elabinelab.in 49.6 KB
Newer Older
1
#!/usr/bin/perl -w
2 3
#
# EMULAB-COPYRIGHT
4
# Copyright (c) 2004-2012 University of Utah and the Flux Group.
5 6 7 8 9 10 11 12
# All rights reserved.
#
# TODO: ntpinfo table.
#       Current source directory? From where?
#
use English;
use Getopt::Std;

13 14 15 16
# Load the Testbed support stuff.
use lib "@prefix@/lib";
use libdb;
use libtestbed;
17
use libtblog;
18
use Experiment;
19
use User;
20
use Lan;
21

22 23 24 25 26
#
# Do things necessary for setting up inner elab experiment. 
#
sub usage()
{
27
    print STDOUT "Usage: elabinelab [-d] [-g] [-u] pid eid\n";
28
    print STDOUT "       elabinelab [-d] [-k | -f] pid eid\n";
29
    print STDOUT "       elabinelab [-d] -r pid eid [node ...]\n";
30 31 32
 
    exit(-1);
}
33
my $optlist  = "dgkfurP";
34
my $debug    = 1;
35
my $verbose  = 0;
36
my $killmode = 0;
37
my $fwboot   = 0;
38
my $dbgooonly= 0;
39 40
my $update   = 0;
my $remove   = 0;
41

42 43 44 45 46 47 48 49
#
# XXX experimental speed hacks.
#     $inparallel    reboots all server in parallel (rather than serially)
#		     after setup
#     $restartnodes  uses a new bootinfo RESTART command to quickly move
#		     inner nodes from control of outer boss to inner boss
#		     avoiding all node reboots
#
50
my $inparallel = 1;
51
my $restartnodes = 0;
52

53 54
sub DumpDBGoo();

55 56 57 58 59 60
#
# Configure variables
#
my $TB		= "@prefix@";
my $TBOPS       = "@TBOPSEMAIL@";
my $CONTROL	= "@USERNODE@";
Russ Fish's avatar
Russ Fish committed
61
my $DBNAME      = "@TBDBNAME@";
62 63
my $TBOPSPID    = TBOPSPID();
my $SSH		= "$TB/bin/sshtb";
64
my $SCP		= "/usr/bin/scp";
65
my $nodereboot  = "$TB/bin/node_reboot";
66
my $noderestart	= "$TB/sbin/bootinfosend -R";
67
my $makeconf    = "$TB/sbin/dhcpd_makeconf";
68
my $nodewait    = "$TB/sbin/node_statewait";
69
my $snmpit      = "$TB/bin/snmpit";
70
my $osselect	= "$TB/bin/os_select";
71 72 73

# Protos
sub TearDownEmulab();
74 75
sub RemoveNodes();
sub UpdateEmulab();
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96

# un-taint path
$ENV{'PATH'} = '/bin:/usr/bin:/usr/local/bin:/usr/site/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("*** $0:\n".
	"    Must be root! Maybe its a development version?\n");
}

# Locals
my $SAVEUID     = $UID;
my $workdir;
97
my $expdir;
98 99 100
my %noderoles	= ();
my $opsnode;
my $bossnode;
Mike Hibler's avatar
Mike Hibler committed
101 102
my $fsnode;
my $routernode;
103
my @expnodes    = ();
104
my $query_result;
105 106
my $inner_experiment;
my $inner_nsfile;
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121

#
# Parse command arguments. Once we return from getopts, all that should
# left are the required arguments.
#
%options = ();
if (! getopts($optlist, \%options)) {
    usage();
}
if (defined($options{"g"})) {
    $dbgooonly = 1;
}
if (defined($options{"d"})) {
    $debug = 1;
}
122 123 124
if (defined($options{"k"})) {
    $killmode = 1;
}
125 126 127
if (defined($options{"f"})) {
    $fwboot = 1;
}
128 129 130 131 132 133
if (defined($options{"u"})) {
    $update = 1;
}
if (defined($options{"r"})) {
    $remove = 1;
}
134 135 136
if (defined($options{"P"})) {
    $inparallel = 1;
}
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
if (! @ARGV) {
    usage();
}
my ($pid,$eid) = @ARGV;

#
# Untaint the arguments.
#
if ($pid =~ /^([-\w]+)$/) {
    $pid = $1;
}
else {
    die("Tainted argument $pid!\n");
}
if ($eid =~ /^([-\w]+)$/) {
    $eid = $1;
}
else {
    die("Tainted argument $eid!\n");
}
157 158 159 160 161
my $experiment = Experiment->Lookup($pid, $eid);
if (!defined($experiment)) {
    die("*** $0:\n".
	"    Could not map $pid/$eid to its object!\n");
}
162
$workdir = TBExptWorkDir($pid, $eid);
163
$expdir = PROJROOT() . "/$pid/exp/$eid";
164

165 166 167 168 169
# Build Logfile names
my $opslogfile  = "$workdir/opsnode.log";
my $fslogfile   = "$workdir/fsnode.log";
my $bosslogfile = "$workdir/bossnode.log";

170 171 172
#
# Verify user and get his DB uid.
#
173 174 175
my $this_user = User->ThisUser();
if (! defined($this_user)) {
    tbdie("You ($UID) do not exist!");
176
}
177 178 179
my $user_uid   = $this_user->uid();
my $user_name  = $this_user->name();
my $user_email = $this_user->email();
180

Leigh Stoller's avatar
Leigh Stoller committed
181
TBDebugTimeStampsOn();
182

183 184 185 186
#
# Get elabinelab status to make sure, and to see if we need to fire off
# an experiment inside once its setup.
#
187 188 189 190 191
my $elabinelab           = $experiment->elabinelab();
my $elabinelab_eid       = $experiment->elabinelab_eid();
my $elabinelab_nosetup   = $experiment->elabinelab_nosetup();
my $elabinelab_singlenet = $experiment->elabinelab_singlenet();

192 193 194
exit(0)
    if (!$elabinelab);

195 196 197 198 199 200
#
# See if the experiment is firewalled
#
my $firewall;
my $firewalled = TBExptFirewall($pid, $eid, \$firewall);

201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
#
# Presetup; turn off firewall.
#
if ($fwboot) {
    exit(0)
	if (!$firewalled);
    
    print "Turning off firewall rules on $firewall\n";
    $UID = 0;
    system("$SSH -host $firewall ipfw add 1 allow all from any to any");
    if ($?) {
	die("*** $0:\n".
	    "    Error turning off firewall rules ($firewall)!\n");
    }
    exit(0);
}

218 219 220 221 222
#
# If we are going to start an inner experiment, grab the stuff we need
# from the DB and save it. 
#
if (defined($elabinelab_eid)) {
223
    $inner_experiment = Experiment->Lookup($pid, $elabinelab_eid);
224 225
    die("*** $0:\n".
	"    No such experiment in DB for $pid/$elabinelab_eid\n")
226
	if (!defined($inner_experiment));
227

228 229 230
    $inner_experiment->GetNSFile(\$inner_nsfile) == 0 or
	die("*** $0:\n".
	    "    Could not get NS file for $inner_experiment\n");
231 232
    
    die("*** $0:\n".
233 234
	"    No nsfile in DB for $inner_experiment")
	if (!defined($inner_nsfile) || $inner_nsfile eq "");
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249
}

#
# Get the role for each node.
#
$query_result =
    DBQueryFatal("select r.node_id,r.inner_elab_role from reserved as r ".
		 "where r.pid='$pid' and r.eid='$eid'");
while (my ($node_id,$role) = $query_result->fetchrow_array()) {
    # Like, the firewall node.
    next
	if (!defined($role));
	
    $noderoles{$node_id} = $role;
    $bossnode = $node_id
250
	if ($role =~ /^boss/);
Mike Hibler's avatar
Mike Hibler committed
251 252
    $routernode = $node_id
	if ($role eq 'router');
253
    $opsnode = $node_id
Mike Hibler's avatar
Mike Hibler committed
254 255 256
	if ($role eq 'ops' || $role eq 'ops+fs');
    $fsnode = $node_id
	if ($role eq 'fs');
257 258 259 260 261 262 263 264 265 266
    push(@expnodes, $node_id)
	if ($role eq 'node');
}

