.\n"));
855 Jmsg(jcr, M_INFO, 0, _("Migration JobId %d started.\n"), stat);
860 static bool find_mediaid_then_jobids(JCR *jcr, idpkt *ids, const char *query1,
864 POOL_MEM query(PM_MESSAGE);
867 /* Basic query for MediaId */
868 Mmsg(query, query1, jcr->rpool->name());
869 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
870 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
873 if (ids->count == 0) {
874 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
875 ok = true; /* Not an error */
877 } else if (ids->count != 1) {
878 Jmsg(jcr, M_FATAL, 0, _("SQL error. Expected 1 MediaId got %d\n"),
882 Dmsg1(dbglevel, "Smallest Vol Jobids=%s\n", ids->list);
884 ok = find_jobids_from_mediaid_list(jcr, ids, type);
890 static bool find_jobids_from_mediaid_list(JCR *jcr, idpkt *ids, const char *type)
893 POOL_MEM query(PM_MESSAGE);
895 Mmsg(query, sql_jobids_from_mediaid, ids->list);
897 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
898 Jmsg(jcr, M_FATAL, 0, _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
901 if (ids->count == 0) {
902 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
909 static bool regex_find_jobids(JCR *jcr, idpkt *ids, const char *query1,
910 const char *query2, const char *type)
914 uitem *last_item = NULL;
919 POOL_MEM query(PM_MESSAGE);
921 item_chain = New(dlist(item, &item->link));
922 if (!jcr->job->selection_pattern) {
923 Jmsg(jcr, M_FATAL, 0, _("No Migration %s selection pattern specified.\n"),
927 Dmsg1(dbglevel, "regex-sel-pattern=%s\n", jcr->job->selection_pattern);
928 /* Basic query for names */
929 Mmsg(query, query1, jcr->rpool->name());
930 Dmsg1(dbglevel, "get name query1=%s\n", query.c_str());
931 if (!db_sql_query(jcr->db, query.c_str(), unique_name_handler,
932 (void *)item_chain)) {
933 Jmsg(jcr, M_FATAL, 0,
934 _("SQL to get %s failed. ERR=%s\n"), type, db_strerror(jcr->db));
937 Dmsg1(dbglevel, "query1 returned %d names\n", item_chain->size());
938 if (item_chain->size() == 0) {
939 Jmsg(jcr, M_INFO, 0, _("Query of Pool \"%s\" returned no Jobs to migrate.\n"),
942 goto bail_out; /* skip regex match */
944 /* Compile regex expression */
945 rc = regcomp(&preg, jcr->job->selection_pattern, REG_EXTENDED);
947 regerror(rc, &preg, prbuf, sizeof(prbuf));
948 Jmsg(jcr, M_FATAL, 0, _("Could not compile regex pattern \"%s\" ERR=%s\n"),
949 jcr->job->selection_pattern, prbuf);
952 /* Now apply the regex to the names and remove any item not matched */
953 foreach_dlist(item, item_chain) {
954 const int nmatch = 30;
955 regmatch_t pmatch[nmatch];
957 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
958 free(last_item->item);
959 item_chain->remove(last_item);
961 Dmsg1(dbglevel, "get name Item=%s\n", item->item);
962 rc = regexec(&preg, item->item, nmatch, pmatch, 0);
964 last_item = NULL; /* keep this one */
970 free(last_item->item);
971 Dmsg1(dbglevel, "Remove item %s\n", last_item->item);
972 item_chain->remove(last_item);
976 if (item_chain->size() == 0) {
977 Jmsg(jcr, M_INFO, 0, _("Regex pattern matched no Jobs to migrate.\n"));
979 goto bail_out; /* skip regex match */
983 * At this point, we have a list of items in item_chain
984 * that have been matched by the regex, so now we need
985 * to look up their jobids.
988 foreach_dlist(item, item_chain) {
989 Dmsg2(dbglevel, "Got %s: %s\n", type, item->item);
990 Mmsg(query, query2, item->item, jcr->rpool->name());
991 Dmsg1(dbglevel, "get id from name query2=%s\n", query.c_str());
992 if (!db_sql_query(jcr->db, query.c_str(), unique_dbid_handler, (void *)ids)) {
993 Jmsg(jcr, M_FATAL, 0,
994 _("SQL failed. ERR=%s\n"), db_strerror(jcr->db));
998 if (ids->count == 0) {
999 Jmsg(jcr, M_INFO, 0, _("No %ss found to migrate.\n"), type);
1004 Dmsg2(dbglevel, "Count=%d Jobids=%s\n", ids->count, ids->list);
1011 * Release resources allocated during backup.
