3 * Bacula Director -- backup.c -- responsible for doing backup jobs
5 * Kern Sibbald, March MM
7 * Basic tasks done here:
8 * Open DB and create records for this job.
9 * Open Message Channel with Storage daemon to tell him a job will be starting.
10 * Open connection with File daemon and pass him commands
12 * When the File daemon finishes the job, update the DB.
18 Copyright (C) 2000-2004 Kern Sibbald and John Walker
20 This program is free software; you can redistribute it and/or
21 modify it under the terms of the GNU General Public License as
22 published by the Free Software Foundation; either version 2 of
23 the License, or (at your option) any later version.
25 This program is distributed in the hope that it will be useful,
26 but WITHOUT ANY WARRANTY; without even the implied warranty of
27 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
28 General Public License for more details.
30 You should have received a copy of the GNU General Public
31 License along with this program; if not, write to the Free
32 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
41 /* Commands sent to File daemon */
42 static char backupcmd[] = "backup\n";
43 static char storaddr[] = "storage address=%s port=%d ssl=%d\n";
45 /* Responses received from File daemon */
46 static char OKbackup[] = "2000 OK backup\n";
47 static char OKstore[] = "2000 OK storage\n";
48 static char EndJob[] = "2800 End Job TermCode=%d JobFiles=%u "
49 "ReadBytes=%" lld " JobBytes=%" lld " Errors=%u\n";
52 /* Forward referenced functions */
53 static void backup_cleanup(JCR *jcr, int TermCode, char *since, FILESET_DBR *fsr);
55 /* External functions */
58 * Do a backup of the specified FileSet
60 * Returns: 0 on failure
63 int do_backup(JCR *jcr)
65 char since[MAXSTRING];
74 if (!get_or_create_client_record(jcr)) {
78 if (!get_or_create_fileset_record(jcr, &fsr)) {
82 get_level_since_time(jcr, since, sizeof(since));
84 jcr->fname = get_pool_memory(PM_FNAME);
87 * Get the Pool record -- first apply any level defined pools
89 switch (jcr->JobLevel) {
92 jcr->pool = jcr->full_pool;
97 jcr->pool = jcr->inc_pool;
102 jcr->pool = jcr->dif_pool;
106 memset(&pr, 0, sizeof(pr));
107 bstrncpy(pr.Name, jcr->pool->hdr.name, sizeof(pr.Name));
109 while (!db_get_pool_record(jcr, jcr->db, &pr)) { /* get by Name */
110 /* Try to create the pool */
111 if (create_pool(jcr, jcr->db, jcr->pool, POOL_OP_CREATE) < 0) {
112 Jmsg(jcr, M_FATAL, 0, _("Pool %s not in database. %s"), pr.Name,
113 db_strerror(jcr->db));
116 Jmsg(jcr, M_INFO, 0, _("Pool %s created in database.\n"), pr.Name);
119 jcr->PoolId = pr.PoolId; /****FIXME**** this can go away */
120 jcr->jr.PoolId = pr.PoolId;
123 /* Print Job Start message */
124 Jmsg(jcr, M_INFO, 0, _("Start Backup JobId %u, Job=%s\n"),
125 jcr->JobId, jcr->Job);
127 set_jcr_job_status(jcr, JS_Running);
128 Dmsg2(100, "JobId=%d JobLevel=%c\n", jcr->jr.JobId, jcr->jr.JobLevel);
129 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
130 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
135 * Open a message channel connection with the Storage
136 * daemon. This is to let him know that our client
137 * will be contacting him for a backup session.
140 Dmsg0(110, "Open connection with storage daemon\n");
141 set_jcr_job_status(jcr, JS_WaitSD);
143 * Start conversation with Storage daemon
145 if (!connect_to_storage_daemon(jcr, 10, SDConnectTimeout, 1)) {
149 * Now start a job with the Storage daemon
151 if (!start_storage_daemon_job(jcr)) {
155 * Now start a Storage daemon message thread
157 if (!start_storage_daemon_message_thread(jcr)) {
160 Dmsg0(150, "Storage daemon connection OK\n");
162 set_jcr_job_status(jcr, JS_WaitFD);
163 if (!connect_to_file_daemon(jcr, 10, FDConnectTimeout, 1)) {
167 set_jcr_job_status(jcr, JS_Running);
168 fd = jcr->file_bsock;
170 if (!send_include_list(jcr)) {
174 if (!send_exclude_list(jcr)) {
178 if (!send_level_command(jcr)) {
183 * send Storage daemon address to the File daemon
185 store = (STORE *)jcr->storage[0]->first();
186 if (store->SDDport == 0) {
187 store->SDDport = store->SDport;
189 bnet_fsend(fd, storaddr, store->address, store->SDDport,
191 if (!response(jcr, fd, OKstore, "Storage", DISPLAY_ERROR)) {
196 if (!send_run_before_and_after_commands(jcr)) {
200 /* Send backup command */
201 bnet_fsend(fd, backupcmd);
202 if (!response(jcr, fd, OKbackup, "backup", DISPLAY_ERROR)) {
206 /* Pickup Job termination data */
207 stat = wait_for_job_termination(jcr);
208 backup_cleanup(jcr, stat, since, &fsr);
212 backup_cleanup(jcr, JS_ErrorTerminated, since, &fsr);
217 * Here we wait for the File daemon to signal termination,
218 * then we wait for the Storage daemon. When both
219 * are done, we return the job status.