#
# Tear down an inner emulab.
# 
if ($killmode) {
    exit(TearDownEmulab());
}
267 268 269 270 271 272
elsif ($remove) {
    exit(RemoveNodes());
}
elsif ($update) {
    exit(UpdateEmulab());
}
273 274 275 276 277 278

#
# Get elabinelab info. If this is a container for an actual experiment,
# then need to fire off the experiment once the inner emulab is ready to
# go.
# 
Leigh Stoller's avatar
Leigh Stoller committed
279
TBDebugTimeStamp("Dumping DB state");
280 281 282 283 284
DumpDBGoo();
exit(0)
    if ($dbgooonly);

#
285
# For SSH and SCP below
286 287
#
$UID = 0;
Leigh Stoller's avatar
Leigh Stoller committed
288

289
#
290 291
# The firewall should be off at this point; called from os_setup with -f.
# 
292

Leigh Stoller's avatar
Leigh Stoller committed
293 294 295
#
# This is temporary. I think I will switch this over to grabbing the latest
# version from the web server.
296
#
297
# XXX ugh, copy over a newer mkextrafs.pl as well (one that supports -2).
298
# XXX ughII, we only copy over a FreeBSD version, this will break a Linux boss.
299
#
300
my $mkelab = "$TB/etc/rc.mkelab";
301 302
if (-e "$expdir/rc.mkelab") {
    $mkelab = "$expdir/rc.mkelab";
303
}
304 305 306 307
my $mkextrafs = "";
if (-e "$TB/etc/mkextrafs.pl") {
    $mkextrafs = "$TB/etc/mkextrafs.pl";
}
308 309 310
print "Copying $mkelab $mkextrafs to ${bossnode}";
print "/${opsnode}"
    if (defined($opsnode));
311 312 313
print "/${fsnode}"
    if (defined($fsnode));
print "\n";
314
system("scp $mkelab $mkextrafs ${bossnode}:/tmp");
315 316
system("scp $mkelab $mkextrafs ${opsnode}:/tmp")
    if (defined($opsnode));
317
system("scp $mkelab $mkextrafs ${fsnode}:/tmp")
318
    if (defined($fsnode));
319

320 321 322
if (defined($fsnode)) {
    TBDebugTimeStamp("Setting up fsnode");
    print "Setting up fsnode on $fsnode\n";
323
    system("$SSH -host $fsnode /tmp/rc.mkelab -s -d > $fslogfile 2>&1");
324 325 326 327 328 329 330
    if ($?) {
	$UID = $SAVEUID;
	SENDMAIL("$user_name <$user_email>",
		 "ElabInElab Failure: $pid/$eid",
		 "Error building the fs node ($fsnode)",
		 $TBOPS,
		 "Cc: $TBOPS",
331
		 ($fslogfile));
332 333 334 335 336
	print STDERR "*** $0:\n".
	    "    Error building the fsnode ($fsnode)!\n";
	exit(($debug ? 0 : -1));
    }
}
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
if (defined($opsnode)) {
    TBDebugTimeStamp("Setting up opsnode");
    print "Setting up opsnode on $opsnode\n";
    system("$SSH -host $opsnode /tmp/rc.mkelab -s -d > $opslogfile 2>&1");
    if ($?) {
	$UID = $SAVEUID;
	SENDMAIL("$user_name <$user_email>",
		 "ElabInElab Failure: $pid/$eid",
		 "Error building the ops node ($opsnode)",
		 $TBOPS,
		 "Cc: $TBOPS",
		 ($opslogfile));
	print STDERR "*** $0:\n".
	             "    Error building the opsnode ($opsnode)!\n";
	exit(($debug ? 0 : -1));
    }
353
}
Leigh Stoller's avatar
Leigh Stoller committed
354
TBDebugTimeStamp("Setting up bossnode");
355
print "Setting up bossnode on $bossnode\n";
356
system("$SSH -host $bossnode /tmp/rc.mkelab -s -d > $bosslogfile 2>&1");
357 358 359 360
if ($?) {
    $UID = $SAVEUID;
    SENDMAIL("$user_name <$user_email>",
	     "ElabInElab Failure: $pid/$eid",
361
	     "Error building the boss node ($bossnode)",
362 363
	     $TBOPS,
	     "Cc: $TBOPS",
364
	     ($bosslogfile));
365 366 367 368 369
    print STDERR "*** $0:\n".
	         "    Error building the bossnode ($bossnode)!\n";
    exit(($debug ? 0 : -1));
}

370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
if ($verbose) {
    # Send these log files off now so that we can look at them.
    if (defined($fsnode)) {
	SENDMAIL("$user_name <$user_email>",
		 "ElabInElab Setup Log: $pid/$eid",
		 "Logs for building fs/ops/boss ($fsnode/$opsnode/$bossnode)",
		 $TBOPS,
		 "Cc: $TBOPS",
		 ($fslogfile, $opslogfile, $bosslogfile));
    }
    else {
	SENDMAIL("$user_name <$user_email>",
		 "ElabInElab Setup Log: $pid/$eid",
		 "Logs for building ops/boss ($opsnode/$bossnode)",
		 $TBOPS,
		 "Cc: $TBOPS",
		 ($opslogfile, $bosslogfile));
    }
388
}
389
$UID  = $SAVEUID;
390 391

# Run as real user for the next few scripts, which are setuid.
392
$EUID = $UID;
393

394
goto skipsetup
395
    if ($elabinelab_nosetup);
396

397
#
398 399 400 401 402
# Restart DHCPD, but first mark the nodes as being ready to boot inside
# the inner emulab, so that dhcpd_makeconf knows what nodes to change
# the entries for.
#
DBQueryFatal("update reserved set inner_elab_boot=1 ".
403
	     "where pid='$pid' and eid='$eid'");
404

405 406 407 408 409 410 411
print "Regenerating DHCPD config file and restarting daemon.\n";
system("$makeconf -i -r");
if ($?) {
    die("*** $0:\n".
	"    Failed to reconfig/restart DHCPD.\n");
}

412
if ($inparallel) {
413 414 415
    my $nodes = "$bossnode";
    $nodes .= " $opsnode"
	if (defined($opsnode));
416 417 418 419 420
    $nodes .= " $fsnode"
	if (defined($fsnode));
    print "Rebooting servers ($nodes).\n";
    TBDebugTimeStamp("Rebooting servers");
    system("$nodereboot -w $nodes");
421
    if ($?) {
422 423 424
	print STDERR "*** $0:\n".
	    "    Error rebooting the servers ($nodes)!\n";
	exit(($debug ? 0 : -1));
425 426 427 428 429 430 431 432
    }
} else {
    if (defined($fsnode)) {
	# Reboot fs and wait for it to come back.
	print "Rebooting fsnode ($fsnode).\n";
	TBDebugTimeStamp("Rebooting fsnode");
	system("$nodereboot -w $fsnode");
	if ($?) {
433 434 435
	    print STDERR "*** $0:\n".
		"    Error rebooting the fsnode ($fsnode)!\n";
	    exit(($debug ? 0 : -1));
436 437
	}
    }
438 439 440 441 442 443
    if (defined($opsnode)) {
	# Reboot ops and wait for it to come back.
	print "Rebooting opsnode ($opsnode).\n";
	TBDebugTimeStamp("Rebooting opsnode");
	system("$nodereboot -w $opsnode");
	if ($?) {
444 445 446
	    print STDERR "*** $0:\n".
		"    Error rebooting the opsnode ($opsnode)!\n";
	    exit(($debug ? 0 : -1));
447
	}
448 449 450 451 452 453
    }
    # Reboot boss and wait for it to come back.
    print "Rebooting bossnode ($bossnode).\n";
    TBDebugTimeStamp("Rebooting bossnode");
    system("$nodereboot -w $bossnode");
    if ($?) {
454 455 456
	print STDERR "*** $0:\n".
	    "    Error rebooting the bossnode ($bossnode)!\n";
	exit(($debug ? 0 : -1));
457
    }
458
}
459
$EUID = 0;
Leigh Stoller's avatar
Leigh Stoller committed
460

