PVE/QemuMigrate.pm

   1 package PVE::QemuMigrate;
   2
   3 use strict;
   4 use warnings;
   5
   6 use IO::File;
   7 use IPC::Open2;
   8 use POSIX qw( WNOHANG );
   9 use Time::HiRes qw( usleep );
  10
  11 use PVE::Cluster;
  12 use PVE::GuestHelpers qw(safe_boolean_ne safe_string_ne);
  13 use PVE::INotify;
  14 use PVE::RPCEnvironment;
  15 use PVE::Replication;
  16 use PVE::ReplicationConfig;
  17 use PVE::ReplicationState;
  18 use PVE::Storage;
  19 use PVE::Tools;
  20
  21 use PVE::QemuConfig;
  22 use PVE::QemuServer::CPUConfig;
  23 use PVE::QemuServer::Drive;
  24 use PVE::QemuServer::Helpers qw(min_version);
  25 use PVE::QemuServer::Machine;
  26 use PVE::QemuServer::Monitor qw(mon_cmd);
  27 use PVE::QemuServer;
  28
  29 use PVE::AbstractMigrate;
  30 use base qw(PVE::AbstractMigrate);
  31
  32 sub fork_command_pipe {
  33     my ($self, $cmd) = @_;
  34
  35     my $reader = IO::File->new();
  36     my $writer = IO::File->new();
  37
  38     my $orig_pid = $$;
  39
  40     my $cpid;
  41
  42     eval { $cpid = open2($reader, $writer, @$cmd); };
  43
  44     my $err = $@;
  45
  46     # catch exec errors
  47     if ($orig_pid != $$) {
  48         $self->log('err', "can't fork command pipe\n");
  49         POSIX::_exit(1);
  50         kill('KILL', $$);
  51     }
  52
  53     die $err if $err;
  54
  55     return { writer => $writer, reader => $reader, pid => $cpid };
  56 }
  57
  58 sub finish_command_pipe {
  59     my ($self, $cmdpipe, $timeout) = @_;
  60
  61     my $cpid = $cmdpipe->{pid};
  62     return if !defined($cpid);
  63
  64     my $writer = $cmdpipe->{writer};
  65     my $reader = $cmdpipe->{reader};
  66
  67     $writer->close();
  68     $reader->close();
  69
  70     my $collect_child_process = sub {
  71         my $res = waitpid($cpid, WNOHANG);
  72         if (defined($res) && ($res == $cpid)) {
  73             delete $cmdpipe->{cpid};
  74             return 1;
  75         } else {
  76             return 0;
  77         }
  78      };
  79
  80     if ($timeout) {
  81         for (my $i = 0; $i < $timeout; $i++) {
  82             return if &$collect_child_process();
  83             sleep(1);
  84         }
  85     }
  86
  87     $self->log('info', "ssh tunnel still running - terminating now with SIGTERM\n");
  88     kill(15, $cpid);
  89
  90     # wait again
  91     for (my $i = 0; $i < 10; $i++) {
  92         return if &$collect_child_process();
  93         sleep(1);
  94     }
  95
  96     $self->log('info', "ssh tunnel still running - terminating now with SIGKILL\n");
  97     kill 9, $cpid;
  98     sleep 1;
  99
 100     $self->log('err', "ssh tunnel child process (PID $cpid) couldn't be collected\n")
 101         if !&$collect_child_process();
 102 }
 103
 104 sub read_tunnel {
 105     my ($self, $tunnel, $timeout) = @_;
 106
 107     $timeout = 60 if !defined($timeout);
 108
 109     my $reader = $tunnel->{reader};
 110
 111     my $output;
 112     eval {
 113         PVE::Tools::run_with_timeout($timeout, sub { $output = <$reader>; });
 114     };
 115     die "reading from tunnel failed: $@\n" if $@;
 116
 117     chomp $output;
 118
 119     return $output;
 120 }
 121
 122 sub write_tunnel {
 123     my ($self, $tunnel, $timeout, $command) = @_;
 124
 125     $timeout = 60 if !defined($timeout);
 126
 127     my $writer = $tunnel->{writer};
 128
 129     eval {
 130         PVE::Tools::run_with_timeout($timeout, sub {
 131             print $writer "$command\n";
 132             $writer->flush();
 133         });
 134     };
 135     die "writing to tunnel failed: $@\n" if $@;
 136
 137     if ($tunnel->{version} && $tunnel->{version} >= 1) {
 138         my $res = eval { $self->read_tunnel($tunnel, 10); };
 139         die "no reply to command '$command': $@\n" if $@;
 140
 141         if ($res eq 'OK') {
 142             return;
 143         } else {
 144             die "tunnel replied '$res' to command '$command'\n";
 145         }
 146     }
 147 }
