newnode.in 14 KB
Newer Older
1 2 3 4
#!/usr/bin/perl -w

#
# EMULAB-COPYRIGHT
5
# Copyright (c) 2003-2007 University of Utah and the Flux Group.
6 7 8 9 10 11 12
# All rights reserved.
#

#
# newnode - a script for moving nodes from the new_ tables into production.
#

13
use lib '@prefix@/lib';
14 15

use libdb;
16
use libxmlrpc;
17
use NodeType;
Robert Ricci's avatar
Robert Ricci committed
18
use English;
19
use OSinfo;
20
use Getopt::Std;
21 22 23

use strict;

24 25 26 27 28
my $TB         = "@prefix@";
my $ELABINELAB = @ELABINELAB@;
my $RPCSERVER  = "@OUTERBOSS_NODENAME@";
my $RPCPORT    = "@OUTERBOSS_XMLRPCPORT@";
my $RPCCERT    = "@OUTERBOSS_SSLCERTNAME@";
29 30

my $switchmac = "$TB/sbin/switchmac";
31 32 33 34
my $os_load = "$TB/bin/os_load";
my $os_select = "$TB/bin/os_select";
my $newnode_reboot = "$TB/sbin/newnode_reboot";
my $named_setup = "$TB/sbin/named_setup";
35 36 37
my $nalloc = "$TB/bin/nalloc";
my $nfree = "$TB/bin/nfree";
my $dhcpd_makeconf = "$TB/sbin/dhcpd_makeconf";
38
my $exports_setup = "$TB/sbin/exports_setup";
39 40 41

my $dhcpd_rc = "/usr/local/etc/rc.d/2.dhcpd.sh";

42
my $sudo = "/usr/local/bin/sudo -S";
43 44

#
45
# MFS to boot the nodes into initially
46
#
47 48
my $MFS_INITIAL = TB_OSID_FREEBSD_MFS();
my $OS_OPSNODE  = "OPSNODE-BSD";
49 50

#
51
# Initial event system state to put the nodes into
52
#
53 54
my $STATE_INITIAL = TBDB_NODESTATE_SHUTDOWN;
my $STATE_OPSNODE = TBDB_NODESTATE_ISUP;
55 56 57 58 59 60

#
# Number of vnodes to create for each physical node
#
my $NUM_VNODES = 50;

61 62 63 64 65
#
# Default number of vnodes that can be hosted on any pnode
#
my $DEFAULT_PCVM_COUNT = 20;

66 67 68 69 70 71 72 73
#
# Experiments we might put nodes into
#
my $PID_HWDOWN = NODEDEAD_PID();
my $EID_HWDOWN = NODEDEAD_EID();
my $PID_OPSNODE = TBOPSPID();
my $EID_OPSNODE = "opsnodes";

74 75
# Turn off line buffering on output
$| = 1;
76

77 78 79 80
if (!TBAdmin()) {
    die "Sorry, only testbed administrators can run this script!\n";
}

81 82 83
#
# Handle command-line options
#
84 85 86
sub usage() {
    die "Usage: $0 [-f] <node_id> ...\n";
}
87 88 89 90 91 92 93 94 95 96
my $force_unconnected_interfaces = 0;
my %options = ();
if (! getopts("f", \%options)) {
    usage();
}

if ($options{f}) {
    $force_unconnected_interfaces = 1;
}

97 98 99 100 101 102
#
# We seem to have to do this so that ssh gets proper root permissions to read
# the key file. Argh.
#
$UID = $EUID = 0;

103
if (@ARGV < 1) {
104
    usage();
105 106
}

107 108 109
#
# The user has to be able to run sudo, so they can restart dhcpd.
#
110
if (system "$sudo /bin/pwd < /dev/null") {
111 112 113
    die "You must be able to sudo to root to use this script\n";
}

114 115 116 117 118 119 120 121 122 123
#
# For vnodes - figure out the jail IP base
#
my $IPBASE;
if (TBDB_JAILIPBASE =~ /^(\d+).(\d+).(\d+).(\d+)/) {
    $IPBASE = "$1.$2";
} else {
    die "Problem with JAILIPBASE\n"
}