461 462 463
# Reboot the experimental nodes. They will come up inside the inner elab.
# DO NOT WAIT! They are not going to report ISUP from this point on. 
if (@expnodes) {
464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
    #
    # First we try the magic pxeboot restart.
    # The nodes should still be in PXEWAIT, so we send them a restart
    # to make them re-DHCP.  This should get them quickly reparented to
    # the inner boss.
    #
    # If this doesn't work, we fall back on rebooting the nodes.
    #
    if ($restartnodes) {
	TBDebugTimeStamp("Redirecting experimental nodes to inner boss");
	my $stat = 0;
	# Run as real user again.
	$EUID = $UID;
	foreach my $node (@expnodes) {
	    $stat = system("$noderestart $node");
	    last if ($stat);
	}
	$EUID = 0;
	if ($stat) {
	    tbwarn("Node restart failed ($stat), falling back to reboot.");
	    goto rebootnodes;
	}

	#
	# Ssh into inner boss and use a utility script to determine
	# when the nodes have reported in and are in PXEWAIT (part of the
	# inner elab). Note the short timeout, since this operation should
	# be virtually instantaneous.
	#
	print "Waiting for nodes to restart and join the inner emulab.\n";
	TBDebugTimeStamp("Waiting for inner nodes to restart");
	$UID  = 0;
	$stat = system("$SSH -host $bossnode ".
		       "/usr/testbed/sbin/node_statewait -t 15 -a");
	$UID  = $SAVEUID;
	if ($stat) {
	    tbwarn("Error ($stat) waiting for nodes to restart, falling back to reboot.");
	    goto rebootnodes;
	}

	goto restartworked;
    }

rebootnodes:
508
    print "Rebooting inner experimental nodes.\n";
Leigh Stoller's avatar
Leigh Stoller committed
509
    TBDebugTimeStamp("Rebooting experimental nodes");
510 511
    # Run as real user again.
    $EUID = $UID;
512
    system("$nodereboot -b @expnodes");
513 514 515 516
    if ($?) {
	die("*** $0:\n".
	    "    Error rebooting the expnodes (@expnodes)!\n");
    }
517
    $EUID = 0;
518 519 520 521 522

    #
    # Instead, we ssh into the node and use a utility script to determine
    # when the nodes have rebooted and are in PXEWAIT (part of the inner elab).
    #
523
    # Run as real root for ssh.
524 525 526 527 528 529 530 531 532 533 534
    $UID  = 0;

    print "Waiting for nodes to reboot and join the inner emulab.\n";
    TBDebugTimeStamp("Waiting for inner nodes to reboot");
    system("$SSH -host $bossnode /usr/testbed/sbin/node_statewait -t 180 -a");
    if ($?) {
	print STDERR "*** $0:\n".
	             "    Error waiting for inner nodes to join!\n";
	exit(($debug ? 0 : -1));
    }
    $UID  = $SAVEUID;
535

536
restartworked:
537 538 539 540 541 542 543 544
    #
    # To avoid confusion later (with swapmod, which wants them to be ISUP),
    # and so the web interface does not show the nodes as down, set the 
    # state to ISUP.
    #
    foreach my $node (@expnodes) {
	TBSetNodeEventState($node, TBDB_NODESTATE_ISUP());
    }
545
}
546 547

#
548 549 550 551 552
# Fire off inner elab experiment.
# 
if (defined($elabinelab_eid)) {
    # Formatted to make batchexp happy.
    my $nsfilename = "/tmp/$pid-$elabinelab_eid-$$.nsfile";
553
    
554 555 556 557 558 559
    #
    # Write NS file to temp file so we can send it over.
    #
    open(NS, "> /tmp/$$.ns")
	or die("*** $0:\n".
	       "    Could not write ns code to tmp file!\n");
560
    print NS $inner_nsfile;
561 562 563 564 565 566 567
    print NS "\n";
    close(NS);

    #
    # Copy the file over.
    #
    $UID = 0;
568
    print "Sending NS file to inner bossnode ($bossnode).\n";
569
    system("cat /tmp/$$.ns | $SSH -host $bossnode '(cat > $nsfilename)'");
570 571
    if ($?) {
	die("*** $0:\n".
572
	    "    Could not copy ns code to inner boss ($bossnode)!\n");
573
    }
574 575

    #
576 577
    # Now run batchexp on the node as the user. If firewalled, experiment
    # must start async (cause we have to turn the firewall back on). 
578
    #
579 580 581 582
    my $optarg = ($firewalled ? "" : "-w");
	
    print "Starting experiment $pid/$elabinelab_eid on inner emulab.\n";
    TBDebugTimeStamp("Starting inner experiment");
583 584
    system("$SSH -host $bossnode ".
	   " 'sudo -u $user_uid /usr/testbed/bin/batchexp ".
585
	   "  -q -i $optarg -S \"ElabInElab Experiment\" ".
586 587 588 589 590
	   "  -L \"ElabInElab ElabInElab\" -E \"ElabInElab Experiment\" ".
	   "  -p $pid -e $elabinelab_eid $nsfilename'");
    
    $UID = $SAVEUID;
    unlink("/tmp/$$.ns");
591
}
592
skipsetup:
593

594
#
595 596 597 598 599 600
# Turn the firewall back on.
#
# XXX If this fails, we have to do something much stronger! We do not want
# nodes coming up and starting something if the firewall is not active.
# Maybe hit the panic button from here (turning off the control network).
#
601 602 603 604 605 606
#
if ($firewalled) {
    print "Turning firewall back on\n";
    $UID = 0;
    system("$SSH -host $firewall ipfw delete 1");
    if ($?) {
607 608 609 610 611 612 613
	print STDERR "*** Error turning back on firewall rules ($firewall)!\n".
		     "    Will retry again.\n";
	system("$SSH -host $firewall ipfw delete 1");
	if ($?) {
	    die("*** $0:\n".
		"    Error turning back on firewall rules! Retry failed.\n");
	}
614 615 616 617
    }
    $UID = $SAVEUID;
}

618 619 620
TBDebugTimeStamp("ElabInElab setup done");
exit(0);

621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
#
# Dump parts of the DB that are needed for inner elab to run. The idea
# is to create a set of files named by the table name. Note that mysqld
# cannot write to the project tree cause of directory permissions. Put the
# files into the workdir for now, and them copy them over. 
#
sub DumpDBGoo()
{
    my $statedir = "$workdir/elabinelab";

    if (-d $statedir) {
	system("rm -rf $statedir");
    }
    mkdir($statedir, 0777) or
	die("*** $0:\n".
	    "    Could not mkdir $statedir\n");
    
    chmod(0777, $statedir) or
	die("*** $0:\n".
	    "    Could not chmod $statedir\n");

642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
    #
    # No place else for this; when ops is a vm on boss, need to generate
    # a proper IP that is routable on the control network.
    #
    my $attributes = $experiment->GetElabInElabAttrs();
    die("*** $0:\n".
	"    Could not get elabinelab attributes\n")
	if (!defined($attributes));

    if (exists($attributes->{'CONFIG_OPSVM'}) && $attributes->{'CONFIG_OPSVM'}){
	#
	# Need to assign an IP to the ops node for the jail.
	#
	my $pnode = Node->Lookup($bossnode);
	die("*** $0:\n".
	    "    Could not lookup $bossnode\n")
	    if (!defined($pnode));
	my $ip = $pnode->GetJailIP();
	die("*** $0:\n".
	    "    Could not generate an IP for OPS jail\n")
	    if (!defined($ip));
	
	print "Setting the IP for OPS jail to $ip\n";
	$experiment->SetElabInElabAttr("boss", "OPSIP", $ip);
	$experiment->SetElabInElabAttr("ops",  "OPSIP", $ip);
	$experiment->SetElabInElabAttr("fs",   "OPSIP", $ip);
    }

670 671 672
    #
    # These tables are dumped completely.
    #
673
    my @FULLTABLES = ("node_types", "node_type_attributes", "interface_types",
674 675
		      "interface_capabilities",
		      "switch_paths", "switch_stack_types", "switch_stacks",
676 677
		      "node_type_features", "node_types_auxtypes", "osid_map",
		      "os_boot_cmd");
678 679 680