220 * Also used by restore.c
222 int wait_for_job_termination(JCR *jcr)
225 BSOCK *fd = jcr->file_bsock;
227 uint32_t JobFiles, Errors;
228 uint64_t ReadBytes, JobBytes;
230 set_jcr_job_status(jcr, JS_Running);
231 /* Wait for Client to terminate */
232 while ((n = bget_dirmsg(fd)) >= 0) {
233 if (!fd_ok && sscanf(fd->msg, EndJob, &jcr->FDJobStatus, &JobFiles,
234 &ReadBytes, &JobBytes, &Errors) == 5) {
236 set_jcr_job_status(jcr, jcr->FDJobStatus);
237 Dmsg1(100, "FDStatus=%c\n", (char)jcr->JobStatus);
239 Jmsg(jcr, M_WARNING, 0, _("Unexpected Client Job message: %s\n"),
242 if (job_canceled(jcr)) {
246 if (is_bnet_error(fd)) {
247 Jmsg(jcr, M_FATAL, 0, _("Network error with FD during %s: ERR=%s\n"),
248 job_type_to_str(jcr->JobType), bnet_strerror(fd));
250 bnet_sig(fd, BNET_TERMINATE); /* tell Client we are terminating */
252 /* Note, the SD stores in jcr->JobFiles/ReadBytes/JobBytes/Errors */
253 wait_for_storage_daemon_termination(jcr);
256 /* Return values from FD */
258 jcr->JobFiles = JobFiles;
259 jcr->Errors = Errors;
260 jcr->ReadBytes = ReadBytes;
261 jcr->JobBytes = JobBytes;
263 Jmsg(jcr, M_FATAL, 0, _("No Job status returned from FD.\n"));
266 // Dmsg4(100, "fd_ok=%d FDJS=%d JS=%d SDJS=%d\n", fd_ok, jcr->FDJobStatus,
267 // jcr->JobStatus, jcr->SDJobStatus);
269 /* Return the first error status we find Dir, FD, or SD */
270 if (!fd_ok || is_bnet_error(fd)) {
271 jcr->FDJobStatus = JS_ErrorTerminated;
273 if (jcr->JobStatus != JS_Terminated) {
274 return jcr->JobStatus;
276 if (jcr->FDJobStatus != JS_Terminated) {
277 return jcr->FDJobStatus;
279 return jcr->SDJobStatus;
283 * Release resources allocated during backup.
285 static void backup_cleanup(JCR *jcr, int TermCode, char *since, FILESET_DBR *fsr)
287 char sdt[50], edt[50];
288 char ec1[30], ec2[30], ec3[30], ec4[30], ec5[30], compress[50];
289 char term_code[100], fd_term_msg[100], sd_term_msg[100];
290 const char *term_msg;
293 double kbps, compression;
296 Dmsg2(100, "Enter backup_cleanup %d %c\n", TermCode, TermCode);
297 dequeue_messages(jcr); /* display any queued messages */
298 memset(&mr, 0, sizeof(mr));
299 set_jcr_job_status(jcr, TermCode);
301 update_job_end_record(jcr); /* update database */
303 if (!db_get_job_record(jcr, jcr->db, &jcr->jr)) {
304 Jmsg(jcr, M_WARNING, 0, _("Error getting job record for stats: %s"),
305 db_strerror(jcr->db));
306 set_jcr_job_status(jcr, JS_ErrorTerminated);
309 bstrncpy(mr.VolumeName, jcr->VolumeName, sizeof(mr.VolumeName));
310 if (!db_get_media_record(jcr, jcr->db, &mr)) {
311 Jmsg(jcr, M_WARNING, 0, _("Error getting Media record for Volume \"%s\": ERR=%s"),
312 mr.VolumeName, db_strerror(jcr->db));
313 set_jcr_job_status(jcr, JS_ErrorTerminated);
316 /* Now update the bootstrap file if any */
317 if (jcr->JobStatus == JS_Terminated && jcr->jr.JobBytes &&
318 jcr->job->WriteBootstrap) {
322 char *fname = jcr->job->WriteBootstrap;
323 VOL_PARAMS *VolParams = NULL;
329 bpipe = open_bpipe(fname, 0, "w");
330 fd = bpipe ? bpipe->wfd : NULL;
332 /* ***FIXME*** handle BASE */
333 fd = fopen(fname, jcr->JobLevel==L_FULL?"w+":"a+");
336 VolCount = db_get_job_volume_parameters(jcr, jcr->db, jcr->JobId,
339 Jmsg(jcr, M_ERROR, 0, _("Could not get Job Volume Parameters to "
340 "update Bootstrap file. ERR=%s\n"), db_strerror(jcr->db));
341 if (jcr->SDJobFiles != 0) {
342 set_jcr_job_status(jcr, JS_ErrorTerminated);
346 for (int i=0; i < VolCount; i++) {
347 /* Write the record */
