#include "md_u.h"
#include <sys/wait.h>
#include <sys/signal.h>
+#include <values.h>
static void alert(char *event, char *dev, char *disc, char *mailaddr, char *cmd);
int Monitor(mddev_dev_t devlist,
char *mailaddr, char *alert_cmd,
- int period, int scan,
- char *config)
+ int period, int daemonise, int scan, int oneshot,
+ char *config, int test, char* pidfile)
{
/*
* Every few seconds, scan every md device looking for changes
int err;
char *spare_group;
int active, working, failed, spare, raid;
+ int expected_spares;
int devstate[MD_SB_DISKS];
int devid[MD_SB_DISKS];
int percent;
int finished = 0;
struct mdstat_ent *mdstat = NULL;
- if (!mailaddr && scan)
+ if (!mailaddr) {
mailaddr = conf_get_mailaddr(config);
- if (!alert_cmd && scan)
+ if (mailaddr && ! scan)
+ fprintf(stderr, Name ": Monitor using email address \"%s\" from config file\n",
+ mailaddr);
+ }
+ if (!alert_cmd) {
alert_cmd = conf_get_program(config);
- if (scan && !mailaddr && !alert_cmd)
- return 0;
+ if (alert_cmd && ! scan)
+ fprintf(stderr, Name ": Monitor using program \"%s\" from config file\n",
+ alert_cmd);
+ }
+ if (scan && !mailaddr && !alert_cmd) {
+ fprintf(stderr, Name ": No mail address or alert command - not monitoring.\n");
+ return 1;
+ }
+
+ if (daemonise) {
+ int pid = fork();
+ if (pid > 0) {
+ if (!pidfile)
+ printf("%d\n", pid);
+ else {
+ FILE *pid_file;
+ pid_file=fopen(pidfile, "w");
+ if (!pid_file)
+ perror("cannot create pid file");
+ else {
+ fprintf(pid_file,"%d\n", pid);
+ fclose(pid_file);
+ }
+ }
+ return 0;
+ }
+ if (pid < 0) {
+ perror("daemonise");
+ return 1;
+ }
+ close(0);
+ open("/dev/null", 3);
+ dup2(0,1);
+ dup2(0,2);
+ setsid();
+ }
if (devlist == NULL) {
mddev_ident_t mdlist = conf_get_ident(config, NULL);
st->devname = strdup(mdlist->devname);
st->utime = 0;
st->next = statelist;
- st->err = 1;
- st->devnum = -1;
+ st->err = 0;
+ st->devnum = MAXINT;
st->percent = -2;
+ st->expected_spares = mdlist->spare_disks;
if (mdlist->spare_group)
st->spare_group = strdup(mdlist->spare_group);
else
} else {
mddev_dev_t dv;
for (dv=devlist ; dv; dv=dv->next) {
+ mddev_ident_t mdlist = conf_get_ident(config, dv->devname);
struct state *st = malloc(sizeof *st);
if (st == NULL)
continue;
st->devname = strdup(dv->devname);
st->utime = 0;
st->next = statelist;
- st->err = 1;
- st->devnum = -1;
+ st->err = 0;
+ st->devnum = MAXINT;
st->percent = -2;
+ st->expected_spares = -1;
st->spare_group = NULL;
+ if (mdlist) {
+ st->expected_spares = mdlist->spare_disks;
+ if (mdlist->spare_group)
+ st->spare_group = strdup(mdlist->spare_group);
+ }
statelist = st;
}
}
while (! finished) {
+ int new_found = 0;
struct state *st;
if (mdstat)
free_mdstat(mdstat);
- mdstat = mdstat_read();
+ mdstat = mdstat_read(oneshot?0:1);
for (st=statelist; st; st=st->next) {
mdu_array_info_t array;
- struct mdstat_ent *mse;
+ struct mdstat_ent *mse = NULL, *mse2;
char *dev = st->devname;
int fd;
- int i;
+ unsigned int i;
+ if (test)
+ alert("TestMessage", dev, NULL, mailaddr, alert_cmd);
fd = open(dev, O_RDONLY);
if (fd < 0) {
if (!st->err)
close(fd);
continue;
}
- if (st->devnum < 0) {
+ if (array.level != 1 && array.level != 5 && array.level != -4 &&
+ array.level != 6 && array.level != 10) {
+ if (!st->err)
+ alert("DeviceDisappeared", dev, "Wrong-Level",
+ mailaddr, alert_cmd);
+ st->err = 1;
+ close(fd);
+ continue;
+ }
+ if (st->devnum == MAXINT) {
struct stat stb;
if (fstat(fd, &stb) == 0 &&
- (S_IFMT&stb.st_mode)==S_IFBLK)
- st->devnum = MINOR(stb.st_rdev);
+ (S_IFMT&stb.st_mode)==S_IFBLK) {
+ if (major(stb.st_rdev) == MD_MAJOR)
+ st->devnum = minor(stb.st_rdev);
+ else
+ st->devnum = -1- (minor(stb.st_rdev)>>6);
+ }
}
- for (mse = mdstat ; mse ; mse=mse->next)
- if (mse->devnum == st->devnum) {
- mse->devnum = -1; /* flag it as "used" */
- break;
+ for (mse2 = mdstat ; mse2 ; mse2=mse2->next)