    #
    # These tables are dumped by role (node/ops). For each one dump the table
681 682
    # as is, unless its the fs or ops node. For those we want to change the
    # node_id to "fs" or "ops" and their type to ops.
683
    #
684 685 686
    my @NODETABLES = ("node_auxtypes", "node_status", "nodes", 
		      "node_rusage", "node_hostkeys", "node_activity",
		      "interface_state");
687 688 689 690

    #
    # These tables are dumped by project ID.
    #
691
    my @PROJTABLES = ("projects", "groups");
692 693

    #
694
    # These tables are dumped by user ID (for the project members).
695
    #
696
    my @USERTABLES = ("users", "user_pubkeys");
697 698 699

    foreach my $table (@FULLTABLES) {
	unlink("$statedir/$table");
700 701 702 703 704
	DBQueryWarn("create temporary table temp_${table} ".
		    "select t.* from $table as t")
	    or die("*** $0:\n".
		   "    Could not dump table $table\n");

705 706
	if ($table eq "node_type_attributes") {
	    my $attributes_result =
707
		DBQueryFatal("select type,attrvalue from temp_${table} ".
708 709 710 711
			     "where attrkey='delay_capacity'");

	    while (my ($ntype,$value) = $attributes_result->fetchrow_array()) {
		my $newvalue = $value - 1;
712 713 714

		next
		    if ($newvalue < 0);
715 716 717
		
		DBQueryFatal("update temp_${table} set ".
			     "   attrvalue='$newvalue' ".
718 719
			     "where type='$ntype' and ".
			     "      attrkey='delay_capacity'");
720
	    }
721 722
	}

723 724 725 726 727
	# filter out community strings
	if ($table eq "switch_stacks" || $table eq "switch_stack_types") {
	    DBQueryFatal("update temp_${table} set snmp_community=NULL");
	}

728
	DBQueryWarn("select * from temp_$table ".
729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
		    "into outfile '$statedir/$table'")
	    or die("*** $0:\n".
		   "    Could not dump table $table\n");
    }

    foreach my $table (@NODETABLES) {
	unlink("$statedir/$table");
	#
	# Create a temporary table.
	#
	DBQueryWarn("create temporary table temp_${table} ".
		    "select t.* from reserved as r ".
		    "left join $table as t on t.node_id=r.node_id ".
		    "left join virt_nodes as v on v.vname=r.vname and ".
		    "     v.pid=r.pid and v.eid=r.eid ".
		    "where r.pid='$pid' and r.eid='$eid' and ".
		    "      t.node_id is not null and ".
746
		    "      v.inner_elab_role in ('node','fs','ops','ops+fs')")
747 748 749
	    or die("*** $0:\n".
		   "    Could not create temporary table temp_$table\n");
	#
750 751
	# Rename the fs and ops node in each table. For the nodes table,
	# there is a bunch of other stuff to do.
752
	#
753 754 755
	DBQueryFatal("update temp_${table} set node_id='fs' ".
		     "where node_id='$fsnode'")
	    if (defined($fsnode));
756
	DBQueryFatal("update temp_${table} set node_id='ops' ".
757 758
		     "where node_id='$opsnode'")
	    if (defined($opsnode));
759 760 761 762

	if ($table eq "nodes") {
	    DBQueryFatal("update temp_${table} set ".
			 " type='ops', ".
763
			 " phys_nodeid=node_id, ".
764 765
			 " role='ctrlnode', ".
			 " op_mode='OPSNODEBSD' ".
766
			 "where node_id in ('fs','ops')");
767 768 769 770 771 772 773

	    # Also add the nodes that correspond to the "trunk" wires.
	    DBQueryFatal("insert into temp_${table} ".
			 "select distinct n.* from wires as w ".
			 "left join nodes as n on w.node_id1=n.node_id or ".
			 "     w.node_id2=n.node_id ".
			 "where w.type='Trunk'");
774 775 776 777

	    # Clear any node reservations on the inside
	    DBQueryFatal("update temp_${table} set ".
			 " reserved_pid=null where reserved_pid is not null");
778 779 780 781 782 783 784 785 786 787

	    # Put the inner nodes into "limbo" so they DTRT when restarted
	    if ($restartnodes) {
		DBQueryFatal("update temp_${table} set".
			     "  op_mode='PXEKERNEL',next_op_mode='',".
			     "  eventstate='". TBDB_NODESTATE_PXELIMBO . "',".
			     "  temp_boot_osid=NULL,next_boot_osid=NULL,".
			     "  osid=NULL".
			     " where role='testnode'");
	    }
788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
	}
    
	DBQueryWarn("select * from temp_$table ".
		    "into outfile '$statedir/$table'")
	    or die("*** $0:\n".
		   "    Could not dump table $table\n");
    }

    foreach my $table (@PROJTABLES) {
	unlink("$statedir/$table");
	DBQueryWarn("select * from $table ".
		    "where pid='$pid' ".
		    "into outfile '$statedir/$table'")
	    or die("*** $0:\n".
		   "    Could not dump table $table\n");
    }
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825
    #
    # Cleared versions of the project/group stats tables.
    #
    unlink("$statedir/project_stats");
    DBQueryFatal("create temporary table temp_project_stats ".
		 "like project_stats");
    DBQueryFatal("insert into temp_project_stats (pid,pid_idx) ".
		 "select pid,pid_idx from project_stats ".
		 "where pid='$pid'");
    DBQueryFatal("select * from temp_project_stats ".
		 "into outfile '$statedir/project_stats'");

    unlink("$statedir/group_stats");
    DBQueryFatal("create temporary table temp_group_stats ".
		 "like group_stats");
    DBQueryFatal("insert into temp_group_stats ".
		 "  (pid,pid_idx,gid,gid_idx,gid_uuid) ".
		 "select pid,pid_idx,gid,gid_idx,gid_uuid ".
		 "   from group_stats ".
		 "where pid='$pid'");
    DBQueryFatal("select * from temp_group_stats ".
		 "into outfile '$statedir/group_stats'");
826

827 828 829 830 831 832 833 834 835 836 837 838
    #
    # Special case the group and user policy tables. Not sure what to
    # really do about this; should there be any restrictions inside the
    # inner elab?
    #
    unlink("$statedir/group_policies");
    DBQueryWarn("select * from group_policies ".
		"where pid='$pid' or pid='+' or pid='-' ".
		"into outfile '$statedir/group_policies'")
	or die("*** $0:\n".
	       "    Could not dump table group_policies\n");

839 840
    foreach my $table (@USERTABLES) {
	unlink("$statedir/$table");
841 842

	DBQueryWarn("create temporary table temp_$table ".
843
		    "select distinct t.* from group_membership as gm ".
844 845
		    "left join users as u on u.uid_idx=gm.uid_idx ".
		    "left join $table as t on t.uid_idx=u.uid_idx ".
846 847
		    "where (gm.pid='$pid' or ".
		    "       gm.pid='" . TBOPSPID() . "') and gm.gid=gm.pid ".
848
		    " and t.uid_idx is not NULL and ".
849 850 851 852
		    " u.status='" . USERSTATUS_ACTIVE() . "'")
	    or die("*** $0:\n".
		   "    Could not create table temp_$table\n");

853 854 855 856 857
	# Clean up ... these are created in the inner elab.
	DBQueryFatal("delete from temp_${table} ".
		     "where uid='elabman' or uid='elabckup' or ".
		     "      uid='operator'");

858
	if ($table eq "users") {
859
	    my $creator_uid = $experiment->creator();
860 861 862
	    
	    DBQueryFatal("update temp_${table} set ".
			 " admin=1 ".
863
			 "where uid='$creator_uid'");
864 865

	    #
866
	    # Save time; force all other users to start out
867 868 869 870 871
	    # frozen since most users in the project do not ever
	    # actually log in. 
	    #
	    DBQueryFatal("update temp_${table} set ".
			 " status='" . USERSTATUS_FROZEN() . "' ".
872
			 "where uid!='$creator_uid'");
873 874 875
	}

