]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Monitor.c
2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
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.
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.
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
22 * Email: <neilb@cse.unsw.edu.au>
24 * School of Computer Science and Engineering
25 * The University of New South Wales
38 static void alert(char *event
, char *dev
, char *disc
, char *mailaddr
, char *mailfrom
,
39 char *cmd
, int dosyslog
);
41 static char *percentalerts
[] = {
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.
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
)
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
64 * For each array, we record:
66 * active/working/failed/spare drives
67 * State of each device.
68 * %rebuilt if rebuilding
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.
78 * An active device had Faulty set or Active/Sync removed
80 * A spare device had Faulty set
82 * An active device had a reverse transition
84 * percent went from -1 to +ve
85 * Rebuild20 Rebuild40 Rebuild60 Rebuild80
86 * percent went from below to not-below that number
88 * Couldn't access a device which was previously visible
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
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
103 int devnum
; /* to sync with mdstat info */
107 int active
, working
, failed
, spare
, raid
;
109 int devstate
[MaxDisks
];
115 struct mdstat_ent
*mdstat
= NULL
;
116 char *mailfrom
= NULL
;
119 mailaddr
= conf_get_mailaddr();
120 if (mailaddr
&& ! scan
)
121 fprintf(stderr
, Name
": Monitor using email address \"%s\" from config file\n",
124 mailfrom
= conf_get_mailfrom();
127 alert_cmd
= conf_get_program();
128 if (alert_cmd
&& ! scan
)
129 fprintf(stderr
, Name
": Monitor using program \"%s\" from config file\n",
132 if (scan
&& !mailaddr
&& !alert_cmd
) {
133 fprintf(stderr
, Name
": No mail address or alert command - not monitoring.\n");
144 pid_file
=fopen(pidfile
, "w");
146 perror("cannot create pid file");
148 fprintf(pid_file
,"%d\n", pid
);
159 open("/dev/null", 3);
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
);
171 st
->devname
= strdup(mdlist
->devname
);
173 st
->next
= statelist
;
175 st
->devnum
= INT_MAX
;
177 st
->expected_spares
= mdlist
->spare_disks
;
178 if (mdlist
->spare_group
)
179 st
->spare_group
= strdup(mdlist
->spare_group
);
181 st
->spare_group
= NULL
;
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
);
191 st
->devname
= strdup(dv
->devname
);
193 st
->next
= statelist
;
195 st
->devnum
= INT_MAX
;
197 st
->expected_spares
= -1;
198 st
->spare_group
= NULL
;
200 st
->expected_spares
= mdlist
->spare_disks
;
201 if (mdlist
->spare_group
)
202 st
->spare_group
= strdup(mdlist
->spare_group
);
215 mdstat
= mdstat_read(oneshot
?0:1, 0);
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
;
226 alert("TestMessage", dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
227 fd
= open(dev
, O_RDONLY
);
230 alert("DeviceDisappeared", dev
, NULL
,
231 mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
232 /* fprintf(stderr, Name ": cannot open %s: %s\n",
233 dev, strerror(errno));
237 fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
238 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)<0) {
240 alert("DeviceDisappeared", dev
, NULL
,
241 mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
242 /* fprintf(stderr, Name ": cannot get array info for %s: %s\n",
243 dev, strerror(errno));
248 /* It's much easier to list what array levels can't
249 * have a device disappear than all of them that can
251 if (array
.level
== 0 || array
.level
== -1) {
253 alert("DeviceDisappeared", dev
, "Wrong-Level",
254 mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
259 if (st
->devnum
== INT_MAX
) {
261 if (fstat(fd
, &stb
) == 0 &&
262 (S_IFMT
&stb
.st_mode
)==S_IFBLK
) {
263 if (major(stb
.st_rdev
) == MD_MAJOR
)
264 st
->devnum
= minor(stb
.st_rdev
);
266 st
->devnum
= -1- (minor(stb
.st_rdev
)>>6);
270 for (mse2
= mdstat
; mse2
; mse2
=mse2
->next
)
271 if (mse2
->devnum
== st
->devnum
) {
272 mse2
->devnum
= INT_MAX
; /* flag it as "used" */
276 if (st
->utime
== array
.utime
&&
277 st
->failed
== array
.failed_disks
&&
278 st
->working
== array
.working_disks
&&
279 st
->spare
== array
.spare_disks
&&
281 mse
->percent
== st
->percent
287 if (st
->utime
== 0 && /* new array */
288 mse
&& /* is in /proc/mdstat */
289 mse
->pattern
&& strchr(mse
->pattern
, '_') /* degraded */
291 alert("DegradedArray", dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
293 if (st
->utime
== 0 && /* new array */
294 st
->expected_spares
> 0 &&
295 array
.spare_disks
< st
->expected_spares
)
296 alert("SparesMissing", dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
300 alert("RebuildStarted", dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
304 (mse
->percent
/ 20) > (st
->percent
/ 20))
305 alert(percentalerts
[mse
->percent
/20],
306 dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
309 mse
->percent
== -1 &&
311 /* Rebuild/sync/whatever just finished.