124 125 126 127 128
my @node_ids = @ARGV;

#
# Now, loop through the nodes given, and add each one
#
Robert Ricci's avatar
Robert Ricci committed
129
my (@succeeded_nodes, @reboot_IPs);
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
NODE: foreach my $node_id (@node_ids) {
    my $query_result;

    #
    # Check to make sure said node does not already exist!
    #
    $query_result = DBQueryFatal("SELECT node_id FROM nodes WHERE " .
	"node_id='$node_id'");
    if ($query_result->num_rows()) {
	warn "Node $node_id failed: a node with that name already exists!\n";
	next NODE;
    }

    #
    # Grab information about the node from the new_nodes table
    #
146
    $query_result = DBQueryFatal("SELECT new_node_id, type, IP, temporary_IP, ".
147
	"floor, building, loc_x, loc_y, contact, phone, room, role " .
148
	"FROM new_nodes WHERE node_id='$node_id'");
149 150 151 152
    if (!$query_result->num_rows()) {
	warn "Node $node_id failed: No pending node with that name exists!\n";
	next NODE;
    }
153
    my ($new_node_id, $type, $IP, $tempIP, $floor, $building, $loc_x, $loc_y,
154
	$contact, $phone, $room, $role) = $query_result->fetchrow();
155

156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
    #
    # Make sure they've given it an IP address that looks valid
    #
    if ($IP !~ /^(\d+).(\d+).(\d+).(\d+)/) {
	warn "Node $node_id has an invalid IP address ($IP) - skipping\n";
	next NODE;
    }

    #
    # Make sure some other node doesn't already have this IP address on its
    # control interface
    #
    $query_result = DBQueryFatal("SELECT node_id FROM interfaces WHERE " .
	"IP='$IP' AND role='" . TBDB_IFACEROLE_CONTROL() . "'");
    if ($query_result->num_rows()) {
	my ($existing_node_id) = $query_result->fetchrow();
	warn "Node $node_id failed: $existing_node_id is already using IP " .
		"address $IP\n";
	next NODE;
    }

177 178 179 180
    #
    # Make sure that the new node is of a valid type, and grab a few other
    # things to fill in as initial values
    #
181 182 183 184 185
    my $typeinfo = NodeType->Lookup($type);
    if (!defined($typeinfo)) {
	warn "No such nodetype $type is defined in the DB!\n";
	next NODE;
    }
186
    my ($control_iface,$virtnode_capacity,$adminmfs,$adminmfs_osid);
187 188 189 190 191 192 193 194 195

    if ($typeinfo->control_interface(\$control_iface) || !$control_iface) {
	warn "No control interface for $type is defined in the DB!\n";
	next NODE;
    }
    if ($typeinfo->virtnode_capacity(\$virtnode_capacity)) {
	warn "Failure looking up virtnode_capacity for $type in the DB!\n";
	next NODE;
    }
196
    if ($typeinfo->adminmfs_osid(\$adminmfs_osid)) {
197
	warn "Failure looking up adminmfs_osid for $type in the DB!\n";
198 199
	next NODE;
    }
200 201
    $virtnode_capacity = 0
	if (!defined($virtnode_capacity));
202 203 204 205 206 207 208 209

    # Find object for the adminfs.
    if (defined($adminmfs_osid)) {
	$adminmfs = OSinfo->Lookup($adminmfs_osid);
    }
    else {
	$adminmfs = OSinfo->Lookup(TBOPSPID(), $MFS_INITIAL);
    }
210 211 212 213

    #
    # Grab the node's MACs from the new_interfaces table
    #
214
    $query_result = DBQueryFatal("SELECT card, MAC, interface_type, " .
215
	"switch_id, switch_card, switch_port, cable, len, role " .
216
	"FROM new_interfaces WHERE new_node_id='$new_node_id'");
217 218 219 220 221 222
    if (!$query_result->num_rows()) {
	warn "Node $node_id failed: Must have at least one interface!\n";
	next NODE;
    }

    my %interfaces;
223
    while (my ($card, $MAC, $iface_type, $switch_id, $switch_card,
224
	    $switch_port, $cable, $len, $ifrole) = $query_result->fetchrow()) {
225 226 227
	#
	# Get some more information about this interface type
	#
228
	my $iface_query = DBQueryFatal("SELECT max_speed, full_duplex, " .
229
	    "connector FROM interface_types WHERE type='$iface_type'");
230
	if (!$iface_query->num_rows()) {
231
	    warn "Node $node_id failed: Interface $card is of unknown type " .
232
	    	"$iface_type\n";
233 234 235
	    next NODE;
	}