	DBQueryWarn("select * from temp_$table ".
876 877
		    "into outfile '$statedir/$table'")
	    or die("*** $0:\n".
878
		   "    Could not dump table temp_$table\n");
879
    }
880 881 882 883 884 885 886 887 888 889 890 891
    #
    # We want a cleared stats table, so do it here.
    #
    DBQueryFatal("create temporary table temp_user_stats ".
		 "like user_stats");
    DBQueryFatal("insert into temp_user_stats ".
		 "  (uid,uid_idx,uid_uuid) ".
		 "select uid,uid_idx,uid_uuid from temp_users");
    DBQueryWarn("select * from temp_user_stats ".
		"into outfile '$statedir/user_stats'")
	or die("*** $0:\n".
	       "    Could not dump table temp_user_stats\n");
892 893 894

    # The group_membership is also special.
    DBQueryWarn("select gm.* from group_membership as gm ".
895
		"left join users as u on u.uid_idx=gm.uid_idx ".
896
		"where (gm.pid='$pid' or ".
897
		"       (gm.pid='" . TBOPSPID() . "' and gm.pid=gm.gid)) and ".
898
		" u.status='" . USERSTATUS_ACTIVE() . "' and ".
899 900
		" gm.uid!='elabman' and gm.uid!='elabckup' and ".
		" gm.uid!='operator' ".
901 902 903 904 905
		"into outfile '$statedir/group_membership'")
	or die("*** $0:\n".
	       "    Could not dump table group_membership\n");

    #
Mike Hibler's avatar
Mike Hibler committed
906
    # Initial images; note that these images are not going to exist inside!
907 908
    # Note that we exclude any encrypted images because we cannot expose
    # the encryption keys that they would need to decode the images!
909 910
    # 
    DBQueryWarn("select * from images ".
911 912
		"where (pid='$pid' or (pid='$TBOPSPID' and global=1)) ".
		"  and decryption_key is null ".
913 914 915 916 917 918 919 920 921 922 923 924 925 926
		"into outfile '$statedir/images'")
	or die("*** $0:\n".
	       "    Could not dump table images\n");
	    
    DBQueryWarn("create temporary table temp_os_info ".
		"select * from os_info ".
		"where pid='$pid' or (pid='$TBOPSPID' and shared=1)")
	or die("*** $0:\n".
	       "    Could not create table temp_os_info\n");

    # Ack. The MFS paths have a hardcoded "boss" in them, but that is going
    # to resolve incorrectly to an inner control IP, which will not work
    # from the pxeboot kernel since it uses the outer control network.
    # Just remove the host spec; pxeboot will do the right thing.
927
    my $query_result =
928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947
	DBQueryFatal("select osid,path from temp_os_info ".
		     "where path like '%:%'");
    
    while (my ($osid,$hostpath) = $query_result->fetchrow_array()) {
	my ($host,$path) = $hostpath =~ /^(.*):(.*)$/;

	DBQueryFatal("update temp_os_info set path='$path' where osid='$osid'");
    }

    DBQueryWarn("select * from temp_os_info ".
		"into outfile '$statedir/os_info'")
	or die("*** $0:\n".
	       "    Could not dump table os_info\n");
	    
    DBQueryWarn("select o.* from osidtoimageid as o ".
		"left join images as i on i.imageid=o.imageid ".
		"where i.pid='$pid' or (i.pid='$TBOPSPID' and i.global=1) ".
		"into outfile '$statedir/osidtoimageid'")
	or die("*** $0:\n".
	       "    Could not dump table osidtoimageid\n");
948 949 950 951 952 953 954 955 956
	    
    # Subosids.  Only take the mapping to parents for those children that
    # are in the e-in-e pid or are in emulab-ops and are shared.
    DBQueryWarn("select osm.* from os_submap as osm ".
		"left join os_info as osi on osm.osid=osi.osid ".
		"where osi.pid='$pid' or (osi.pid='$TBOPSPID' and osi.shared=1) ".
		"into outfile '$statedir/os_submap'")
	or die("*** $0:\n".
	       "    Could not dump table os_submap\n");
957 958 959 960 961 962 963 964 965 966 967 968

    #
    # interfaces table. Need to tag the interfaces being used as the control
    # network, with the proper tag so they do not say they experimental
    # interfaces in the inner emulab. Use a temp table again.
    #
    DBQueryWarn("create temporary table temp_interfaces ".
		"select t.* from reserved as r ".
		"left join interfaces as t on t.node_id=r.node_id ".
		"left join virt_nodes as v on v.vname=r.vname and ".
		"     v.pid=r.pid and v.eid=r.eid ".
		"where r.pid='$pid' and r.eid='$eid' and ".
969
		"      v.inner_elab_role in ('node','ops','fs','ops+fs')")
970 971 972
	or die("*** $0:\n".
	       "    Could not create temporary table temp_interfaces\n");

973
    if (! $elabinelab_singlenet) {
974 975 976 977 978 979 980
	# First, mark the real control network as "other" to avoid it being
	# thought of as the control network!.
	DBQueryWarn("update temp_interfaces ".
		    "set role='" . TBDB_IFACEROLE_OUTER_CONTROL() . "' " .
		    "where role='" . TBDB_IFACEROLE_CONTROL() . "'")
	    or die("*** $0:\n".
		   "    Could not delete control ifaces from temp_interfaces\n");
981

982 983 984 985 986 987
	DBQueryWarn("update temp_interfaces set ".
		    " role='" . TBDB_IFACEROLE_CONTROL() . "' " .
		    "where IP!='' and role='" . TBDB_IFACEROLE_EXPERIMENT() . "'")
	    or die("*** $0:\n".
		   "    Could not update roles in temp_interfaces\n");
    }
988

989 990 991 992 993 994 995
    # And rename the fs/ops nodes as above.
    if (defined($fsnode)) {
	DBQueryWarn("update temp_interfaces set node_id='fs' ".
		    "where node_id='$fsnode'")
	    or die("*** $0:\n".
		   "    Could not fs node_id in temp_interfaces\n");
    }
996 997 998 999 1000 1001
    if (defined($opsnode)) {
	DBQueryWarn("update temp_interfaces set node_id='ops' ".
		    "where node_id='$opsnode'")
	    or die("*** $0:\n".
		   "    Could not ops node_id in temp_interfaces\n");
    }
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019

    # Also add the interfaces that correspond to the "trunk" wires.
    DBQueryFatal("insert into temp_interfaces ".
		 "select distinct i.* from wires as w ".
		 "left join interfaces as i on w.node_id1=i.node_id or ".
		 "     w.node_id2=i.node_id ".
		 "where w.type='Trunk'");

    DBQueryWarn("select * from temp_interfaces ".
		"into outfile '$statedir/interfaces'")
	or die("*** $0:\n".
	       "    Could not dump table interfaces\n");

    # And the wires table. Strip out the control wires; not needed.
    DBQueryWarn("create temporary table temp_wires ".
		"select t.* from reserved as r ".
		"left join virt_nodes as v on v.vname=r.vname and ".
		"     v.pid=r.pid and v.eid=r.eid ".
1020
		"left join wires as t on t.node_id1=r.node_id ".
1021
		($elabinelab_singlenet == 0 ? " and t.type='Node' " : " ") .
1022
		"where r.pid='$pid' and r.eid='$eid' and ".
1023
		"      v.inner_elab_role in ('node','ops','fs','ops+fs') ")
1024 1025 1026
	or die("*** $0:\n".
	       "    Could not create temporary table temp_wires\n");

1027 1028 1029 1030 1031 1032 1033
    # And rename the fs/ops node as above.
    if (defined($fsnode)) {
	DBQueryWarn("update temp_wires set node_id1='fs' ".
		    "where node_id1='$fsnode'")
	    or die("*** $0:\n".
		   "    Could not fs node_id in temp_wires\n");
    }
1034 1035 1036 1037 1038 1039
    if (defined($opsnode)) {
	DBQueryWarn("update temp_wires set node_id1='ops' ".
		    "where node_id1='$opsnode'")
	    or die("*** $0:\n".
		   "    Could not ops node_id in temp_wires\n");
    }
1040

