X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=Monitor.c;h=c94bfd85d6e188d8710c6cc764a12751cfd2cac1;hp=fa0d7088df03425535eb7323c13af1ae21e0e7f2;hb=4f589ad0c5e95154dc6ffabfc2e9493feabe9ab6;hpb=b83d95f362ff41e5ce63baa3bbcf0a76ea1f15aa diff --git a/Monitor.c b/Monitor.c index fa0d7088..c94bfd85 100644 --- a/Monitor.c +++ b/Monitor.c @@ -1,7 +1,7 @@ /* * mdadm - manage Linux "md" devices aka RAID arrays. * - * Copyright (C) 2001-2002 Neil Brown + * Copyright (C) 2001-2006 Neil Brown * * * This program is free software; you can redistribute it and/or modify @@ -32,8 +32,11 @@ #include "md_u.h" #include #include +#include +#include -static void alert(char *event, char *dev, char *disc, char *mailaddr, char *cmd); +static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom, + char *cmd, int dosyslog); static char *percentalerts[] = { "RebuildStarted", @@ -43,10 +46,15 @@ static char *percentalerts[] = { "Rebuild80", }; +/* The largest number of disks current arrays can manage is 384 + * This really should be dynamically, but that will have to wait + * At least it isn't MD_SB_DISKS. + */ +#define MaxDisks 384 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, + int dosyslog, char *config, int test, char* pidfile) { /* * Every few seconds, scan every md device looking for changes @@ -97,20 +105,62 @@ int Monitor(mddev_dev_t devlist, int err; char *spare_group; int active, working, failed, spare, raid; - int devstate[MD_SB_DISKS]; - int devid[MD_SB_DISKS]; + int expected_spares; + int devstate[MaxDisks]; + int devid[MaxDisks]; int percent; struct state *next; } *statelist = NULL; int finished = 0; struct mdstat_ent *mdstat = NULL; + char *mailfrom = 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); + } + mailfrom = conf_get_mailfrom(config); + + 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); @@ -121,9 +171,10 @@ int Monitor(mddev_dev_t devlist, 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 @@ -133,40 +184,50 @@ int Monitor(mddev_dev_t devlist, } 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, 0); 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, mailfrom, alert_cmd, dosyslog); fd = open(dev, O_RDONLY); if (fd < 0) { if (!st->err) alert("DeviceDisappeared", dev, NULL, - mailaddr, alert_cmd); + mailaddr, mailfrom, alert_cmd, dosyslog); /* fprintf(stderr, Name ": cannot open %s: %s\n", dev, strerror(errno)); */ st->err=1; @@ -175,24 +236,37 @@ int Monitor(mddev_dev_t devlist, if (ioctl(fd, GET_ARRAY_INFO, &array)<0) { if (!st->err) alert("DeviceDisappeared", dev, NULL, - mailaddr, alert_cmd); + mailaddr, mailfrom, alert_cmd, dosyslog); /* fprintf(stderr, Name ": cannot get array info for %s: %s\n", dev, strerror(errno)); */ st->err=1; 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, mailfrom, alert_cmd, dosyslog); + 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 && @@ -206,57 +280,80 @@ int Monitor(mddev_dev_t devlist, 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, mailfrom, alert_cmd, dosyslog); + + if (st->utime == 0 && /* new array */ + st->expected_spares > 0 && + array.spare_disks < st->expected_spares) + alert("SparesMissing", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog); if (mse && st->percent == -1 && mse->percent >= 0) - alert("RebuildStarted", dev, NULL, mailaddr, alert_cmd); + alert("RebuildStarted", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog); if (mse && st->percent >= 0 && mse->percent >= 0 && (mse->percent / 20) > (st->percent / 20)) alert(percentalerts[mse->percent/20], - dev, NULL, mailaddr, alert_cmd); + dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog); + + if (mse && + mse->percent == -1 && + st->percent >= 0) + alert("RebuildFinished", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog); if (mse) st->percent = mse->percent; - - for (i=0; i= 0) { + if (i > array.raid_disks + array.nr_disks) { + newstate = 0; + disc.major = disc.minor = 0; + } else 