+ if (mse2->devnum == st->devnum) {
+ mse2->devnum = MAXINT; /* flag it as "used" */
+ mse = mse2;
}
if (st->utime == array.utime &&
st->err = 0;
continue;
}
+ if (st->utime == 0 && /* new array */
+ mse && /* is in /proc/mdstat */
+ mse->pattern && strchr(mse->pattern, '_') /* degraded */
+ )
+ alert("DegradedArray", dev, NULL, mailaddr, alert_cmd);
+
+ if (st->utime == 0 && /* new array */
+ st->expected_spares > 0 &&
+ array.spare_disks < st->expected_spares)
+ alert("SparesMissing", dev, NULL, mailaddr, alert_cmd);
if (mse &&
st->percent == -1 &&
mse->percent >= 0)
alert(percentalerts[mse->percent/20],
dev, NULL, mailaddr, alert_cmd);
+ if (mse &&
+ mse->percent == -1 &&
+ st->percent >= 0)
+ alert("RebuildFinished", dev, NULL, mailaddr, alert_cmd);
+
if (mse)
st->percent = mse->percent;
if (ioctl(fd, GET_DISK_INFO, &disc)>= 0) {
newstate = disc.state;
dv = map_dev(disc.major, disc.minor);
- } else if (mse && i < strlen(mse->pattern))
+ } else if (mse && mse->pattern && i < strlen(mse->pattern))
switch(mse->pattern[i]) {
case 'U': newstate = 6 /* ACTIVE/SYNC */; break;
case '_': newstate = 0; break;
}
change = newstate ^ st->devstate[i];
if (st->utime && change && !st->err) {
- if (i < array.raid_disks &&
+ if (i < (unsigned)array.raid_disks &&
(((newstate&change)&(1<<MD_DISK_FAULTY)) ||
((st->devstate[i]&change)&(1<<MD_DISK_ACTIVE)) ||
((st->devstate[i]&change)&(1<<MD_DISK_SYNC)))
)
alert("Fail", dev, dv, mailaddr, alert_cmd);
- else if (i>=array.raid_disks &&
+ else if (i >= (unsigned)array.raid_disks &&
(disc.major || disc.minor) &&
- st->devid[i] == MKDEV(disc.major, disc.minor) &&
+ st->devid[i] == makedev(disc.major, disc.minor) &&
((newstate&change)&(1<<MD_DISK_FAULTY))
)
alert("FailSpare", dev, dv, mailaddr, alert_cmd);
- else if (i < array.raid_disks &&
+ else if (i < (unsigned)array.raid_disks &&
(((st->devstate[i]&change)&(1<<MD_DISK_FAULTY)) ||
((newstate&change)&(1<<MD_DISK_ACTIVE)) ||
((newstate&change)&(1<<MD_DISK_SYNC)))
alert("SpareActive", dev, dv, mailaddr, alert_cmd);
}
st->devstate[i] = disc.state;
- st->devid[i] = MKDEV(disc.major, disc.minor);
+ st->devid[i] = makedev(disc.major, disc.minor);
}
close(fd);
st->active = array.active_disks;
if (scan) {
struct mdstat_ent *mse;
for (mse=mdstat; mse; mse=mse->next)
- if (mse->devnum > 0) {
+ if (mse->devnum != MAXINT &&
+ (strcmp(mse->level, "raid1")==0 ||
+ strcmp(mse->level, "raid5")==0 ||
+ strcmp(mse->level, "multipath")==0)
+ ) {
struct state *st = malloc(sizeof *st);
+ mdu_array_info_t array;
+ int fd;
if (st == NULL)
continue;
st->devname = strdup(get_md_name(mse->devnum));
+ if ((fd = open(st->devname, O_RDONLY)) < 0 ||
+ ioctl(fd, GET_ARRAY_INFO, &array)< 0) {
+ /* no such array */
+ if (fd >=0) close(fd);
+ free(st->devname);
+ free(st);
+ continue;
+ }
+ close(fd);
st->utime = 0;
st->next = statelist;
st->err = 1;
st->devnum = mse->devnum;
st->percent = -2;
st->spare_group = NULL;
+ st->expected_spares = -1;
statelist = st;
alert("NewArray", st->devname, NULL, mailaddr, alert_cmd);
+ new_found = 1;
}
}
/* If an array has active < raid && spare == 0 && spare_group != NULL
close(fd2);
}
}
-
- sleep(period);
+ if (!new_found) {
+ if (oneshot)
+ break;
+ else
+ mdstat_wait(period);
+ }
+ test = 0;
}
+ if (pidfile)
+ unlink(pidfile);
return 0;
}
exit(2);
}
}
- if (mailaddr && strncmp(event, "Fail", 4)==0) {
+ if (mailaddr &&
+ (strncmp(event, "Fail", 4)==0 ||
+ strncmp(event, "Test", 4)==0 ||
+ strncmp(event, "Degrade", 7)==0)) {
FILE *mp = popen(Sendmail, "w");
if (mp) {
char hname[256];
fprintf(mp, "A %s event had been detected on md device %s.\n\n", event, dev);
if (disc)
- fprintf(mp, "It could be related to componenet device %s.\n\n", disc);
+ fprintf(mp, "It could be related to component device %s.\n\n", disc);
fprintf(mp, "Faithfully yours, etc.\n");
fclose(mp);