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.
43 $SIG{'__DIE__'} = sub {
45 exit $CODE{'UNKNOWN'};
50 $cmd = "sudo hpssacli $cmd";
51 open(FH, $cmd."|") or die ("Cannot run $cmd: $!");
54 die ("no results from $cmd\n") if (scalar @lines == 0);
60 die "code $newexit not defined\n" unless defined $CODE{$newexit};
62 if ($CODE{$newexit} > $CODE{$EXITCODE}) {
67 my $usage = "$PROGRAM_NAME: Usage: $PROGRAM_NAME [--no-battery] [--ignore-cache] [--ignore-controller=<regex>] [--no-controller-ok] [--ignore-transfer-speed=<pd> [--ignore-transfer-speed=<pd> ...]]\n";
69 Getopt::Long::Configure('bundling');
71 '--help' => \$params->{'help'},
72 '--no-battery' => \$params->{'no-battery'},
73 '--no-controller-ok' => \$params->{'no-controller-ok'},
74 '--ignore-cache' => \$params->{'ignore-cache'},
75 '--ignore-controller=s' => \$params->{'ignore-controller'},
76 '--ignore-transfer-speed=s@' => \$params->{'ignore-transfer-speed'},
80 if ($params->{'help'}) {
84 die ($usage) unless (scalar @ARGV == 0);
86 my $ctrlallshow = runcmd("controller all show");
91 next if ($params->{'ignore-controller'} && /$params->{'ignore-controller'}/);
92 if (/in Slot ([0-9a-z]+)/) {
93 push @controllers, $1;
96 die ("Cannot read line '$_' gotten from hpssacli controller all show\n");
99 if (scalar @controllers == 0) {
100 if ($params->{'no-controller-ok'}) {
101 print "No smartarray controllers found with hpssacli\n";
104 print "UNKNOWN: No smartarray controllers found with hpssacli\n";
105 exit $CODE{'UNKNOWN'}
111 for my $slot (sort @controllers) {
117 my $ldallshow = runcmd("controller slot=$slot ld all show");
122 next if (/^\S.*in Slot $slot/);
123 next if /^ *array [A-Z]$/;
124 if (/logicaldrive ([0-9a-z]+)/) {
125 push @logicaldrives, $1;
127 } elsif (/^Error: The specified device does not have any logical drives.$/) {
130 die ("Cannot read line '$_' gotten from hpssacli controller slot = $slot logicaldrive all show\n");
134 # check logicaldrives
135 for my $logicaldrive (sort @logicaldrives) {
136 my $lds = runcmd("controller slot=$slot ld $logicaldrive show");
140 if (/^ *Parity Initialization Status: (Initialization Completed|Initialization Failed|Rebuilding)$/) {
142 if ($status eq 'Initialization Completed') {
143 push @{$status{'OK'}}, "Parity LD$logicaldrive";
144 } elsif ($status eq 'Rebuilding') {
145 push @{$status{'Failed'}}, "Parity LD$logicaldrive";
147 } elsif ($status eq 'Initialization Failed') {
148 push @{$status{'Failed'}}, "Parity LD$logicaldrive";
154 if (/^ *LD Acceleration Method: (.*)$/) {
156 # can at least be "Controller Cache" or HP SSD Smart Path", both OK
157 if ($status eq 'All disabled') {
158 push @{$status{'Acceleration method'}}, "LD$logicaldrive disabled";
165 if (!$nodrives && scalar @logicaldrives == 0) {
166 push @resultstr, "Slot $slot: unexpectedly, found no logical drives in list.";
168 } elsif ($nodrives && scalar keys %status > 0) {
169 push @resultstr, "Slot $slot: have no logical drives but status results?";
172 } elsif ($nodrives) {
173 push @resultstr, "Slot $slot: no logical drives";
177 my $pds = runcmd("controller slot=$slot pd all show");
181 next if (/^\S.*in Slot $slot/);
182 next if /^ *array [A-Z]$/;
183 next if /^ *unassigned/;
184 if (/^ *HBA Drives/) {
185 # HBA mode implies no logical drives, thus reset the "drives found" check and proceed with
186 # checking physical drives.
190 if (/^ *(array [A-Z]) \(Failed\)$/) {
192 push @{$status{'Failed'}}, $1;
193 } elsif (/^Error: The specified controller does not have any physical drives on it.$/) {
195 } elsif (/^ *physicaldrive (\S+) .* (OK|Predictive Failure|Failed|Rebuilding)(?:, (?:active )?spare.*)?\)$/) {
198 push @{$status{$status}}, $drive;
199 if ($status eq 'OK') {
200 } elsif ($status eq 'Predictive Failure' ||
201 $status eq 'Rebuilding') {
203 } elsif ($status eq 'Failed') {
208 push @drives, $drive;
210 die ("Cannot read line '$_' gotten from hpssacli controller slot=$slot pd all show\n");
214 # Check that all drives have the proper transfer speed.