1013 void migration_cleanup(JCR *jcr, int TermCode)
1015 char sdt[MAX_TIME_LENGTH], edt[MAX_TIME_LENGTH];
1016 char ec1[30], ec2[30], ec3[30], ec4[30], ec5[30], elapsed[50];
1017 char ec6[50], ec7[50], ec8[50];
1018 char term_code[100], sd_term_msg[100];
1019 const char *term_msg;
1020 int msg_type = M_INFO;
1024 JCR *mig_jcr = jcr->mig_jcr;
1025 POOL_MEM query(PM_MESSAGE);
1027 Dmsg2(100, "Enter migrate_cleanup %d %c\n", TermCode, TermCode);
1028 update_job_end(jcr, TermCode);
1029 memset(&mr, 0, sizeof(mr));
1032 * Check if we actually did something.
1033 * mig_jcr is jcr of the newly migrated job.
1036 mig_jcr->JobFiles = jcr->JobFiles = jcr->SDJobFiles;
1037 mig_jcr->JobBytes = jcr->JobBytes = jcr->SDJobBytes;
1038 mig_jcr->VolSessionId = jcr->VolSessionId;
1039 mig_jcr->VolSessionTime = jcr->VolSessionTime;
1040 mig_jcr->jr.RealEndTime = 0;
1041 mig_jcr->jr.PriorJobId = jcr->previous_jr.JobId;
1043 update_job_end(mig_jcr, TermCode);
1045 /* Update final items to set them to the previous job's values */
1046 Mmsg(query, "UPDATE Job SET StartTime='%s',EndTime='%s',"
1047 "JobTDate=%s WHERE JobId=%s",
1048 jcr->previous_jr.cStartTime, jcr->previous_jr.cEndTime,
1049 edit_uint64(jcr->previous_jr.JobTDate, ec1),
1050 edit_uint64(mig_jcr->jr.JobId, ec2));
1051 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1053 /* Now mark the previous job as migrated if it terminated normally */
1054 if (jcr->JobStatus == JS_Terminated) {
1055 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1056 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1057 db_sql_query(mig_jcr->db, query.c_str(), NULL, NULL);
1060 if (!db_get_job_record(jcr, jcr->db, &jcr->jr)) {
1061 Jmsg(jcr, M_WARNING, 0, _("Error getting job record for stats: %s"),
1062 db_strerror(jcr->db));
1063 set_jcr_job_status(jcr, JS_ErrorTerminated);
1066 bstrncpy(mr.VolumeName, jcr->VolumeName, sizeof(mr.VolumeName));
1067 if (!db_get_media_record(jcr, jcr->db, &mr)) {
1068 Jmsg(jcr, M_WARNING, 0, _("Error getting Media record for Volume \"%s\": ERR=%s"),
1069 mr.VolumeName, db_strerror(jcr->db));
1070 set_jcr_job_status(jcr, JS_ErrorTerminated);
1073 update_bootstrap_file(mig_jcr);
1075 if (!db_get_job_volume_names(mig_jcr, mig_jcr->db, mig_jcr->jr.JobId, &mig_jcr->VolumeName)) {
1077 * Note, if the job has failed, most likely it did not write any
1078 * tape, so suppress this "error" message since in that case
1079 * it is normal. Or look at it the other way, only for a
1080 * normal exit should we complain about this error.