 148
 149 sub fork_tunnel {
 150     my ($self, $ssh_forward_info) = @_;
 151
 152     my @localtunnelinfo = ();
 153     foreach my $addr (@$ssh_forward_info) {
 154         push @localtunnelinfo, '-L', $addr;
 155     }
 156
 157     my $cmd = [@{$self->{rem_ssh}}, '-o ExitOnForwardFailure=yes', @localtunnelinfo, '/usr/sbin/qm', 'mtunnel' ];
 158
 159     my $tunnel = $self->fork_command_pipe($cmd);
 160
 161     eval {
 162         my $helo = $self->read_tunnel($tunnel, 60);
 163         die "no reply\n" if !$helo;
 164         die "no quorum on target node\n" if $helo =~ m/^no quorum$/;
 165         die "got strange reply from mtunnel ('$helo')\n"
 166             if $helo !~ m/^tunnel online$/;
 167     };
 168     my $err = $@;
 169
 170     eval {
 171         my $ver = $self->read_tunnel($tunnel, 10);
 172         if ($ver =~ /^ver (\d+)$/) {
 173             $tunnel->{version} = $1;
 174             $self->log('info', "ssh tunnel $ver\n");
 175         } else {
 176             $err = "received invalid tunnel version string '$ver'\n" if !$err;
 177         }
 178     };
 179
 180     if ($err) {
 181         $self->finish_command_pipe($tunnel);
 182         die "can't open migration tunnel - $err";
 183     }
 184     return $tunnel;
 185 }
 186
 187 sub finish_tunnel {
 188     my ($self, $tunnel) = @_;
 189
 190     eval { $self->write_tunnel($tunnel, 30, 'quit'); };
 191     my $err = $@;
 192
 193     $self->finish_command_pipe($tunnel, 30);
 194
 195     if (my $unix_sockets = $tunnel->{unix_sockets}) {
 196         # ssh does not clean up on local host
 197         my $cmd = ['rm', '-f', @$unix_sockets];
 198         PVE::Tools::run_command($cmd);
 199
 200         # .. and just to be sure check on remote side
 201         unshift @{$cmd}, @{$self->{rem_ssh}};
 202         PVE::Tools::run_command($cmd);
 203     }
 204
 205     die $err if $err;
 206 }
 207
 208 sub start_remote_tunnel {
 209     my ($self, $raddr, $rport, $ruri, $unix_socket_info) = @_;
 210
 211     my $nodename = PVE::INotify::nodename();
 212     my $migration_type = $self->{opts}->{migration_type};
 213
 214     if ($migration_type eq 'secure') {
 215
 216         if ($ruri =~ /^unix:/) {
 217             my $ssh_forward_info = ["$raddr:$raddr"];
 218             $unix_socket_info->{$raddr} = 1;
 219
 220             my $unix_sockets = [ keys %$unix_socket_info ];
 221             for my $sock (@$unix_sockets) {
 222                 push @$ssh_forward_info, "$sock:$sock";
 223                 unlink $sock;
 224             }
 225
 226             $self->{tunnel} = $self->fork_tunnel($ssh_forward_info);
 227
 228             my $unix_socket_try = 0; # wait for the socket to become ready
 229             while ($unix_socket_try <= 100) {
 230                 $unix_socket_try++;
 231                 my $available = 0;
 232                 foreach my $sock (@$unix_sockets) {
 233                     if (-S $sock) {
 234                         $available++;
 235                     }
 236                 }
 237
 238                 if ($available == @$unix_sockets) {
 239                     last;
 240                 }
 241
 242                 usleep(50000);
 243             }
 244             if ($unix_socket_try > 100) {
 245                 $self->{errors} = 1;
 246                 $self->finish_tunnel($self->{tunnel});
 247                 die "Timeout, migration socket $ruri did not get ready";
 248             }
 249             $self->{tunnel}->{unix_sockets} = $unix_sockets if (@$unix_sockets);
 250
 251         } elsif ($ruri =~ /^tcp:/) {
 252             my $ssh_forward_info = [];
 253             if ($raddr eq "localhost") {
 254                 # for backwards compatibility with older qemu-server versions
 255                 my $pfamily = PVE::Tools::get_host_address_family($nodename);
 256                 my $lport = PVE::Tools::next_migrate_port($pfamily);
 257                 push @$ssh_forward_info, "$lport:localhost:$rport";