236
	my ($max_speed, $full_duplex, $connector) = $iface_query->fetchrow();
237

238 239 240 241
	#
	# Do a sanity check - make sure that we have a switch recorded for all
	# experimental interfaces
	#
242
	unless ($force_unconnected_interfaces || ($role eq "ctrlnode")) {
243
	    my $iface = "eth$card";
244 245
	    if (($iface ne $control_iface) && ($connector !~ /wireless/i)
		&& !$switch_id) {
246 247 248 249 250 251
		warn "Node $node_id failed: Don't know which switch " .
		    "card $iface is connected to\n";
		next NODE;
	    }
	}

252 253 254
	#
	# Stash it away...
	#
255
	$interfaces{$card} = [$MAC, $iface_type, $max_speed, $full_duplex,
256
		$switch_id, $switch_card, $switch_port, $cable, $len, $ifrole];
257 258 259 260 261 262

    }

    #
    # Make up a priority (just used for sorting)
    #
263 264
    $node_id =~ /^(.*\D)(\d+)$/;
    my ($prefix,$nodenum) = ($1, $2);
265
    my $priority;
266 267
    if (defined $nodenum) {
	$priority = $nodenum;
268 269 270 271
    } else {
	$priority = 1;
    }

272 273 274 275 276 277 278 279 280 281 282 283 284
    #
    # Figure out which project and experiment this node needs to go into
    #
    my ($pid, $eid);
    if ($role eq "ctrlnode") {
	$pid = $PID_OPSNODE;
	$eid = $EID_OPSNODE;
    } else {
	$pid = $PID_HWDOWN;
	$eid = $EID_HWDOWN;
    }

    #
285
    # Figure out what osid/opmode this node should run
286
    #
287 288
    my ($osid,$opmode);
    
289
    if ($role eq "ctrlnode") {
290
	my $osinfo = OSinfo->Lookup(TBOPSPID(), $OS_OPSNODE);
291

292
	$osid   = $osinfo->osid();
Leigh B. Stoller's avatar
Leigh B. Stoller committed
293
	$opmode = $osinfo->op_mode();
294 295 296
    }
    else {
	$osid   = $adminmfs->osid();
Leigh B. Stoller's avatar
Leigh B. Stoller committed
297
	$opmode = $adminmfs->op_mode();
298 299 300 301 302 303 304 305 306 307 308 309
    }

    #
    # Figure out what state to start the node in
    #
    my $state;
    if ($role eq "ctrlnode") {
	$state = $STATE_OPSNODE;
    } else {
	$state = $STATE_INITIAL;
    }

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
    #
    # When operating inside an inner elab, we need an extra bit of information
    # from the outer boss; which the IP of the outer control network. This is
    # strictly for bootinfo which has some really gross hacks in it.
    #
    my $outer_control_ip;
    
    if ($ELABINELAB) {
        libxmlrpc::Config({"server"  => $RPCSERVER,
			   "verbose" => 0,
			   "cert"    => $RPCCERT,
			   "portnum" => $RPCPORT});

	my $rval = libxmlrpc::CallMethod("elabinelab", "newnode_info",
					 {"mac" => $interfaces{0}[0]});

	if (!defined($rval)) {
	    print "XMPRPC server did not return control iface info!\n";
	    next NODE;
	}
	else {
	    #
	    # Grab what we need by searching the interface marked as
	    # the outer control network.
	    #
	    my %iface_array = %{ $rval->{"interfaces"} };
		
	    foreach my $mac (keys(%iface_array)) {
338 339
		my $aref   = $iface_array{$mac};
		my $ifrole = $aref->{"role"};
340