1082 if (jcr->JobStatus == JS_Terminated && jcr->jr.JobBytes) {
1083 Jmsg(jcr, M_ERROR, 0, "%s", db_strerror(mig_jcr->db));
1085 mig_jcr->VolumeName[0] = 0; /* none */
1087 switch (jcr->JobStatus) {
1089 if (jcr->Errors || jcr->SDErrors) {
1090 term_msg = _("%s OK -- with warnings");
1092 term_msg = _("%s OK");
1096 case JS_ErrorTerminated:
1097 term_msg = _("*** %s Error ***");
1098 msg_type = M_ERROR; /* Generate error message */
1099 if (jcr->store_bsock) {
1100 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1101 if (jcr->SD_msg_chan) {
1102 pthread_cancel(jcr->SD_msg_chan);
1107 term_msg = _("%s Canceled");
1108 if (jcr->store_bsock) {
1109 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
1110 if (jcr->SD_msg_chan) {
1111 pthread_cancel(jcr->SD_msg_chan);
1116 term_msg = _("Inappropriate %s term code");
1120 if (jcr->previous_jr.JobId != 0) {
1121 /* Mark previous job as migrated */
1122 Mmsg(query, "UPDATE Job SET Type='%c' WHERE JobId=%s",
1123 (char)JT_MIGRATED_JOB, edit_uint64(jcr->previous_jr.JobId, ec1));
1124 Dmsg1(000, "Mark: %s\n", query.c_str());
1125 db_sql_query(jcr->db, query.c_str(), NULL, NULL);
1127 term_msg = _("%s -- no files to migrate");
1130 bsnprintf(term_code, sizeof(term_code), term_msg, "Migration");
1131 bstrftimes(sdt, sizeof(sdt), jcr->jr.StartTime);
1132 bstrftimes(edt, sizeof(edt), jcr->jr.EndTime);
1133 RunTime = jcr->jr.EndTime - jcr->jr.StartTime;
1137 kbps = (double)jcr->SDJobBytes / (1000 * RunTime);
1141 jobstatus_to_ascii(jcr->SDJobStatus, sd_term_msg, sizeof(sd_term_msg));
1143 Jmsg(jcr, msg_type, 0, _("Bacula %s (%s): %s\n"
1144 " Prev Backup JobId: %s\n"
1145 " New Backup JobId: %s\n"
1146 " Migration JobId: %s\n"
1147 " Migration Job: %s\n"
1148 " Backup Level: %s%s\n"
1150 " FileSet: \"%s\" %s\n"
1151 " Read Pool: \"%s\" (From %s)\n"
1152 " Read Storage: \"%s\" (From %s)\n"
1153 " Write Pool: \"%s\" (From %s)\n"
1154 " Write Storage: \"%s\" (From %s)\n"
1157 " Elapsed time: %s\n"
1159 " SD Files Written: %s\n"
1160 " SD Bytes Written: %s (%sB)\n"
1161 " Rate: %.1f KB/s\n"
1162 " Volume name(s): %s\n"
1163 " Volume Session Id: %d\n"
1164 " Volume Session Time: %d\n"
1165 " Last Volume Bytes: %s (%sB)\n"
1167 " SD termination status: %s\n"
1168 " Termination: %s\n\n"),
1172 edit_uint64(jcr->previous_jr.JobId, ec6),
1173 mig_jcr ? edit_uint64(mig_jcr->jr.JobId, ec7) : "0",
1174 edit_uint64(jcr->jr.JobId, ec8),
1176 level_to_str(jcr->JobLevel), jcr->since,
1177 jcr->client->name(),
1178 jcr->fileset->name(), jcr->FSCreateTime,
1179 jcr->rpool->name(), jcr->rpool_source,
1180 jcr->rstore?jcr->rstore->name():"*None*",
1181 NPRT(jcr->rstore_source),
1182 jcr->pool->name(), jcr->pool_source,
1183 jcr->wstore?jcr->wstore->name():"*None*",
1184 NPRT(jcr->wstore_source),
1187 edit_utime(RunTime, elapsed, sizeof(elapsed)),
1189 edit_uint64_with_commas(jcr->SDJobFiles, ec1),
1190 edit_uint64_with_commas(jcr->SDJobBytes, ec2),
1191 edit_uint64_with_suffix(jcr->SDJobBytes, ec3),
1193 mig_jcr ? mig_jcr->VolumeName : "",
1195 jcr->VolSessionTime,
1196 edit_uint64_with_commas(mr.VolBytes, ec4),
1197 edit_uint64_with_suffix(mr.VolBytes, ec5),
1202 Dmsg1(100, "migrate_cleanup() mig_jcr=0x%x\n", jcr->mig_jcr);
1204 free_jcr(jcr->mig_jcr);
1205 jcr->mig_jcr = NULL;
1207 Dmsg0(100, "Leave migrate_cleanup()\n");
1211 * Return next DBId from comma separated list
1214 * 1 if next DBId returned
1215 * 0 if no more DBIds are in list
1216 * -1 there is an error
1218 static int get_next_dbid_from_list(char **p, DBId_t *DBId)
1224 for (int i=0; i<(int)sizeof(id); i++) {
1227 } else if (*q == ',') {
1236 } else if (!is_a_number(id)) {
1237 return -1; /* error */
1240 *DBId = str_to_int64(id);
1244 bool set_migration_wstorage(JCR *jcr, POOL *pool)
1246 POOL *wpool = pool->NextPool;
1249 Jmsg(jcr, M_FATAL, 0, _("No Next Pool specification found in Pool \"%s\".\n"),
1254 if (!wpool->storage || wpool->storage->size() == 0) {
1255 Jmsg(jcr, M_FATAL, 0, _("No Storage specification found in Next Pool \"%s\".\n"),
1260 /* If pool storage specified, use it instead of job storage for backup */
1261 copy_wstorage(jcr, wpool->storage, _("Storage from Pool's NextPool resource"));