3 # check _physical_ disk status of disks on HP smart array controllers
6 # does _not_ check raid status. use arrayprobe for that.
8 # Copyright (c) 2008,2009,2010,2011 Peter Palfrader <peter@palfrader.org>
10 # Permission is hereby granted, free of charge, to any person obtaining
11 # a copy of this software and associated documentation files (the
12 # "Software"), to deal in the Software without restriction, including
13 # without limitation the rights to use, copy, modify, merge, publish,
14 # distribute, sublicense, and/or sell copies of the Software, and to
15 # permit persons to whom the Software is furnished to do so, subject to
16 # the following conditions:
18 # The above copyright notice and this permission notice shall be
19 # included in all copies or substantial portions of the Software.
21 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
22 # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
23 # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
24 # NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
25 # LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
26 # OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
27 # WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
44 $SIG{'__DIE__'} = sub {
46 exit $CODE{'UNKNOWN'};
51 my ($os, $host, $kernel, $date, $arch) = uname();
52 if ($kernel =~ m/^3(.*)/) {
53 $cmd = "sudo setarch x86_64 --uname-2.6 hpacucli $cmd";
55 $cmd = "sudo hpacucli $cmd";
57 open(FH, $cmd."|") or die ("Cannot run $cmd: $!");
60 die ("no results from $cmd\n") if (scalar @lines == 0);
66 die "code $newexit not defined\n" unless defined $CODE{$newexit};
68 if ($CODE{$newexit} > $CODE{$EXITCODE}) {
73 my $usage = "$PROGRAM_NAME: Usage: $PROGRAM_NAME [--no-battery] [--ignore-transfer-speed=<pd> [--ignore-transfer-speed=<pd> ...]]\n";
75 Getopt::Long::Configure('bundling');
77 '--help' => \$params->{'help'},
78 '--no-battery' => \$params->{'no-battery'},
79 '--ignore-transfer-speed=s@' => \$params->{'ignore-transfer-speed'},
83 if ($params->{'help'}) {
87 die ($usage) unless (scalar @ARGV == 0);
89 my $ctrlallshow = runcmd("controller all show");
94 if (/in Slot ([0-9a-z]+)/) {
95 push @controllers, $1;
98 die ("Cannot read line '$_' gotten from hpacucli controller all show\n");
101 if (scalar @controllers == 0) {
102 print "UNKNONW: No smartarray controllers found with hpacucli\n";
103 exit $CODE{'UNKNOWN'}
108 for my $slot (sort @controllers) {
109 my $pds = runcmd("controller slot=$slot pd all show");
116 next if (/^\S.*in Slot $slot/);
117 next if /^ *array [A-Z]$/;
118 next if /^ *unassigned/;
119 if (/^ *(array [A-Z]) \(Failed\)$/) {
121 push @{$status{'Failed'}}, $1;
122 } elsif (/^Error: The specified controller does not have any physical drives on it.$/) {
124 } elsif (/^ *physicaldrive (\S+) .* (OK|Predictive Failure|Failed|Rebuilding)(?:, (?:active )?spare)?\)$/) {
127 push @{$status{$status}}, $drive;
128 if ($status eq 'OK') {
129 } elsif ($status eq 'Predictive Failure' ||
130 $status eq 'Rebuilding') {
132 } elsif ($status eq 'Failed') {
137 push @drives, $drive;
139 die ("Cannot read line '$_' gotten from hpacucli controller slot=$slot pd all show\n");
143 # Check that all drives have the proper transfer speed.
144 # sometimes stuff breaks and they fall back to 10mb/sec.
145 for my $drive (@drives) {
146 # skip drives that are known to have failed
147 next if (exists $status{'Failed'} && grep {$drive eq $_} @{$status{'Failed'}});
149 if ($drive =~ /^[0-9]+:[0-9]+$/) { # scsi drives
151 } elsif ($drive =~ /^[0-9]+[EI]:[0-9]+:[0-9]+$/) { # SAS
153 } elsif ($drive =~ /^[0-9]+[C]:[0-9]+:[0-9]+$/) { # New 6GBPS SAS
156 # I'm not going to run pass arguments of unknown form to the shell..
157 warn ("Unknown diskdrive ID $drive\n");
161 my $pd = runcmd("controller slot=$slot pd $drive show");
162 while (defined $pd->[0] && !($pd->[0] =~ /physicaldrive/)) {
168 if (m/^\s*(.*?):\s*(.*?)\s*$/) {
175 if ($type eq 'SCSI') {
176 $key = 'Transfer Speed';
177 if (!defined $value{'Transfer Mode'}) {
179 push @{$status{'unknown transfer mode'}}, $drive;
181 } elsif ($value{'Transfer Mode'} eq 'Ultra 3 Wide') {
182 $expected = '160 MB/Sec';
183 } elsif ($value{'Transfer Mode'} eq 'Ultra 320 Wide') {
184 $expected = '320 MB/Sec';
187 push @{$status{'unknown transfer mode'}}, $drive."(".$value{'Transfer Mode'}.")";
190 } elsif ($type eq 'SAS' || $type eq 'SAS+') {
191 $key = 'PHY Transfer Rate';
192 if ($value{'PHY Count'} eq '2') {
193 if (defined($value{'Redundant Path(s)'})) {
194 $expected = [ '3.0GBPS, 3.0GBPS', '6.0GBPS, 6.0GBPS' ];
196 $expected = [ '3.0GBPS, Unknown', 'Unknown, 3.0GBPS',
197 '6.0GBPS, Unknown', 'Unknown, 6.0GBPS' ];
200 $expected = [ '3.0GBPS', '6.0GBPS' ];
203 warn "Should not be here. Do not know what to do with type '$type'\n";
207 if ($params->{'ignore-transfer-speed'}) {
208 if (grep { $drive eq $_ } @{$params->{'ignore-transfer-speed'}}) {
209 push @{$status{'ignored transfer speed'}}, $drive."(".$value{$key}.")";
213 if (!defined $value{$key}) {
215 push @{$status{'unknown transfer speed'}}, $drive;
216 } elsif (ref($expected) eq 'ARRAY') {
217 if (scalar(grep { $value{$key} eq $_ } @$expected) == 0) {
219 push @{$status{'bad transfer speed'}}, $drive."(".$value{$key}.")";
221 } elsif ($value{$key} ne $expected) {
223 push @{$status{'bad transfer speed'}}, $drive."(".$value{$key}.")";
227 if ($nodrives && scalar keys %status > 0) {
228 push @resultstr, "Slot $slot: have no drives but status results?";
231 } elsif ($nodrives) {
232 push @resultstr, "Slot $slot: no drives";
236 my $cst = runcmd("controller slot=$slot show status");
240 next if (/^\S.*in Slot $slot/);
241 if (/^ *(.*) Status: (.*)$/) {
244 push @{$status{$status}}, $system;
245 if ($status ne 'OK') {
246 next if ($params->{'no-battery'} && $system eq 'Cache');
247 next if ($params->{'no-battery'} && $system eq 'Battery/Capacitor');
251 die ("Cannot read line '$_' gotten from hpacucli controller slot=$slot show status\n");
255 my $status = join(" - ", (map { $_.": ".join(", ", @{$status{$_}}) } keys %status));
256 push @resultstr, "Slot $slot: $status";
259 print "$EXITCODE: ", join(" --- ", @resultstr), "\n";
260 exit $CODE{$EXITCODE};