348 fprintf(fd, "Volume=\"%s\"\n", VolParams[i].VolumeName);
349 fprintf(fd, "VolSessionId=%u\n", jcr->VolSessionId);
350 fprintf(fd, "VolSessionTime=%u\n", jcr->VolSessionTime);
351 fprintf(fd, "VolFile=%u-%u\n", VolParams[i].StartFile,
352 VolParams[i].EndFile);
353 fprintf(fd, "VolBlock=%u-%u\n", VolParams[i].StartBlock,
354 VolParams[i].EndBlock);
355 fprintf(fd, "FileIndex=%d-%d\n", VolParams[i].FirstIndex,
356 VolParams[i].LastIndex);
368 Jmsg(jcr, M_ERROR, 0, _("Could not open WriteBootstrap file:\n"
369 "%s: ERR=%s\n"), fname, be.strerror());
370 set_jcr_job_status(jcr, JS_ErrorTerminated);
374 msg_type = M_INFO; /* by default INFO message */
375 switch (jcr->JobStatus) {
377 if (jcr->Errors || jcr->SDErrors) {
378 term_msg = _("Backup OK -- with warnings");
380 term_msg = _("Backup OK");
384 case JS_ErrorTerminated:
385 term_msg = _("*** Backup Error ***");
386 msg_type = M_ERROR; /* Generate error message */
387 if (jcr->store_bsock) {
388 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
389 if (jcr->SD_msg_chan) {
390 pthread_cancel(jcr->SD_msg_chan);
395 term_msg = _("Backup Canceled");
396 if (jcr->store_bsock) {
397 bnet_sig(jcr->store_bsock, BNET_TERMINATE);
398 if (jcr->SD_msg_chan) {
399 pthread_cancel(jcr->SD_msg_chan);
404 term_msg = term_code;
405 sprintf(term_code, _("Inappropriate term code: %c\n"), jcr->JobStatus);
408 bstrftime(sdt, sizeof(sdt), jcr->jr.StartTime);
409 bstrftime(edt, sizeof(edt), jcr->jr.EndTime);
410 RunTime = jcr->jr.EndTime - jcr->jr.StartTime;
414 kbps = (double)jcr->jr.JobBytes / (1000 * RunTime);
416 if (!db_get_job_volume_names(jcr, jcr->db, jcr->jr.JobId, &jcr->VolumeName)) {
418 * Note, if the job has erred, most likely it did not write any
419 * tape, so suppress this "error" message since in that case
420 * it is normal. Or look at it the other way, only for a
421 * normal exit should we complain about this error.
423 if (jcr->JobStatus == JS_Terminated && jcr->jr.JobBytes) {
424 Jmsg(jcr, M_ERROR, 0, "%s", db_strerror(jcr->db));
426 jcr->VolumeName[0] = 0; /* none */
429 if (jcr->ReadBytes == 0) {
430 bstrncpy(compress, "None", sizeof(compress));
432 compression = (double)100 - 100.0 * ((double)jcr->JobBytes / (double)jcr->ReadBytes);
433 if (compression < 0.5) {
434 bstrncpy(compress, "None", sizeof(compress));
436 bsnprintf(compress, sizeof(compress), "%.1f %%", (float)compression);
439 jobstatus_to_ascii(jcr->FDJobStatus, fd_term_msg, sizeof(fd_term_msg));
440 jobstatus_to_ascii(jcr->SDJobStatus, sd_term_msg, sizeof(sd_term_msg));
442 // bmicrosleep(15, 0); /* for debugging SIGHUP */
444 Jmsg(jcr, msg_type, 0, _("Bacula " VERSION " (" LSMDATE "): %s\n\
447 Backup Level: %s%s\n\
449 FileSet: \"%s\" %s\n\
453 FD Files Written: %s\n\
454 SD Files Written: %s\n\
455 FD Bytes Written: %s\n\
456 SD Bytes Written: %s\n\
458 Software Compression: %s\n\
459 Volume name(s): %s\n\
460 Volume Session Id: %d\n\
461 Volume Session Time: %d\n\
462 Last Volume Bytes: %s\n\
463 Non-fatal FD errors: %d\n\
465 FD termination status: %s\n\
466 SD termination status: %s\n\
467 Termination: %s\n\n"),
471 level_to_str(jcr->JobLevel), since,
472 jcr->client->hdr.name,
473 jcr->fileset->hdr.name, fsr->cCreateTime,
477 edit_uint64_with_commas(jcr->jr.JobFiles, ec1),
478 edit_uint64_with_commas(jcr->SDJobFiles, ec4),
479 edit_uint64_with_commas(jcr->jr.JobBytes, ec2),
480 edit_uint64_with_commas(jcr->SDJobBytes, ec5),
486 edit_uint64_with_commas(mr.VolBytes, ec3),
493 Dmsg0(100, "Leave backup_cleanup()\n");