]> git.ipfire.org Git - thirdparty/mdadm.git/blob - Monitor.c
Fix spare migration and other problems with --monitor.
[thirdparty/mdadm.git] / Monitor.c
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neilb@cse.unsw.edu.au>
23 * Paper: Neil Brown
24 * School of Computer Science and Engineering
25 * The University of New South Wales
26 * Sydney, 2052
27 * Australia
28 */
29
30 #include "mdadm.h"
31 #include "md_p.h"
32 #include "md_u.h"
33 #include <sys/wait.h>
34 #include <sys/signal.h>
35 #include <values.h>
36 #include <syslog.h>
37
38 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom,
39 char *cmd, int dosyslog);
40
41 static char *percentalerts[] = {
42 "RebuildStarted",
43 "Rebuild20",
44 "Rebuild40",
45 "Rebuild60",
46 "Rebuild80",
47 };
48
49 /* The largest number of disks current arrays can manage is 384
50 * This really should be dynamically, but that will have to wait
51 * At least it isn't MD_SB_DISKS.
52 */
53 #define MaxDisks 384
54 int Monitor(mddev_dev_t devlist,
55 char *mailaddr, char *alert_cmd,
56 int period, int daemonise, int scan, int oneshot,
57 int dosyslog, int test, char* pidfile)
58 {
59 /*
60 * Every few seconds, scan every md device looking for changes
61 * When a change is found, log it, possibly run the alert command,
62 * and possibly send Email
63 *
64 * For each array, we record:
65 * Update time
66 * active/working/failed/spare drives
67 * State of each device.
68 * %rebuilt if rebuilding
69 *
70 * If the update time changes, check out all the data again
71 * It is possible that we cannot get the state of each device
72 * due to bugs in the md kernel module.
73 * We also read /proc/mdstat to get rebuild percent,
74 * and to get state on all active devices incase of kernel bug.
75 *
76 * Events are:
77 * Fail
78 * An active device had Faulty set or Active/Sync removed
79 * FailSpare
80 * A spare device had Faulty set
81 * SpareActive
82 * An active device had a reverse transition
83 * RebuildStarted
84 * percent went from -1 to +ve
85 * Rebuild20 Rebuild40 Rebuild60 Rebuild80
86 * percent went from below to not-below that number
87 * DeviceDisappeared
88 * Couldn't access a device which was previously visible
89 *
90 * if we detect an array with active<raid and spare==0
91 * we look at other arrays that have same spare-group
92 * If we find one with active==raid and spare>0,
93 * and if we can get_disk_info and find a name
94 * Then we hot-remove and hot-add to the other array
95 *
96 * If devlist is NULL, then we can monitor everything because --scan
97 * was given. We get an initial list from config file and add anything
98 * that appears in /proc/mdstat
99 */
100
101 struct state {
102 char *devname;
103 int devnum; /* to sync with mdstat info */
104 long utime;
105 int err;
106 char *spare_group;
107 int active, working, failed, spare, raid;
108 int expected_spares;
109 int devstate[MaxDisks];
110 int devid[MaxDisks];
111 int percent;
112 struct state *next;
113 } *statelist = NULL;
114 int finished = 0;
115 struct mdstat_ent *mdstat = NULL;
116 char *mailfrom = NULL;
117
118 if (!mailaddr) {
119 mailaddr = conf_get_mailaddr();
120 if (mailaddr && ! scan)
121 fprintf(stderr, Name ": Monitor using email address \"%s\" from config file\n",
122 mailaddr);
123 }
124 mailfrom = conf_get_mailfrom();
125
126 if (!alert_cmd) {
127 alert_cmd = conf_get_program();
128 if (alert_cmd && ! scan)