312 * If there is a number in /mismatch_cnt,
313 * we should report that.
316 sysfs_read(-1, st
->devnum
, GET_MISMATCH
);
317 if (sra
&& sra
->mismatch_cnt
> 0) {
319 sprintf(cnt
, " mismatches found: %d", sra
->mismatch_cnt
);
320 alert("RebuildFinished", dev
, cnt
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
322 alert("RebuildFinished", dev
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
328 st
->percent
= mse
->percent
;
331 for (i
=0; i
<MaxDisks
&& i
<= array
.raid_disks
+ array
.nr_disks
;
333 mdu_disk_info_t disc
;
335 if (ioctl(fd
, GET_DISK_INFO
, &disc
) >= 0) {
336 info
[i
].state
= disc
.state
;
337 info
[i
].major
= disc
.major
;
338 info
[i
].minor
= disc
.minor
;
340 info
[i
].major
= info
[i
].minor
= 0;
344 for (i
=0; i
<MaxDisks
; i
++) {
345 mdu_disk_info_t disc
= {0};
350 if (i
> array
.raid_disks
+ array
.nr_disks
) {
352 disc
.major
= disc
.minor
= 0;
353 } else if (info
[i
].major
|| info
[i
].minor
) {
354 newstate
= info
[i
].state
;
355 dv
= map_dev(info
[i
].major
, info
[i
].minor
, 1);
356 disc
.state
= newstate
;
357 disc
.major
= info
[i
].major
;
358 disc
.minor
= info
[i
].minor
;
359 } else if (mse
&& mse
->pattern
&& i
< strlen(mse
->pattern
)) {
360 switch(mse
->pattern
[i
]) {
361 case 'U': newstate
= 6 /* ACTIVE/SYNC */; break;
362 case '_': newstate
= 0; break;
364 disc
.major
= disc
.minor
= 0;
366 if (dv
== NULL
&& st
->devid
[i
])
367 dv
= map_dev(major(st
->devid
[i
]),
368 minor(st
->devid
[i
]), 1);
369 change
= newstate
^ st
->devstate
[i
];
370 if (st
->utime
&& change
&& !st
->err
) {
371 if (i
< (unsigned)array
.raid_disks
&&
372 (((newstate
&change
)&(1<<MD_DISK_FAULTY
)) ||
373 ((st
->devstate
[i
]&change
)&(1<<MD_DISK_ACTIVE
)) ||
374 ((st
->devstate
[i
]&change
)&(1<<MD_DISK_SYNC
)))
376 alert("Fail", dev
, dv
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
377 else if (i
>= (unsigned)array
.raid_disks
&&
378 (disc
.major
|| disc
.minor
) &&
379 st
->devid
[i
] == makedev(disc
.major
, disc
.minor
) &&
380 ((newstate
&change
)&(1<<MD_DISK_FAULTY
))
382 alert("FailSpare", dev
, dv
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
383 else if (i
< (unsigned)array
.raid_disks
&&
384 (((st
->devstate
[i
]&change
)&(1<<MD_DISK_FAULTY
)) ||
385 ((newstate
&change
)&(1<<MD_DISK_ACTIVE
)) ||
386 ((newstate
&change
)&(1<<MD_DISK_SYNC
)))
388 alert("SpareActive", dev
, dv
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
390 st
->devstate
[i
] = newstate
;
391 st
->devid
[i
] = makedev(disc
.major
, disc
.minor
);
393 st
->active
= array
.active_disks
;
394 st
->working
= array
.working_disks
;
395 st
->spare
= array
.spare_disks
;
396 st
->failed
= array
.failed_disks
;
397 st
->utime
= array
.utime
;
398 st
->raid
= array
.raid_disks
;
401 /* now check if there are any new devices found in mdstat */
403 struct mdstat_ent
*mse
;
404 for (mse
=mdstat
; mse
; mse
=mse
->next
)
405 if (mse
->devnum
!= INT_MAX
&&
407 (strcmp(mse
->level
, "raid0")!=0 &&
408 strcmp(mse
->level
, "linear")!=0)
410 struct state
*st
= malloc(sizeof *st
);
411 mdu_array_info_t array
;
415 st
->devname
= strdup(get_md_name(mse
->devnum
));
416 if ((fd
= open(st
->devname
, O_RDONLY
)) < 0 ||
417 ioctl(fd
, GET_ARRAY_INFO
, &array
)< 0) {
419 if (fd
>=0) close(fd
);
420 put_md_name(st
->devname
);
427 st
->next
= statelist
;
429 st
->devnum
= mse
->devnum
;
431 st
->spare_group
= NULL
;
432 st
->expected_spares
= -1;
434 alert("NewArray", st
->devname
, NULL
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
438 /* If an array has active < raid && spare == 0 && spare_group != NULL
439 * Look for another array with spare > 0 and active == raid and same spare_group
