Don't WARN when device is being checked, only when it's rebuilding

tags/zabbix-agent-addons-0.2.24-1
Daniel Berteaud 7 years ago
parent 0c9fad57bd
commit b4f79b94e4
  1. 148
      zabbix_scripts/check_raid_mdadm

@ -27,20 +27,19 @@ use Getopt::Long;
#
# unused devices: <none>
my $file = "/proc/mdstat";
my $file = "/proc/mdstat";
my $device = "all";
# Get command line options.
GetOptions ('file=s' => \$file,
'device=s' => \$device,
'help' => sub { &usage() } );
GetOptions (
'file=s' => \$file,
'device=s' => \$device,
'help' => sub { &usage() }
);
## Strip leading "/dev/" from --device in case it has been given
$device =~ s/^\/dev\///;
## Return codes for Nagios
my %ERRORS=('OK'=>0,'WARNING'=>1,'CRITICAL'=>2,'UNKNOWN'=>3,'DEPENDENT'=>4);
## This is a global return value - set to the worst result we get overall
my $retval = 0;
@ -50,78 +49,85 @@ my $result = 'OK';
open FILE, "< $file" or die "Can't open $file : $!";
while (<FILE>) {
next if ! /^(md\d+)+\s*:/;
next if $device ne "all" and $device ne $1;
my $dev = $1;
push @raids, $dev;
my @array = split(/ /);
$devs_total{$dev} = 0;
my $devs_up = 0;
my $missing = 0;
for $_ (@array) {
$level{$dev} = $1 if /^(raid\d+)$/;
next if ! /(\w+)\[\d+\](\(.\))*/;
$devs_total{$dev}++;
if ($2 eq "(F)") {
$failed_devs{$dev} .= "$1,";
}
elsif ($2 eq "(S)") {
$spare_devs{$dev} .= "$1,";
}
else {
$active_devs{$dev} .= "$1,";
$devs_up++;
}
}
if (! defined($active_devs{$dev})) { $active_devs{$dev} = "none"; }
else { $active_devs{$dev} =~ s/,$//; }
if (! defined($spare_devs{$dev})) { $spare_devs{$dev} = "none"; }
else { $spare_devs{$dev} =~ s/,$//; }
if (! defined($failed_devs{$dev})) { $failed_devs{$dev} = "none"; }
else { $failed_devs{$dev} =~ s/,$//; }
$_ = <FILE>;
/(\d+)\ blocks\ (.*)(\[.*\])\s?$/;
$size{$dev} = int($1/1024);
#print "$3\n";
$missing = 1 if ($3 =~ m/_/);
if ($size{$dev} > 1024){
$size{$dev} = int($size{$dev}/1024)."GB";
}
else{
$size{$dev} .= "MB";
}
$_ = <FILE>;
if (($devs_total{$dev} > $devs_up) || ($failed_devs{$dev} ne "none") || (($missing) && (!/recovery/))) {
$status{$dev} = "Degraded";
$result = "CRITICAL";
$retval = $ERRORS{"CRITICAL"};
}
else {
$status{$dev} = "Optimal";
}
if (/recovery|resync/){
$status{$dev} = "Rebuilding";
if ($result eq "OK"){
$result = "WARNING";
$retval = $ERRORS{"WARNING"};
}
}
next if ! /^(md\d+)+\s*:/;
next if $device ne "all" and $device ne $1;
my $dev = $1;
push @raids, $dev;
my @array = split(/ /);
$devs_total{$dev} = 0;
my $devs_up = 0;
my $missing = 0;
for $_ (@array) {
$level{$dev} = $1 if /^(raid\d+)$/;
next if ! /(\w+)\[\d+\](\(.\))*/;
$devs_total{$dev}++;
if ($2 eq "(F)") {
$failed_devs{$dev} .= "$1,";
}
elsif ($2 eq "(S)") {
$spare_devs{$dev} .= "$1,";
}
else {
$active_devs{$dev} .= "$1,";
$devs_up++;
}
}
if (! defined($active_devs{$dev})){
$active_devs{$dev} = "none";
}
else {
$active_devs{$dev} =~ s/,$//;
}
if (! defined($spare_devs{$dev})){
$spare_devs{$dev} = "none";
}
else {
$spare_devs{$dev} =~ s/,$//;
}
if (! defined($failed_devs{$dev})){
$failed_devs{$dev} = "none";
}
else {
$failed_devs{$dev} =~ s/,$//;
}
$_ = <FILE>;
/(\d+)\ blocks\ (.*)(\[.*\])\s?$/;
$size{$dev} = int($1/1024);
$missing = 1 if ($3 =~ m/_/);
if ($size{$dev} > 1024){
$size{$dev} = int($size{$dev}/1024)."GB";
}
else{
$size{$dev} .= "MB";
}
$_ = <FILE>;
if (($devs_total{$dev} > $devs_up) || ($failed_devs{$dev} ne "none") || (($missing) && (!/recovery/))) {
$status{$dev} = "Degraded";
$result = "CRITICAL";
}
else {
$status{$dev} = "Optimal";
}
if (/(recovery|resync)\s*=\s*\d{1,2}(\.\d)?%/){
$status{$dev} = "Rebuilding";
if ($result eq "OK"){
$result = "WARNING";
}
}
}
print "$result: ";
foreach my $raid (@raids){
print "$raid:$level{$raid}:$devs_total{$raid} drives:$size{$raid}:$status{$raid} ";
print "$raid:$level{$raid}:$devs_total{$raid} drives:$size{$raid}:$status{$raid} ";
}
print "\n";
close FILE;
exit $retval;
exit 0;
# =====
sub usage()
{
sub usage(){
printf("
Check status of Linux SW RAID

Loading…
Cancel
Save