129 fprintf(stderr, Name ": Monitor using program \"%s\" from config file\n",
130 alert_cmd);
131 }
132 if (scan && !mailaddr && !alert_cmd) {
133 fprintf(stderr, Name ": No mail address or alert command - not monitoring.\n");
134 return 1;
135 }
136
137 if (daemonise) {
138 int pid = fork();
139 if (pid > 0) {
140 if (!pidfile)
141 printf("%d\n", pid);
142 else {
143 FILE *pid_file;
144 pid_file=fopen(pidfile, "w");
145 if (!pid_file)
146 perror("cannot create pid file");
147 else {
148 fprintf(pid_file,"%d\n", pid);
149 fclose(pid_file);
150 }
151 }
152 return 0;
153 }
154 if (pid < 0) {
155 perror("daemonise");
156 return 1;
157 }
158 close(0);
159 open("/dev/null", 3);
160 dup2(0,1);
161 dup2(0,2);
162 setsid();
163 }
164
165 if (devlist == NULL) {
166 mddev_ident_t mdlist = conf_get_ident(NULL);
167 for (; mdlist; mdlist=mdlist->next) {
168 struct state *st = malloc(sizeof *st);
169 if (st == NULL)
170 continue;
171 st->devname = strdup(mdlist->devname);
172 st->utime = 0;
173 st->next = statelist;
174 st->err = 0;
175 st->devnum = MAXINT;
176 st->percent = -2;
177 st->expected_spares = mdlist->spare_disks;
178 if (mdlist->spare_group)
179 st->spare_group = strdup(mdlist->spare_group);
180 else
181 st->spare_group = NULL;
182 statelist = st;
183 }
184 } else {
185 mddev_dev_t dv;
186 for (dv=devlist ; dv; dv=dv->next) {
187 mddev_ident_t mdlist = conf_get_ident(dv->devname);
188 struct state *st = malloc(sizeof *st);
189 if (st == NULL)
190 continue;
191 st->devname = strdup(dv->devname);
192 st->utime = 0;
193 st->next = statelist;
194 st->err = 0;
195 st->devnum = MAXINT;
196 st->percent = -2;
197 st->expected_spares = -1;
198 st->spare_group = NULL;
199 if (mdlist) {
200 st->expected_spares = mdlist->spare_disks;
201 if (mdlist->spare_group)
202 st->spare_group = strdup(mdlist->spare_group);
203 }
204 statelist = st;
205 }
206 }
207
208
209 while (! finished) {
210 int new_found = 0;
211 struct state *st;
212
213 if (mdstat)
214 free_mdstat(mdstat);
215 mdstat = mdstat_read(oneshot?0:1, 0);
216
217 for (st=statelist; st; st=st->next) {
218 struct { int state, major, minor; } info[MaxDisks];
219 mdu_array_info_t array;
220 struct mdstat_ent *mse = NULL, *mse2;
221 char *dev = st->devname;
222 int fd;
223 unsigned int i;
224
225 if (test)
226 alert("TestMessage", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
227 fd = open(dev, O_RDONLY);
228 if (fd < 0) {
229 if (!st->err)
230 alert("DeviceDisappeared", dev, NULL,
231 mailaddr, mailfrom, alert_cmd, dosyslog);
232 /* fprintf(stderr, Name ": cannot open %s: %s\n",
233 dev, strerror(errno));
234 */ st->err=1;
235 continue;
236 }
237 if (ioctl(fd, GET_ARRAY_INFO, &array)<0) {
238 if (!st->err)
239 alert("DeviceDisappeared", dev, NULL,
240 mailaddr, mailfrom, alert_cmd, dosyslog);
241 /* fprintf(stderr, Name ": cannot get array info for %s: %s\n",
242 dev, strerror(errno));
243 */ st->err=1;
244 close(fd);
245 continue;
246 }
247 if (array.level != 1 && array.level != 5 && array.level != -4 &&
248 array.level != 6 && array.level != 10) {
249 if (!st->err)
250 alert("DeviceDisappeared", dev, "Wrong-Level",
251 mailaddr, mailfrom, alert_cmd, dosyslog);
252 st->err = 1;
253 close(fd);
254 continue;
255 }
256 if (st->devnum == MAXINT) {
257 struct stat stb;
258 if (fstat(fd, &stb) == 0 &&
259 (S_IFMT&stb.st_mode)==S_IFBLK) {
260 if (major(stb.st_rdev) == MD_MAJOR)
261 st->devnum = minor(stb.st_rdev);
262 else