341
		if ($ifrole eq TBDB_IFACEROLE_OUTER_CONTROL()) {
342 343 344 345 346 347 348 349 350 351 352
		    $outer_control_ip = $aref->{"IP"};
		    last;
		}
	    }
	}
	if (!defined($outer_control_ip)) {
	    print "Could not determine outer control interface IP!\n";
	    next NODE;
	}
    }

353 354 355 356 357
    #
    # Okay, time to actually add the node!
    #
    
    DBQueryFatal("INSERT INTO nodes SET node_id='$node_id', type='$type', " .
358 359 360
	"phys_nodeid='$node_id', role='$role', priority=$priority, " .
    	"eventstate='$state', op_mode='$opmode', " .
	"def_boot_osid='$MFS_INITIAL', " .
361 362
	"state_timestamp=unix_timestamp(NOW()), " .
	"op_mode_timestamp=unix_timestamp(NOW())");
363

364 365
    DBQueryFatal("insert into node_hostkeys (node_id) ".
		 "values ('$node_id')");
366 367 368 369 370
    DBQueryFatal("insert into node_status ".
		 "(node_id, status, status_timestamp) ".
		 "values ('$node_id', 'down', now()) ");
    DBQueryFatal("insert into node_activity ".
		 "(node_id) values ('$node_id')");
371

372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
    #
    # Copy data into the location_info table, if they provided any
    #
    if ($floor || $building || $loc_x || $loc_y || $contact
	|| $phone || $room) {

	#
	# Handle columns that could have come to us NULL (and thus are undef)
	#
	if (defined $floor) {
	    $floor = "'$floor'";
	} else {
	    $floor = "NULL";
	}
	if (defined $building) {
	    $building = "'$building'";
	} else {
	    $building = "NULL";
	}
	if (defined $room) {
	    $room = "'$room'";
	} else {
	    $room = "NULL";
	}

	DBQueryFatal("INSERT INTO location_info SET node_id='$node_id', " .
	    "floor=$floor, building=$building, loc_x=$loc_x, loc_y=$loc_y, ".
	    "contact='$contact', phone='$phone', room=$room;");
    }

402 403 404
    #
    # Add some vnodes
    #
405
    if (defined $nodenum && ($role eq "testnode")) {
406 407 408 409
	my $vtype = $type;
	if (!($vtype =~ s/pc/pcvm/)) {
	    $vtype = "$vtype-vm";
	}
410
	DBQueryFatal("INSERT INTO node_auxtypes set node_id='$node_id', " .
411
	    "type='pcvm', count=$virtnode_capacity");
412
	DBQueryFatal("INSERT INTO node_auxtypes set node_id='$node_id', " .
413
	    "type='$vtype', count=$virtnode_capacity");
414 415
    }

416
    while (my ($card, $aref) = each %interfaces) {
417
	my ($MAC, $iface_type, $max_speed, $duplex, $switch_id, $switch_card,
418
	    $switch_port, $cable, $len, $ifrole) = @$aref;
419
	my $iface = "eth$card";
420 421 422 423 424
	my $iface_IP;
	my $wire_type;
	my $iface_role;

	# Role from new_interfaces overrides anything we decide here.
425 426
	if (defined($ifrole)) {
	    $iface_role = $ifrole;
427 428
	}
	elsif ($iface eq $control_iface) {
429
	    $iface_role = TBDB_IFACEROLE_CONTROL();
430
	}
431 432 433 434 435 436 437 438 439 440 441 442 443
	else {
	    $iface_role = TBDB_IFACEROLE_EXPERIMENT();
	}

	# Okay, now set up some stuff for the inserts.
	if ($iface_role eq TBDB_IFACEROLE_CONTROL()) {
	    $iface_IP  = $IP;
	    $wire_type = "Control";
	}
	elsif ($iface_role eq TBDB_IFACEROLE_OUTER_CONTROL()) {
	    $iface_IP  = $outer_control_ip;
	    $wire_type = "OuterControl";
	}
444 445
	else {
	    # Experimental interfaces start with no speed set. 
Kirk Webb's avatar
Kirk Webb committed
446
	    $max_speed = 0;
447 448
	    $iface_IP  = "";
	    $wire_type = "Node";
449
	}
450

