node_reboot.in 6.1 KB
Newer Older
1 2 3 4 5 6 7
#!/usr/bin/perl -wT
use English;
use Getopt::Std;

#
# Reboot a node. Will power cycle the node as a last resort.
#
8
# usage: node_reboot [-d] [-f] node [node ...]
9 10 11 12 13
#        Exit value is 0 if all nodes reboot okay, or the number of nodes
#        could not be rebooted.
#
sub usage()
{
14
    print STDOUT "Usage: node_reboot [-d] [-f] node [node ...]\n" .
15 16
	"Use the -d option to turn on debugging\n" .
	"Use the -f option to shoot the node in the head\n";
17 18
    exit(-1);
}
19
my  $optlist = "df";
20 21 22 23 24

#
# Configure variables
#
my $TB		= "@prefix@";
25 26 27 28 29 30

#
# Load the Testbed support stuff. 
#
push(@INC, "$TB/lib");
require libdb;
31

Robert Ricci's avatar
Robert Ricci committed
32
my $ssh		= "$TB/bin/sshtb -n";
33 34 35 36 37 38 39
my $power	= "$TB/bin/power";
my $ipod	= "$TB/sbin/ipod";
my $ping	= "/sbin/ping";
my %pids	= ();
my @row;
my @nodes       = ();
my $debug       = 0;
40
my $force       = 0;
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68
my $failed      = 0;

# un-taint path
$ENV{'PATH'} = '/bin:/sbin:/usr/bin:/usr/local/bin';
delete @ENV{'IFS', 'CDPATH', 'ENV', 'BASH_ENV'};

# Turn off line buffering on output
$| = 1; 

#
# We don't want to run this script unless its the real version.
#
if ($EUID != 0) {
    die("Must be root! Maybe its a development version?");
}

#
# Parse command arguments. Once we return from getopts, all that should
# left are the required arguments.
#
%options = ();
if (! getopts($optlist, \%options)) {
    usage();
}
if (@ARGV == 0) {
    usage();
}
if (defined($options{"d"})) {
69 70 71 72
    $debug = 1;
}
if (defined($options{"f"})) {
    $force = 1;
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90
}

# Untaint the nodes.
foreach my $node ( @ARGV ) {
    if ($node =~ /^([-\@\w]+)$/) {
	$node = $1;
    }
    else {
	die("Bad node name: $node.");
    }
    
    push(@nodes, $node);
}

#
# Figure out who called us. Root and admin types can do whatever they
# want. Normal users can reboot nodes in their experiment.
#
91 92 93 94
if ($UID && !TBAdmin($UID)) {
    foreach my $node (@nodes) {
	if (! NodeAccessCheck(\$node)) {
	    die("You do not have permission to reboot $node\n");
95 96 97 98
	}
    }
}

99 100 101 102 103 104 105 106 107
#
# In force mode, just call the power program for the whole bunch and
# be done with it.
#
if ($force) {
    system("$power cycle @nodes");
    exit $? >> 8;
}

108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158
#
# Fire off a reboot process so that we can overlap them all.
# We need the pid so we can wait for them all before preceeding.
#
foreach my $node ( @nodes ) {
    $mypid = RebootNode($node);
    $pids{$node} = $mypid;
}

#
# Wait for all the reboot children to exit before continuing.
#
foreach my $node ( @nodes ) {
    my $mypid     = $pids{$node};

    waitpid($mypid, 0);
    if ($?) {
	$failed++;
	print STDERR "Reboot of node $node failed!\n";
    }
    else {
	print STDOUT "$node rebooting ...\n";
    }
}

if ($debug && $failed) {
    print STDERR "$failed nodes could not be rebooted\n";
}
exit $failed;

#
# Reboot a node in a child process. Return the pid to the parent so
# that it can wait on all the children later.
# 
sub RebootNode {
    local($pc) = @_;
    local($status, $syspid, $mypid);

    print STDOUT "Rebooting $pc ...\n";