1041
    if (! $elabinelab_singlenet) {
1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
	# But we need to take out the wires that are being used as the
	# inner control network, or at least mark them as Control.
	$query_result =
	    DBQueryWarn("select node_id,card,port from temp_interfaces ".
			"where role='" . TBDB_IFACEROLE_CONTROL() . "' ");

	while (my ($node_id,$card,$port) = $query_result->fetchrow_array()) {
	    DBQueryWarn("update temp_wires set type='Control' ".
			"where node_id1='$node_id' and card1=$card and ".
			"      port1=$port");
	}
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
    }
    # Okay, now add the "trunk" wires in without any alteration.
    DBQueryWarn("insert into temp_wires ".
		"select * from wires where type='Trunk'") 
	or die("*** $0:\n".
	       "    Could not add trunk lines to temp_wires\n");

    DBQueryWarn("select * from temp_wires ".
		"into outfile '$statedir/wires'")
	or die("*** $0:\n".
	       "    Could not dump table wires\n");

    #
1066 1067
    # Ack, we need to create a reservation for the fs and ops nodes,
    # or else they will look free and it will not be able to check in.
1068 1069 1070 1071 1072 1073
    #
    DBQueryWarn("create temporary table temp_reserved ".
		"select r.* from reserved as r ".
		"left join virt_nodes as v on v.vname=r.vname and ".
		"     v.pid=r.pid and v.eid=r.eid ".
		"where r.pid='$pid' and r.eid='$eid' ".
1074
		"      and v.inner_elab_role in ('fs','ops','ops+fs')")
1075 1076
	or die("*** $0:\n".
	       "    Could not create temporary table temp_reserved\n");
1077 1078 1079 1080
    if (defined($fsnode)) {
	DBQueryWarn("update temp_reserved set ".
		    "   node_id='fs', ".
		    "   pid='$TBOPSPID', ".
1081 1082
		    "   eid='opsnodes', ".
		    "   exptidx=1 ".
1083 1084 1085 1086
		    "where node_id='$fsnode'")
	    or die("*** $0:\n".
		   "    Could not update temporary table temp_reserved\n");
    }
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
    if (defined($opsnode)) {
	DBQueryWarn("update temp_reserved set ".
		    "   node_id='ops', ".
		    "   pid='$TBOPSPID', ".
		    "   eid='opsnodes', ".
		    "   exptidx=1 ".
		    "where node_id='$opsnode'")
	    or die("*** $0:\n".
		   "    Could not update temporary table temp_reserved\n");
    }
1097 1098 1099 1100 1101
    DBQueryWarn("select * from temp_reserved ".
		"into outfile '$statedir/reserved'")
	or die("*** $0:\n".
	       "    Could not dump table reserved\n");

1102
    # Copy tiplines table for all nodes so web form gives us a console icon!
1103
    DBQueryWarn("select t.tipname,t.node_id,'',t.disabled,0,0,NULL ".
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113
		"from reserved as r ".
		"left join virt_nodes as v on v.vname=r.vname and ".
		"     v.pid=r.pid and v.eid=r.eid ".
		"left join tiplines as t on t.node_id=r.node_id ".
		"where r.pid='$pid' and r.eid='$eid' and ".
		"      v.inner_elab_role='node' ".
		"into outfile '$statedir/tiplines'")
	or die("*** $0:\n".
	       "    Could not dump table tiplines\n");

1114 1115 1116 1117 1118
    #
    # Dump the DB schema too, so we can check in the inner Elab that this data
    # is compatible with the sql/database-create.sql schema file there, *before*
    # loading it into the db.  Added/removed columns would misalign row data.
    #
Russ Fish's avatar
Russ Fish committed
1119 1120 1121 1122 1123 1124 1125 1126 1127
    my $schemafile = "$expdir/outer_db_schema";
    system("rm -f $schemafile")
	if (-f $schemafile);
    my $isvers5     = system("mysql -V | egrep -q -s 'Distrib 5.'") == 0;
    my $extraopts   = ($isvers5 ? "--skip-quote-names" : "");
    #
    # XXX: Requires that mysqldump be in caller's $PATH - probably an OK
    # assumption, but maybe not always
    #
1128 1129
    my $mysqldump   = "mysqldump -d $extraopts $DBNAME " .
	"@FULLTABLES @NODETABLES @PROJTABLES @USERTABLES";
Russ Fish's avatar
Russ Fish committed
1130
    system("$mysqldump 2> /dev/null > $schemafile");
1131

1132
    #
1133
    # Tar up the directory and send it over to (real) ops.
1134 1135 1136
    #
    $UID = 0;
    system("tar cf - -C $statedir . | ".
1137
	   "   gzip | $SSH -F /dev/null -host $CONTROL ".
1138
	   "   '(cat > $expdir/dbstate.tar.gz)'");
1139 1140 1141 1142 1143 1144 1145 1146
    if ($?) {
	die("*** $0:\n".
	    "    Could not create dbstate.tar.gz\n");
    }
    $UID = $SAVEUID;
    return 0;
}

1147
#
1148 1149
# Tear down an inner Emulab as cleanly as possible to avoid power cycling
# nodes.
1150 1151 1152 1153 1154 1155
# 
sub TearDownEmulab()
{
    my $tbdir      = "/usr/testbed";
    my $wap        = "$tbdir/sbin/withadminprivs";
    my $nodereboot = "$tbdir/bin/node_reboot";
1156

1157 1158 1159 1160 1161 1162 1163
    #
    # We want to rebuild the DHCPD file so that when we reboot the inner nodes
    # they come back to the outer emulab. We cannot just free the nodes, cause
    # then the reload daemon might beat us to it, and end up power cycling the
    # nodes, and that would be bad. So, munge the DB and clear the "role" slot
    # for inner nodes. 
    #
1164
    DBQueryFatal("update reserved set inner_elab_role=NULL,inner_elab_boot=0 ".
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
		 "where pid='$pid' and eid='$eid'");

    #
    # XXX Failure at this point will leave things in an inconsistent state
    # cause we have just munged the reserved table. Since we were trying
    # to swap out the experiment, I think this will be okay. Wait and see.
    #
    return 0
	if (!defined($bossnode));

    #
    # Now regen the DHCPD file.
    #
1178
    # Run as real user since script is setuid.
1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
    $EUID = $UID;
    
    print "Regenerating DHCPD config file and restarting daemon.\n";
    system("$makeconf -i -r");
    if ($?) {
	die("*** $0:\n".
	    "    Failed to reconfig/restart DHCPD.\n");
    }
    $EUID = 0;

1189 1190 1191 1192 1193 1194 1195 1196 1197
    #
    # Kill inner vlans table entries; this is the table that maps
    # inner to outer vlans. We do not care about that anymore since
    # all of the vlans are going to be torn down (using the outer
    # ids).
    #
    DBQueryFatal("delete from elabinelab_vlans ".
		 "where pid='$pid' and eid='$eid'");

1198
    #
1199 1200
    # If firewalled, just return now since all nodes will be powered
    # off anyway.
1201
    #
1202 1203
    if ($firewalled) {
	print "Skipping clean shutdown since experiment is firewalled.\n";
1204 1205 1206
	return 0;
    }

1207 1208 1209 1210 1211
    #
    # When the nodes reboot, we want them to do something reasonable. We
    # have no idea what is loaded on the disk, so they should go into an
    # MFS and wait, but then a bunch of nodes will all try to load the big
    # MFS at once, and that could wreak havoc. So, clear the boot osids
1212 1213
    # so they go into PXEWAIT.
    #
1214 1215 1216 1217 1218 1219 1220
    if (@expnodes) {
	system("$osselect -w @expnodes");
	if ($?) {
	    print STDERR "*** $0:\n".
		         "    Could not clear bootinfo for inner nodes!\n".
			 "    Continuing anyway.\n";
	}
1221
    }
Leigh Stoller's avatar
Leigh Stoller committed
1222
    
1223 1224 1225 1226 1227
    #
    # SSH in and kill the inner DHCPD daemon so that it does not reply
    # to rebooting nodes along the inner control network.
    #
    $UID = 0;
1228