451 452 453
	DBQueryFatal("INSERT INTO interfaces SET node_id='$node_id', " .
	    "card=$card, port=1, mac='$MAC', IP='$iface_IP', " .
	    "interface_type='$iface_type', iface='$iface', " .
Kirk Webb's avatar
Kirk Webb committed
454
	    "current_speed='$max_speed', duplex=$duplex, role='$iface_role'");
455

456 457 458 459
	if (!$switch_id) {
	    print "No switch found for ${node_id}:$iface - skipping\n";
	    next;
	}
460 461 462 463 464 465 466 467

	my $cable_len = "";
	if ($cable) {
	    $cable_len .= ", cable=$cable";
	}
	if ($len) {
	    $cable_len .= ", len=$len";
	}
468 469 470
	DBQueryFatal("INSERT INTO wires SET type='$wire_type', " .
	    "node_id1='$node_id', card1=$card, port1=1, " .
	    "node_id2='$switch_id', card2='$switch_card', " .
471
	    "port2='$switch_port' $cable_len");
472 473 474
    }

    #
475 476
    # Put it into hwdown for now - I would put them in reloading, but I'm
    # afriad the reload_daemon might do the wrong thing to them
477
    #
478
    system "$nalloc $pid $eid $node_id";
479 480 481 482

    #
    # Remove the node from the new_ tables
    #
483
    DBQueryFatal("DELETE FROM new_nodes WHERE new_node_id=$new_node_id");
484
    DBQueryFatal("DELETE FROM new_interfaces WHERE new_node_id=$new_node_id");
485

486
    print "$node_id succesfully added!\n";
487 488

    push @succeeded_nodes, $node_id;
Robert Ricci's avatar
Robert Ricci committed
489 490 491 492 493 494 495

    #
    # Only reboot regular nodes, not ops nodes
    #
    if ($role ne "ctrlnode") {
	push @reboot_IPs, $tempIP;
    }
496 497 498 499 500 501 502 503 504 505 506 507
}

#
# No point in restarting dhcpd, etc. if there are no nodes that succeeded
#
if (!@succeeded_nodes) {
    die "No nodes succeeded, exiting early\n";
}

#
# Re-generate dhcpd.conf
#
508
print "Re-generating dhcpd.conf\n";
509
if (system("$dhcpd_makeconf -i -r")) {
510 511
    die "Unable to open regenerate dhcpd config file\n";
}
512 513 514 515 516 517 518 519 520 521 522 523

print "Restarting dhcpd: $sudo $dhcpd_rc stop\n";
my $sudo_rv = system "$sudo $dhcpd_rc stop";
if ($sudo_rv) {
    die "Error stopping dhcpd - return value was $sudo_rv\n";
}
sleep 2;
print "Restarting dhcpd: $sudo $dhcpd_rc start\n";
$sudo_rv = system "$sudo $dhcpd_rc start";
if ($sudo_rv) {
    die "Error starting dhcpd - return value was $sudo_rv\n";
}
524

525 526 527 528 529
print "Setting up nameserver\n";
my $named_rv = system "$named_setup";
if ($named_rv) {
    die "Error running named_setup - return value was $named_rv\n";
}
530

531 532 533 534 535 536 537 538 539
#
# Before we boot nodes into the MFS, we have to make sure they can mount
# NFS filesystems
#
print "Running exports_setup\n";
my $exports_rv = system "$exports_setup";
if ($exports_rv) {
    warn "WARNING - exports_setup returned $exports_rv";
}
540

541 542 543 544
#
# Start rebooting nodes
#
print "Rebooting nodes...\n";
Robert Ricci's avatar
Robert Ricci committed
545
foreach my $IP (@reboot_IPs) {
546 547
    print "Rebooting $IP\n";
    system "$newnode_reboot $IP\n";
548 549
}

550 551 552 553 554
print "\n\n";
print "Finished - when you are satisifed that the nodes are working\n";
print "correctly, use nfree on boss to free them from the emulab-ops/hwdown\n";
print "experiment.\n";

555
#
556 557 558
# TODO -
#    disable interfaces
#    console setup
559
#