263 st->devnum = -1- (minor(stb.st_rdev)>>6);
264 }
265 }
266
267 for (mse2 = mdstat ; mse2 ; mse2=mse2->next)
268 if (mse2->devnum == st->devnum) {
269 mse2->devnum = MAXINT; /* flag it as "used" */
270 mse = mse2;
271 }
272
273 if (st->utime == array.utime &&
274 st->failed == array.failed_disks &&
275 st->working == array.working_disks &&
276 st->spare == array.spare_disks &&
277 (mse == NULL || (
278 mse->percent == st->percent
279 ))) {
280 close(fd);
281 st->err = 0;
282 continue;
283 }
284 if (st->utime == 0 && /* new array */
285 mse && /* is in /proc/mdstat */
286 mse->pattern && strchr(mse->pattern, '_') /* degraded */
287 )
288 alert("DegradedArray", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
289
290 if (st->utime == 0 && /* new array */
291 st->expected_spares > 0 &&
292 array.spare_disks < st->expected_spares)
293 alert("SparesMissing", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
294 if (mse &&
295 st->percent == -1 &&
296 mse->percent >= 0)
297 alert("RebuildStarted", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
298 if (mse &&
299 st->percent >= 0 &&
300 mse->percent >= 0 &&
301 (mse->percent / 20) > (st->percent / 20))
302 alert(percentalerts[mse->percent/20],
303 dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
304
305 if (mse &&
306 mse->percent == -1 &&
307 st->percent >= 0) {
308 /* Rebuild/sync/whatever just finished.
309 * If there is a number in /mismatch_cnt,
310 * we should report that.
311 */
312 struct sysarray *sra =
313 sysfs_read(-1, st->devnum, GET_MISMATCH);
314 if (sra && sra->mismatch_cnt > 0) {
315 char cnt[40];
316 sprintf(cnt, " mismatches found: %d", sra->mismatch_cnt);
317 alert("RebuildFinished", dev, cnt, mailaddr, mailfrom, alert_cmd, dosyslog);
318 } else
319 alert("RebuildFinished", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
320 if (sra)
321 free(sra);
322 }
323
324 if (mse)
325 st->percent = mse->percent;
326
327
328 for (i=0; i<MaxDisks && i <= array.raid_disks + array.nr_disks;
329 i++) {
330 mdu_disk_info_t disc;
331 disc.number = i;
332 if (ioctl(fd, GET_DISK_INFO, &disc) >= 0) {
333 info[i].state = disc.state;
334 info[i].major = disc.major;
335 info[i].minor = disc.minor;
336 } else
337 info[i].major = info[i].minor = 0;
338 }
339 close(fd);
340
341 for (i=0; i<MaxDisks; i++) {
342 mdu_disk_info_t disc = {0};
343 int newstate=0;
344 int change;
345 char *dv = NULL;
346 disc.number = i;
347 if (i > array.raid_disks + array.nr_disks) {
348 newstate = 0;
349 disc.major = disc.minor = 0;
350 } else if (info[i].major || info[i].minor) {
351 newstate = info[i].state;
352 dv = map_dev(info[i].major, info[i].minor, 1);
353 disc.state = newstate;
354 disc.major = info[i].major;
355 disc.minor = info[i].minor;
356 } else if (mse && mse->pattern && i < strlen(mse->pattern)) {
357 switch(mse->pattern[i]) {
358 case 'U': newstate = 6 /* ACTIVE/SYNC */; break;
359 case '_': newstate = 0; break;
360 }
361 disc.major = disc.minor = 0;
362 }
363 if (dv == NULL && st->devid[i])
364 dv = map_dev(major(st->devid[i]),
365 minor(st->devid[i]), 1);
366 change = newstate ^ st->devstate[i];
367 if (st->utime && change && !st->err) {
368 if (i < (unsigned)array.raid_disks &&
369 (((newstate&change)&(1<<MD_DISK_FAULTY)) ||
370 ((st->devstate[i]&change)&(1<<MD_DISK_ACTIVE)) ||
371 ((st->devstate[i]&change)&(1<<MD_DISK_SYNC)))
372 )
373 alert("Fail", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