1229 1230 1231
    print "Killing DHCPD on inner boss ($bossnode)\n";
    system("$SSH -host $bossnode /usr/local/etc/rc.d/2.dhcpd.sh stop");
    if ($?) {
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
	#
	# This error is non-fatal. If DHCPD cannot be killed, then the inner
	# boss is scrogged or never set up properly. Just return and let
	# the nodes get power cycled (if need be). At some point we need a
	# state machine to control this setup stuff. 
	# 
	print STDERR "*** $0:\n".
	             "    Could not stop DHCPD on inner bossnode ($bossnode)!\n".
		     "    Continuing anyway; outer boss will use power cycle.\n";
	return 0;
1242
    }
1243

1244 1245 1246
    #
    # Now we ask inner boss to reboot all of the testnodes. Maybe need an
    # option to node_reboot, but for now just pass them on the command line.
1247 1248 1249 1250 1251 1252
    #
    if (! @expnodes) {
	$UID = $SAVEUID;
	return 0;
    }
    
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
    print "Asking inner boss ($bossnode) to reboot inner nodes\n";
    system("$SSH -host $bossnode $wap $nodereboot -b @expnodes");
    if ($?) {
	#
	# This error is non-fatal; Outer boss will just resort to power cycle.
	#
	print STDERR "*** $0:\n".
	             "    Could not reboot some inner nodes!\n".
		     "    Continuing anyway; outer boss will use power cycle.\n";
    }
1263
    $UID = $SAVEUID;
1264

1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
    #
    # Now we wait for them to reach PXEWAIT. Again, use our utility script
    # instead of stated stuff.
    #
    $EUID = $UID;
    print "Waiting for inner nodes to reach PXEWAIT\n";
    system("$nodewait @expnodes");
    if ($?) {
	#
	# This error is non-fatal; Outer boss will just resort to power cycle.
	#
	print STDERR "*** $0:\n".
	             "    Some machines did not reboot properly!\n".
		     "    Continuing anyway; outer boss will use power cycle.\n";
    }
    return 0;
}
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291

#
# Remove nodes from an inner Emulab.
# 
sub RemoveNodes()
{
    my $tbdir      = "/usr/testbed";
    my $wap        = "$tbdir/sbin/withadminprivs";
    my $nodereboot = "$tbdir/bin/node_reboot";
    my $deletenode = "$tbdir/sbin/deletenode";
1292
    my $creator    = $experiment->creator();
1293
    my @nodes	   = ();
1294
    my $paniced    = 0;
1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307

    #
    # If firewalled, check to see if paniced. Right now that means the nodes
    # are going to be powered off, so need to do the clean shutdown dance.
    # 
    if ($firewalled) {
	TBExptGetPanicBit($pid, $eid, \$paniced);
    }

    #
    # Actually, this should not even happen; a paniced experiment cannot be
    # modified at all.
    #
1308 1309
    if ($paniced) {
	print "A paniced experiment cannot be modified! What happened?\n";
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
	return -1;
    }

    #
    # Grab the list of nodes. We want to clear the reserved table bits so
    # that we can regen the DHCPD file. 
    #
    shift(@ARGV);	# pid
    shift(@ARGV);	# eid

    foreach my $node (@ARGV) {
	# Untaint the nodes.
	if ($node =~ /^([-\w]+)$/) {
	    $node = $1;
	}
	else {
	    die("*** Tainted node name: $node\n");
	}
	push(@nodes, $node);
    }
    return 0
	if (!@nodes);

    #
    # Grab the vlans table. We need to find any ports used by the nodes
    # getting deleted, and move them back to the default vlan. 
    #
1337 1338
    my @delmembers = ();
    my @todelete   = ();
1339 1340
    
    my $query_result =
1341 1342
	DBQueryWarn("select inner_id,outer_id from elabinelab_vlans ".
		    "where pid='$pid' and eid='$eid'");
1343 1344 1345
    return -1
	if (!$query_result);

1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
    while (my ($inner_id,$outer_id) = $query_result->fetchrow_array()) {
	my $vlan = VLan->Lookup($outer_id);
	if (!defined($vlan)) {
	    print STDERR "*** No such vlan $outer_id ($inner_id)\n";
	    return -1;
	}
	my @members;

	if ($vlan->MemberList(\@members) != 0) {
	    print STDERR "*** Unable to load members for $vlan\n";
	    return -1;
	}
	my $id         = $outer_id;
	my $changed    = 0;
1360

1361 1362 1363
	foreach my $member (@members) {
	    my $node;
	    my $iface;
1364

1365 1366 1367
	    if ($member->GetNodeIface(\$node, \$iface) != 0) {
		print STDERR "Missing attributes for $member in $vlan\n";
		return -1;
1368
	    }
1369 1370 1371 1372 1373 1374
	    my $nodeid = $node->node_id();
	    
	    # See if this node is in the list of nodes to be deleted,
	    if (grep {$_ eq $nodeid} @nodes) {
		push(@todelete, "$nodeid:$iface");
		push(@delmembers, $member);
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
		$changed = 1;
	    }
	}
    }

    # Remove ports from the vlans.
    if (@todelete) {
	print "Removing ports from deleted nodes: @todelete\n";
	system("$snmpit -m default @todelete");
	if ($?) {
	    return -1;
	}
    }
    # Only if the above succeeds, do we update the vlans table.
1389 1390
    foreach my $member (@delmembers) {
	$member->Delete() == 0
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
	    or return -1;
    }
    
    #
    # We want to rebuild the DHCPD file so that when we reboot the inner nodes
    # they come back to the outer emulab. We cannot just free the nodes, cause
    # then the reload daemon might beat us to it, and end up power cycling the
    # nodes, and that would be bad. So, munge the DB and clear the "role" and
    # boot slots for nodes about to be released (by tbswap).
    #
    DBQueryWarn("update reserved set inner_elab_role=NULL,inner_elab_boot=0 ".
		"where pid='$pid' and eid='$eid' and (".
		join(" or ", map("node_id='$_'", @nodes)) . ")")
	or return -1;

    #
    # Now regen the DHCPD file.
    #
    # Run as real user since script is setuid.
    $EUID = $UID;
    
    print "Regenerating DHCPD config file and restarting daemon.\n";
    system("$makeconf -i -r");
    if ($?) {
	die("*** $0:\n".
	    "    Failed to reconfig/restart DHCPD.\n");
    }
    $EUID = 0;

    #
    # When the nodes reboot, we want them to do something reasonable. We
    # have no idea what is loaded on the disk, so they should go into an
    # MFS and wait, but then a bunch of nodes will all try to load the big
    # MFS at once, and that could wreak havoc. So, clear the boot osids
1425 1426 1427 1428 1429 1430 1431 1432
    # so they go into PXEWAIT.
    #
    system("$osselect -w @nodes");
    if ($?) {
	print STDERR "*** $0:\n".
	             "    Could not clear bootinfo for freed nodes!\n".
		     "    Continuing anyway.\n";
    }
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461
    
    #
    # SSH in and kill the inner DHCPD daemon so that it does not reply
    # to rebooting nodes along the inner control network.
    #
    $UID = 0;

    #
    # We are going to do this in a loop, one node at a time. I do not like
    # doing it this way, but its the only reasonable thing to do until we
    # can reboot the inner nodes ourselves (via the outer control network).
    # The reason for doing it one node at a time, is that I cannot delete the
    # node from the inner testbed until its been rebooted. Note that the
    # delete node script regens the dhcpd.conf file, so no need to do that
    # explicitly.
    #
    foreach my $node (@nodes) {
	print "Asking inner boss ($bossnode) to reboot $node\n";
	system("$SSH -host $bossnode $wap $nodereboot -b $node");
	if ($?) {
	    #
	    # This error is non-fatal;
	    # Outer boss will just resort to power cycle.
	    #
	    print STDERR "*** $0:\n".
		         "    Could not reboot $node! Continuing anyway.\n".
			 "    Outer boss will use power cycle.\n";
	}
	print "Asking inner boss ($bossnode) to delete $node\n";
1462
	system("$SSH -host $bossnode sudo -u $creator ".
1463
	       "     $wap $deletenode -b -q -f $node");
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
	if ($?) {
	    #
	    # This error is bad. 
	    #
	    print STDERR "*** $0:\n".
		         "    Could not delete $node! Modify will fail!\n";
	    return -1;
	}
    }
    $UID = $SAVEUID;