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)) + dv = map_dev(disc.major, disc.minor, 1); + } else if (mse && mse->pattern && i < strlen(mse->pattern)) { switch(mse->pattern[i]) { case 'U': newstate = 6 /* ACTIVE/SYNC */; break; case '_': newstate = 0; break; } + disc.major = disc.minor = 0; + } + if (dv == NULL && st->devid[i]) + dv = map_dev(major(st->devid[i]), + minor(st->devid[i]), 1); 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<devstate[i]&change)&(1<devstate[i]&change)&(1<=array.raid_disks && + alert("Fail", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog); + 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<devstate[i]&change)&(1<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; @@ -271,19 +368,36 @@ int Monitor(mddev_dev_t devlist, 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); + alert("NewArray", st->devname, NULL, mailaddr, mailfrom, alert_cmd, dosyslog); + new_found = 1; } } /* If an array has active < raid && spare == 0 && spare_group != NULL @@ -311,7 +425,7 @@ int Monitor(mddev_dev_t devlist, if (fd2>=0) close(fd2); continue; } - for (d=st2->raid; draid; d < MaxDisks; d++) { if (st2->devid[d] > 0 && st2->devstate[d] == 0) { dev = st2->devid[d]; @@ -323,7 +437,7 @@ int Monitor(mddev_dev_t devlist, (unsigned long)dev) == 0) { if (ioctl(fd1, HOT_ADD_DISK, (unsigned long)dev) == 0) { - alert("MoveSpare", st->devname, st2->devname, mailaddr, alert_cmd); + alert("MoveSpare", st->devname, st2->devname, mailaddr, mailfrom, alert_cmd, dosyslog); close(fd1); close(fd2); break; @@ -335,15 +449,25 @@ int Monitor(mddev_dev_t devlist, close(fd2); } } - - sleep(period); + if (!new_found) { + if (oneshot) + break; + else + mdstat_wait(period); + } + test = 0; } + if (pidfile) + unlink(pidfile); return 0; } -static void alert(char *event, char *dev, char *disc, char *mailaddr, char *cmd) +static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom, char *cmd, + int dosyslog) { + int priority; + if (!cmd && !mailaddr) { time_t now = time(0); @@ -362,13 +486,21 @@ static void alert(char *event, char *dev, char *disc, char *mailaddr, char *cmd) exit(2); } } - if (mailaddr && strncmp(event, "Fail", 4)==0) { + if (mailaddr && + (strncmp(event, "Fail", 4)==0 || + strncmp(event, "Test", 4)==0 || + strncmp(event, "Spares", 6)==0 || + strncmp(event, "Degrade", 7)==0)) { FILE *mp = popen(Sendmail, "w"); if (mp) { + FILE *mdstat; char hname[256]; gethostname(hname, sizeof(hname)); signal(SIGPIPE, SIG_IGN); - fprintf(mp, "From: " Name " monitoring \n"); + if (mailfrom) + fprintf(mp, "From: %s\n", mailfrom); + else + fprintf(mp, "From: " Name " monitoring \n"); fprintf(mp, "To: %s\n", mailaddr); fprintf(mp, "Subject: %s event on %s:%s\n\n", event, dev, hname); @@ -378,12 +510,45 @@ static void alert(char *event, char *dev, char *disc, char *mailaddr, char *cmd) 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"); + + mdstat = fopen("/proc/mdstat", "r"); + if (mdstat) { + char buf[8192]; + int n; + fprintf(mp, "\nP.S. The /proc/mdstat file current contains the following:\n\n"); + while ( (n=fread(buf, 1, sizeof(buf), mdstat)) > 0) + fwrite(buf, 1, n, mp); + fclose(mdstat); + } fclose(mp); } } - /* FIXME log the event to syslog maybe */ + + /* log the event to syslog maybe */ + if (dosyslog) { + /* Log at a different severity depending on the event. + * + * These are the critical events: */ + if (strncmp(event, "Fail", 4)==0 || + strncmp(event, "Degrade", 7)==0 || + strncmp(event, "DeviceDisappeared", 17)==0) + priority = LOG_CRIT; + /* Good to know about, but are not failures: */ + else if (strncmp(event, "Rebuild", 7)==0 || + strncmp(event, "MoveSpare", 9)==0 || + strncmp(event, "Spares", 6) != 0) + priority = LOG_WARNING; + /* Everything else: */ + else + priority = LOG_INFO; + + if (disc) + syslog(priority, "%s event detected on md device %s, component device %s", event, dev, disc); + else + syslog(priority, "%s event detected on md device %s", event, dev); + } }