374 else if (i >= (unsigned)array.raid_disks &&
375 (disc.major || disc.minor) &&
376 st->devid[i] == makedev(disc.major, disc.minor) &&
377 ((newstate&change)&(1<<MD_DISK_FAULTY))
378 )
379 alert("FailSpare", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
380 else if (i < (unsigned)array.raid_disks &&
381 (((st->devstate[i]&change)&(1<<MD_DISK_FAULTY)) ||
382 ((newstate&change)&(1<<MD_DISK_ACTIVE)) ||
383 ((newstate&change)&(1<<MD_DISK_SYNC)))
384 )
385 alert("SpareActive", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
386 }
387 st->devstate[i] = newstate;
388 st->devid[i] = makedev(disc.major, disc.minor);
389 }
390 st->active = array.active_disks;
391 st->working = array.working_disks;
392 st->spare = array.spare_disks;
393 st->failed = array.failed_disks;
394 st->utime = array.utime;
395 st->raid = array.raid_disks;
396 st->err = 0;
397 }
398 /* now check if there are any new devices found in mdstat */
399 if (scan) {
400 struct mdstat_ent *mse;
401 for (mse=mdstat; mse; mse=mse->next)
402 if (mse->devnum != MAXINT &&
403 (strcmp(mse->level, "raid1")==0 ||
404 strcmp(mse->level, "raid5")==0 ||
405 strcmp(mse->level, "multipath")==0)
406 ) {
407 struct state *st = malloc(sizeof *st);
408 mdu_array_info_t array;
409 int fd;
410 if (st == NULL)
411 continue;
412 st->devname = strdup(get_md_name(mse->devnum));
413 if ((fd = open(st->devname, O_RDONLY)) < 0 ||
414 ioctl(fd, GET_ARRAY_INFO, &array)< 0) {
415 /* no such array */
416 if (fd >=0) close(fd);
417 put_md_name(st->devname);
418 free(st->devname);
419 free(st);
420 continue;
421 }
422 close(fd);
423 st->utime = 0;
424 st->next = statelist;
425 st->err = 1;
426 st->devnum = mse->devnum;
427 st->percent = -2;
428 st->spare_group = NULL;
429 st->expected_spares = -1;
430 statelist = st;
431 alert("NewArray", st->devname, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
432 new_found = 1;
433 }
434 }
435 /* If an array has active < raid && spare == 0 && spare_group != NULL
436 * Look for another array with spare > 0 and active == raid and same spare_group
437 * if found, choose a device and hotremove/hotadd
438 */
439 for (st = statelist; st; st=st->next)
440 if (st->active < st->raid &&
441 st->spare == 0 &&
442 st->spare_group != NULL) {
443 struct state *st2;
444 for (st2=statelist ; st2 ; st2=st2->next)
445 if (st2 != st &&
446 st2->spare > 0 &&
447 st2->active == st2->raid &&
448 st2->spare_group != NULL &&
449 strcmp(st->spare_group, st2->spare_group) == 0) {
450 /* try to remove and add */
451 int fd1 = open(st->devname, O_RDONLY);
452 int fd2 = open(st2->devname, O_RDONLY);
453 int dev = -1;
454 int d;
455 if (fd1 < 0 || fd2 < 0) {
456 if (fd1>=0) close(fd1);
457 if (fd2>=0) close(fd2);
458 continue;
459 }
460 for (d=st2->raid; d < MaxDisks; d++) {
461 if (st2->devid[d] > 0 &&
462 st2->devstate[d] == 0) {
463 dev = st2->devid[d];
464 break;
465 }
466 }
467 if (dev > 0) {
468 if (ioctl(fd2, HOT_REMOVE_DISK,
469 (unsigned long)dev) == 0) {
470 if (ioctl(fd1, HOT_ADD_DISK,
471 (unsigned long)dev) == 0) {
472 alert("MoveSpare", st->devname, st2->devname, mailaddr, mailfrom, alert_cmd, dosyslog);
473 close(fd1);
474 close(fd2);
475 break;
476 }
477 else ioctl(fd2, HOT_ADD_DISK, (unsigned long) dev);
478 }
479 }
480 close(fd1);
481 close(fd2);
482 }
483 }
484 if (!new_found) {
485 if (oneshot)
486 break;
487 else
488 mdstat_wait(period);
489 }
490 test = 0;