440 * if found, choose a device and hotremove/hotadd
442 for (st
= statelist
; st
; st
=st
->next
)
443 if (st
->active
< st
->raid
&&
445 st
->spare_group
!= NULL
) {
447 for (st2
=statelist
; st2
; st2
=st2
->next
)
450 st2
->active
== st2
->raid
&&
451 st2
->spare_group
!= NULL
&&
452 strcmp(st
->spare_group
, st2
->spare_group
) == 0) {
453 /* try to remove and add */
454 int fd1
= open(st
->devname
, O_RDONLY
);
455 int fd2
= open(st2
->devname
, O_RDONLY
);
458 if (fd1
< 0 || fd2
< 0) {
459 if (fd1
>=0) close(fd1
);
460 if (fd2
>=0) close(fd2
);
463 for (d
=st2
->raid
; d
< MaxDisks
; d
++) {
464 if (st2
->devid
[d
] > 0 &&
465 st2
->devstate
[d
] == 0) {
471 if (ioctl(fd2
, HOT_REMOVE_DISK
,
472 (unsigned long)dev
) == 0) {
473 if (ioctl(fd1
, HOT_ADD_DISK
,
474 (unsigned long)dev
) == 0) {
475 alert("MoveSpare", st
->devname
, st2
->devname
, mailaddr
, mailfrom
, alert_cmd
, dosyslog
);
480 else ioctl(fd2
, HOT_ADD_DISK
, (unsigned long) dev
);
501 static void alert(char *event
, char *dev
, char *disc
, char *mailaddr
, char *mailfrom
, char *cmd
,
506 if (!cmd
&& !mailaddr
) {
507 time_t now
= time(0);
509 printf("%1.15s: %s on %s %s\n", ctime(&now
)+4, event
, dev
, disc
?disc
:"unknown device");
515 waitpid(pid
, NULL
, 0);
520 execl(cmd
, cmd
, event
, dev
, disc
, NULL
);
525 (strncmp(event
, "Fail", 4)==0 ||
526 strncmp(event
, "Test", 4)==0 ||
527 strncmp(event
, "Spares", 6)==0 ||
528 strncmp(event
, "Degrade", 7)==0)) {
529 FILE *mp
= popen(Sendmail
, "w");
533 gethostname(hname
, sizeof(hname
));
534 signal(SIGPIPE
, SIG_IGN
);
536 fprintf(mp
, "From: %s\n", mailfrom
);
538 fprintf(mp
, "From: " Name
" monitoring <root>\n");
539 fprintf(mp
, "To: %s\n", mailaddr
);
540 fprintf(mp
, "Subject: %s event on %s:%s\n\n", event
, dev
, hname
);
542 fprintf(mp
, "This is an automatically generated mail message from " Name
"\n");
543 fprintf(mp
, "running on %s\n\n", hname
);
545 fprintf(mp
, "A %s event had been detected on md device %s.\n\n", event
, dev
);
547 if (disc
&& disc
[0] != ' ')
548 fprintf(mp
, "It could be related to component device %s.\n\n", disc
);
549 if (disc
&& disc
[0] == ' ')
550 fprintf(mp
, "Extra information:%s.\n\n", disc
);
552 fprintf(mp
, "Faithfully yours, etc.\n");
554 mdstat
= fopen("/proc/mdstat", "r");
558 fprintf(mp
, "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
559 while ( (n
=fread(buf
, 1, sizeof(buf
), mdstat
)) > 0)
560 n
=fwrite(buf
, 1, n
, mp
); /* yes, i don't care about the result */
568 /* log the event to syslog maybe */
570 /* Log at a different severity depending on the event.
572 * These are the critical events: */
573 if (strncmp(event
, "Fail", 4)==0 ||
574 strncmp(event
, "Degrade", 7)==0 ||
575 strncmp(event
, "DeviceDisappeared", 17)==0)
577 /* Good to know about, but are not failures: */
578 else if (strncmp(event
, "Rebuild", 7)==0 ||
579 strncmp(event
, "MoveSpare", 9)==0 ||
580 strncmp(event
, "Spares", 6) != 0)
581 priority
= LOG_WARNING
;
582 /* Everything else: */
587 syslog(priority
, "%s event detected on md device %s, component device %s", event
, dev
, disc
);
589 syslog(priority
, "%s event detected on md device %s", event
, dev
);
593 /* Not really Monitor but ... */
600 if (stat(dev
, &stb
) != 0) {
601 fprintf(stderr
, Name
": Cannot find %s: %s\n", dev
,
605 if (major(stb
.st_rdev
) == MD_MAJOR
)
606 devnum
= minor(stb
.st_rdev
);
608 devnum
= -1-(minor(stb
.st_rdev
)/64);
611 struct mdstat_ent
*ms
= mdstat_read(1, 0);
612 struct mdstat_ent
*e
;
614 for (e
=ms
; e
; e
=e
->next
)
615 if (e
->devnum
== devnum
)
618 if (!e
|| e
->percent
< 0) {