    #
    # Now we wait for them to reach PXEWAIT. Again, use our utility script
    # instead of stated stuff.
    #
    $EUID = $UID;
    print "Waiting for inner nodes to reach PXEWAIT\n";
    system("$nodewait @nodes");
    if ($?) {
	#
	# This error is non-fatal; Outer boss will just resort to power cycle.
	#
	print STDERR "*** $0:\n".
	             "    Some machines did not reboot properly!\n".
		     "    Continuing anyway; outer boss will use power cycle.\n";
    }
    return 0;
}

#
# Update an Emulab (add nodes).
# 
sub UpdateEmulab()
{
    my $tbdir      = "/usr/testbed";
1499
    my $statedir   = "$workdir/elabinelab";
1500 1501 1502
    my $wap        = "$tbdir/sbin/withadminprivs";
    my $nodereboot = "$tbdir/bin/node_reboot";
    my $nodewait   = "$tbdir/sbin/node_statewait";
1503
    my $creator    = $experiment->creator();
1504
    my @nodes      = ();
1505
    my $paniced    = 0;
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

    #
    # If firewalled, check to see if paniced. Right now that means the nodes
    # are going to be powered off, so need to do the clean shutdown dance.
    # 
    if ($firewalled) {
	TBExptGetPanicBit($pid, $eid, \$paniced);
    }

    #
    # Actually, this should not even happen; a paniced experiment cannot be
    # modified at all.
    #
1519 1520
    if ($paniced) {
	print "A paniced experiment cannot be modified! What happened?\n";
1521 1522 1523 1524 1525 1526 1527
	return -1;
    }

    #
    # Grab the list of nodes that have been added to the inner elab.
    #
    my $query_result =
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
	DBQueryFatal("select r.node_id,n.type,i.IP from reserved as r ".
		     "left join nodes as n on n.node_id=r.node_id ".
		     "left join interfaces as i on i.node_id=r.node_id and ".
		     "     i.role='" . TBDB_IFACEROLE_CONTROL() . "' ".
		     "where r.pid='$pid' and r.eid='$eid' and ".
		     "      r.inner_elab_boot=0 and r.inner_elab_role='node'");
    return 0
	if (!$query_result->numrows);

    DBQueryFatal("create temporary table temp_new_nodes like new_nodes");
    DBQueryFatal("create temporary table temp_new_interfaces ".
		 " like new_interfaces");

    my %idmap = ();
    while (my ($node,$type,$ip) = $query_result->fetchrow_array()) {
	my $result =
	    DBQueryFatal("insert into temp_new_nodes set ".
			 " new_node_id=NULL, node_id='$node', ".
			 " type='$type', IP='$ip'");
	
	$idmap{$node} = $result->insertid;
    }
    @nodes = keys(%idmap);

    $query_result =
	DBQueryFatal("select r.node_id,i.ip,i.role,i.card,i.mac, ".
		     "  i.interface_type,w.node_id2,w.card2,w.port2 ".
		     " from reserved as r ".
		     "left join interfaces as i on i.node_id=r.node_id ".
		     "left join wires as w on w.node_id1=i.node_id and ".
		     "     w.card1=i.card and w.port1=i.port ".
		     "where r.pid='$pid' and r.eid='$eid' and ".
		     "      r.inner_elab_boot=0 and r.inner_elab_role='node'");

    while (my ($node,$ip,$role,$card,$mac,$type,$switch,
	       $switch_card,$switch_port) = $query_result->fetchrow_array()) {
	my $nid = $idmap{$node};

	if (! $elabinelab_singlenet) {
	    #
	    # Mark the real control network as "other" to avoid it being
	    # thought of as the control network.
	    #
	    if ($role eq TBDB_IFACEROLE_CONTROL()) {
		$role = TBDB_IFACEROLE_OUTER_CONTROL();
	    }
	    # And mark the inner control network.
	    if ($role eq TBDB_IFACEROLE_EXPERIMENT() &&
		defined($ip) && $ip ne "") {
		$role = TBDB_IFACEROLE_CONTROL();
Mike Hibler's avatar
Mike Hibler committed
1578 1579 1580
		# update the temp_new_nodes table with this IP
		DBQueryFatal("update temp_new_nodes set IP='$ip' ".
			     " where node_id='$node'");
1581 1582 1583 1584 1585 1586 1587
	    }
	}
	DBQueryFatal("insert into temp_new_interfaces set ".
		     " new_node_id='$nid', role='$role', card='$card', ".
		     " mac='$mac', interface_type='$type', ".
		     " switch_id='$switch',switch_card='$switch_card', ".
		     " switch_port='$switch_port'");
1588
    }
1589 1590 1591 1592 1593 1594 1595
    unlink("$statedir/new_nodes");
    DBQueryFatal("select * from temp_new_nodes ".
		 "into outfile '$statedir/new_nodes'");
    unlink("$statedir/new_interfaces");
    DBQueryFatal("select * from temp_new_interfaces ".
		 "into outfile '$statedir/new_interfaces'");

1596
    return 0
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
	if ($dbgooonly);
    
    # For SSH and SCP below
    $UID = 0;

    #
    # Send the new_nodes and new_interfaces tables over to the
    # inner elab.
    #
    print "Sending new DB tables to inner boss ($bossnode)\n";
    system("$SCP $statedir/new_nodes $statedir/new_interfaces ".
	   "    ${bossnode}:/tmp");
    if ($?) {
	print STDERR "*** $0:\n".
	             "    Could not scp tables to inner boss\n";
	return -1;
    }
    system("$SSH -host $bossnode mysqlimport -r tbdb ".
	   "     /tmp/new_nodes /tmp/new_interfaces");
    if ($?) {
	print STDERR "*** $0:\n".
	             "    Could not load tables on inner boss\n";
	return -1;
    }
    print "Telling inner boss ($bossnode) to incorporate new nodes: @nodes\n";
    system("$SSH -host $bossnode sudo -u $creator /usr/testbed/sbin/wap ".
Mike Hibler's avatar
Mike Hibler committed
1623
	   "     /usr/testbed/sbin/newnode -f -q -n @nodes");
1624 1625 1626 1627 1628
    if ($?) {
	print STDERR "*** $0:\n".
	             "    Could not incorporate new nodes on inner boss\n";
	return -1;
    }
1629 1630

    # Run as real user for the next few scripts, which are setuid.
1631
    $UID  = $SAVEUID;
1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
    $EUID = $UID;

    #
    # Restart DHCPD, but first mark the nodes as being ready to boot inside
    # the inner emulab, so that dhcpd_makeconf knows what nodes to change
    # the entries for.
    #
    DBQueryFatal("update reserved set inner_elab_boot=1 ".
		 "where pid='$pid' and eid='$eid' and ".
		 "      inner_elab_boot=0 and inner_elab_role='node'");

    print "Regenerating DHCPD config file and restarting daemon.\n";
    system("$makeconf -i -r");
    if ($?) {
	die("*** $0:\n".
	    "    Failed to reconfig/restart DHCPD.\n");
    }

    # Reboot the experimental nodes. They will come up inside the inner elab.
    # DO NOT WAIT! They are not going to report ISUP from this point on. 
    print "Rebooting inner new experimental nodes.\n";
    TBDebugTimeStamp("Rebooting experimental nodes");
    system("$nodereboot @nodes");
    if ($?) {
	die("*** $0:\n".
	    "    Error rebooting the nodes (@nodes)!\n");
    }
    $EUID = 0;

    #
    # At this point, not much I can think of do. The nodes will reboot and
    # enter the newnode MFS. I could add a script to wait for that in the
    # inner elab, but not going to bother yet. 
1665 1666 1667 1668 1669 1670 1671 1672
    #
    # To avoid confusion later (with swapmod, which wants them to be ISUP),
    # and so the web interface does not show the nodes as down, set the 
    # state to ISUP.
    #
    foreach my $node (@nodes) {
	TBSetNodeEventState($node, TBDB_NODESTATE_ISUP());
    }
1673 1674
    return 0;
}