491 }
492 if (pidfile)
493 unlink(pidfile);
494 return 0;
495 }
496
497
498 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom, char *cmd,
499 int dosyslog)
500 {
501 int priority;
502
503 if (!cmd && !mailaddr) {
504 time_t now = time(0);
505
506 printf("%1.15s: %s on %s %s\n", ctime(&now)+4, event, dev, disc?disc:"unknown device");
507 }
508 if (cmd) {
509 int pid = fork();
510 switch(pid) {
511 default:
512 waitpid(pid, NULL, 0);
513 break;
514 case -1:
515 break;
516 case 0:
517 execl(cmd, cmd, event, dev, disc, NULL);
518 exit(2);
519 }
520 }
521 if (mailaddr &&
522 (strncmp(event, "Fail", 4)==0 ||
523 strncmp(event, "Test", 4)==0 ||
524 strncmp(event, "Spares", 6)==0 ||
525 strncmp(event, "Degrade", 7)==0)) {
526 FILE *mp = popen(Sendmail, "w");
527 if (mp) {
528 FILE *mdstat;
529 char hname[256];
530 gethostname(hname, sizeof(hname));
531 signal(SIGPIPE, SIG_IGN);
532 if (mailfrom)
533 fprintf(mp, "From: %s\n", mailfrom);
534 else
535 fprintf(mp, "From: " Name " monitoring <root>\n");
536 fprintf(mp, "To: %s\n", mailaddr);
537 fprintf(mp, "Subject: %s event on %s:%s\n\n", event, dev, hname);
538
539 fprintf(mp, "This is an automatically generated mail message from " Name "\n");
540 fprintf(mp, "running on %s\n\n", hname);
541
542 fprintf(mp, "A %s event had been detected on md device %s.\n\n", event, dev);
543
544 if (disc && disc[0] != ' ')
545 fprintf(mp, "It could be related to component device %s.\n\n", disc);
546 if (disc && disc[0] == ' ')
547 fprintf(mp, "Extra information:%s.\n\n", disc);
548
549 fprintf(mp, "Faithfully yours, etc.\n");
550
551 mdstat = fopen("/proc/mdstat", "r");
552 if (mdstat) {
553 char buf[8192];
554 int n;
555 fprintf(mp, "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
556 while ( (n=fread(buf, 1, sizeof(buf), mdstat)) > 0)
557 n=fwrite(buf, 1, n, mp); /* yes, i don't care about the result */
558 fclose(mdstat);
559 }
560 fclose(mp);
561 }
562
563 }
564
565 /* log the event to syslog maybe */
566 if (dosyslog) {
567 /* Log at a different severity depending on the event.
568 *
569 * These are the critical events: */
570 if (strncmp(event, "Fail", 4)==0 ||
571 strncmp(event, "Degrade", 7)==0 ||
572 strncmp(event, "DeviceDisappeared", 17)==0)
573 priority = LOG_CRIT;
574 /* Good to know about, but are not failures: */
575 else if (strncmp(event, "Rebuild", 7)==0 ||
576 strncmp(event, "MoveSpare", 9)==0 ||
577 strncmp(event, "Spares", 6) != 0)
578 priority = LOG_WARNING;
579 /* Everything else: */
580 else
581 priority = LOG_INFO;
582
583 if (disc)
584 syslog(priority, "%s event detected on md device %s, component device %s", event, dev, disc);
585 else
586 syslog(priority, "%s event detected on md device %s", event, dev);
587 }
588 }
589
590 /* Not really Monitor but ... */
591 int Wait(char *dev)
592 {
593 struct stat stb;
594 int devnum;
595 int rv = 1;
596
597 if (stat(dev, &stb) != 0) {
598 fprintf(stderr, Name ": Cannot find %s: %s\n", dev,
599 strerror(errno));
600 return 2;
601 }
602 if (major(stb.st_rdev) == MD_MAJOR)
603 devnum = minor(stb.st_rdev);
604 else
605 devnum = -1-(minor(stb.st_rdev)/64);
606
607 while(1) {
608 struct mdstat_ent *ms = mdstat_read(1, 0);
609 struct mdstat_ent *e;
610
611 for (e=ms ; e; e=e->next)
612 if (e->devnum == devnum)
613 break;
614
615 if (!e || e->percent < 0) {
616 free_mdstat(ms);
617 return rv;
618 }
619 free(ms);
620 rv = 0;
621 mdstat_wait(5);
622 }
623 }