    $mypid = fork();
    if ($mypid) {
	return $mypid;
    }

    #
    # See if the machine is pingable. If its not pingable, then we just
    # power cycle the machine rather than wait for ssh to time out.
    #
    # ping returns 0 if any packets make it through. 
    #
159
    if (! DoesPing($pc)) {
160 161 162 163 164 165 166 167 168 169 170
	print STDERR "$pc appears to be dead. Power cycling ...\n" if $debug;
	if (PowerCycle($pc)) {
	    exit(-1);
	}
	exit(0);
    }

    #
    # Machine is pingable at least. Try to reboot it gracefully,
    # or power cycle anyway if that does not work. 
    #
171
    print STDERR "Trying ssh reboot of $pc ...\n" if $debug;
172 173 174 175 176 177 178 179 180

    #
    # Run an ssh command in a child process, protected by an alarm to
    # ensure that the ssh is not hung up forever if the machine is in
    # some funky state.
    # 
    $syspid = fork();
    if ($syspid) {
	local $SIG{ALRM} = sub { kill("TERM", $syspid); };
181
	alarm 20;
182 183 184 185 186 187 188 189 190 191
	waitpid($syspid, 0);
	alarm 0;

	#
	# The ssh can return non-zero exit status, but still have worked.
	# FreeBSD for example.
	#
	print STDERR "reboot returned $?.\n" if $debug;
    
	#
192 193
	# If either ssh is not running or it timed out,
	# send it a ping of death.
194
	# 
195 196 197 198 199 200 201 202
	if ($? == 256 || $? == 15) {
	    if ($? == 256) {
		print STDERR "$pc is not running sshd.\n" if $debug;
	    } else {
		print STDERR "$pc is wedged.\n" if $debug;
	    }
	    print STDERR "Trying Ping-of-Death on $pc ...\n" if $debug;

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
	    system("$ipod $pc");
	}
    }
    else {
	# Must change our real UID to root so that ssh will work.
	$UID = 0;
	exec("$ssh $pc /sbin/reboot");
	exit(0);
    }

    #
    # Okay, before we power cycle lets really make sure. We wait a while
    # for it to stop responding to pings, and if it never goes silent,
    # punch the power button.
    #
    if (WaitTillDead($pc) == 0) {
	exit(0);
    }
	
    print STDERR "$pc is still running. Power cycling ...\n" if $debug;
    if (PowerCycle($pc)) {
	exit(-1);
    }
    exit(0);
}

#
# Power cycle a PC using the testbed power program.
#
sub PowerCycle {
    local($pc) = @_;

    system("$power cycle $pc");
    return $? >> 8;
}

#
# Wait until a machine stops returning ping packets.
# 
sub WaitTillDead {
    local($pc) = @_;
    local($status);

    print STDERR "Waiting for $pc to die off\n" if $debug;
    
    #
    # Sigh, a long ping results in the script waiting until all the
    # packets are sent from all the pings, before it will exit. So,
    # loop doing a bunch of shorter pings.
    #
253 254 255
    for ($i = 0; $i < 30; $i++) {
	if (! DoesPing($pc)) {
	    print STDERR "$pc is rebooting.\n" if $debug;
256 257 258 259 260 261 262
	    return 0;
	}
    }
    print STDERR "$pc is still alive.\n" if $debug;
    return 1;
}

263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
#
# Returns 1 if host is responding to pings, 0 otherwise
#
sub DoesPing {
    local($pc) = @_;
    local($status);
    local($saveuid);

    $saveuid = $UID;
    $UID = 0;
    system("$ping -q -i 0.25 -c 8 -t 2 $pc >/dev/null 2>&1");
    $UID = $saveuid;
    $status = $? >> 8;

    #
    # Returns 0 if any packets are returned. Returns 2 if pingable
    # but no packets are returned. Other non-zero error codes indicate
    # other problems.  Any non-zero return indicates "not pingable" to us.
    # 
    print STDERR "$ping $pc returned $status\n" if $debug;
    if ($status) {
	return 0;
    }
    return 1;
}