215 # sometimes stuff breaks and they fall back to 10mb/sec.
216 for my $drive (@drives) {
217 # skip drives that are known to have failed
218 next if (exists $status{'Failed'} && grep {$drive eq $_} @{$status{'Failed'}});
220 if ($drive =~ /^[0-9]+:[0-9]+$/) { # scsi drives
222 } elsif ($drive =~ /^[0-9]+[EI]:[0-9]+:[0-9]+$/) { # SAS
224 } elsif ($drive =~ /^[0-9]+[C]:[0-9]+:[0-9]+$/) { # New 6GBPS SAS
227 # I'm not going to run pass arguments of unknown form to the shell..
228 warn ("Unknown diskdrive ID $drive\n");
232 my $pd = runcmd("controller slot=$slot pd $drive show");
233 while (defined $pd->[0] && !($pd->[0] =~ /physicaldrive/)) {
239 if (m/^\s*(.*?):\s*(.*?)\s*$/) {
246 if ($type eq 'SCSI') {
247 $key = 'Transfer Speed';
248 if (!defined $value{'Transfer Mode'}) {
250 push @{$status{'unknown transfer mode'}}, $drive;
252 } elsif ($value{'Transfer Mode'} eq 'Ultra 3 Wide') {
253 $expected = '160 MB/Sec';
254 } elsif ($value{'Transfer Mode'} eq 'Ultra 320 Wide') {
255 $expected = '320 MB/Sec';
258 push @{$status{'unknown transfer mode'}}, $drive."(".$value{'Transfer Mode'}.")";
261 } elsif ($type eq 'SAS' || $type eq 'SAS+') {
262 $key = 'PHY Transfer Rate';
263 if ($value{'Interface Type'} eq 'SATA') {
264 $expected = [ '1.5Gbps', '3.0Gbps', '6.0Gbps' ];
265 } elsif ($value{'PHY Count'} eq '2') {
266 if (defined($value{'Redundant Path(s)'})) {
267 $expected = [ '3.0GBPS, 3.0GBPS', '6.0GBPS, 6.0GBPS',
268 '12.0GBPS, 12.0GBPS' ];
270 $expected = [ '3.0GBPS, Unknown', 'Unknown, 3.0GBPS',
271 '6.0GBPS, Unknown', 'Unknown, 6.0GBPS',
272 '12.0GBPS, Unknown', 'Unknown, 12.0GBPS' ];
275 $expected = [ '3.0GBPS', '6.0GBPS', '12.0GBPS' ];
278 warn "Should not be here. Do not know what to do with type '$type'\n";
282 if ($params->{'ignore-transfer-speed'}) {
283 if (grep { $drive eq $_ } @{$params->{'ignore-transfer-speed'}}) {
284 push @{$status{'ignored transfer speed'}}, $drive."(".$value{$key}.")";
288 if (!defined $value{$key}) {
290 push @{$status{'unknown transfer speed'}}, $drive;
291 } elsif (ref($expected) eq 'ARRAY') {
292 if (scalar(grep { uc($value{$key}) eq uc($_) } @$expected) == 0) {
294 push @{$status{'bad transfer speed'}}, $drive."(".$value{$key}.")";
296 } elsif (uc($value{$key}) ne uc($expected)) {
298 push @{$status{'bad transfer speed'}}, $drive."(".$value{$key}.")";
302 if ($nodrives && scalar keys %status > 0) {
303 push @resultstr, "Slot $slot: have no drives but status results?";
306 } elsif ($nodrives) {
307 push @resultstr, "Slot $slot: no drives";
311 my $cst = runcmd("controller slot=$slot show detail");
315 next if (/^\S.*in Slot $slot/);
316 if (/^ *(Controller|Cache|Battery\/Capacitor) Status: (.*)$/) {
320 if ($system eq 'Cache') {
323 # - 'Not Configured' (for e.g. HP SSD Smart Path)
324 # - 'Permanently Disabled'
326 next if $status =~ /^(OK|Not Configured)$/;
327 if ($params->{'ignore-cache'}) {
328 push @freetext, "$system: $status (ignored)";
333 push @freetext, "$system: $status";
334 if ($status ne 'OK') {
335 next if ($params->{'no-battery'} && $system eq 'Battery/Capacitor');
338 } elsif (/^ *(Cache Status Details): (Cable Error)/) {
341 } elsif (/^ *(Battery\/Capacitor Count): (.*)/) {
342 next if $params->{'no-battery'} || int($2) > 0;
343 push @freetext, "Battery count: $2";
348 my $status = join(" - ", ((map { $_.": ".join(", ", @{$status{$_}}) } keys %status), @freetext));
350 push @resultstr, "Slot $slot: $status";
353 print "$EXITCODE: ", join(" --- ", @resultstr), "\n";
354 exit $CODE{$EXITCODE};