 258             }
 259
 260             $self->{tunnel} = $self->fork_tunnel($ssh_forward_info);
 261
 262         } else {
 263             die "unsupported protocol in migration URI: $ruri\n";
 264         }
 265     } else {
 266         #fork tunnel for insecure migration, to send faster commands like resume
 267         $self->{tunnel} = $self->fork_tunnel();
 268     }
 269 }
 270
 271 sub lock_vm {
 272     my ($self, $vmid, $code, @param) = @_;
 273
 274     return PVE::QemuConfig->lock_config($vmid, $code, @param);
 275 }
 276
 277 sub prepare {
 278     my ($self, $vmid) = @_;
 279
 280     my $online = $self->{opts}->{online};
 281
 282     $self->{storecfg} = PVE::Storage::config();
 283
 284     # test if VM exists
 285     my $conf = $self->{vmconf} = PVE::QemuConfig->load_config($vmid);
 286
 287     my $repl_conf = PVE::ReplicationConfig->new();
 288     $self->{replication_jobcfg} = $repl_conf->find_local_replication_job($vmid, $self->{node});
 289     $self->{is_replicated} = $repl_conf->check_for_existing_jobs($vmid, 1);
 290
 291     if ($self->{replication_jobcfg} && defined($self->{replication_jobcfg}->{remove_job})) {
 292         die "refusing to migrate replicated VM whose replication job is marked for removal\n";
 293     }
 294
 295     PVE::QemuConfig->check_lock($conf);
 296
 297     my $running = 0;
 298     if (my $pid = PVE::QemuServer::check_running($vmid)) {
 299         die "can't migrate running VM without --online\n" if !$online;
 300         $running = $pid;
 301
 302         if ($self->{is_replicated} && !$self->{replication_jobcfg}) {
 303             if ($self->{opts}->{force}) {
 304                 $self->log('warn', "WARNING: Node '$self->{node}' is not a replication target. Existing " .
 305                                    "replication jobs will fail after migration!\n");
 306             } else {
 307                 die "Cannot live-migrate replicated VM to node '$self->{node}' - not a replication " .
 308                     "target. Use 'force' to override.\n";
 309             }
 310         }
 311
 312         $self->{forcemachine} = PVE::QemuServer::Machine::qemu_machine_pxe($vmid, $conf);
 313
 314         # To support custom CPU types, we keep QEMU's "-cpu" parameter intact.
 315         # Since the parameter itself contains no reference to a custom model,
 316         # this makes migration independent of changes to "cpu-models.conf".
 317         if ($conf->{cpu}) {
 318             my $cpuconf = PVE::JSONSchema::parse_property_string('pve-cpu-conf', $conf->{cpu});
 319             if ($cpuconf && PVE::QemuServer::CPUConfig::is_custom_model($cpuconf->{cputype})) {
 320                 $self->{forcecpu} = PVE::QemuServer::CPUConfig::get_cpu_from_running_vm($pid);
 321             }
 322         }
 323     }
 324
 325     my $loc_res = PVE::QemuServer::check_local_resources($conf, 1);
 326     if (scalar @$loc_res) {
 327         if ($self->{running} || !$self->{opts}->{force}) {
 328             die "can't migrate VM which uses local devices: " . join(", ", @$loc_res) . "\n";
 329         } else {
 330             $self->log('info', "migrating VM which uses local devices");
 331         }
 332     }
 333
 334     my $vollist = PVE::QemuServer::get_vm_volumes($conf);
 335
 336     foreach my $volid (@$vollist) {
 337         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
 338
 339         # check if storage is available on both nodes
 340         my $targetsid = PVE::QemuServer::map_storage($self->{opts}->{storagemap}, $sid);
 341
 342         my $scfg = PVE::Storage::storage_check_node($self->{storecfg}, $sid);
 343         PVE::Storage::storage_check_node($self->{storecfg}, $targetsid, $self->{node});
 344
 345         if ($scfg->{shared}) {
 346             # PVE::Storage::activate_storage checks this for non-shared storages
 347             my $plugin = PVE::Storage::Plugin->lookup($scfg->{type});
 348             warn "Used shared storage '$sid' is not online on source node!\n"
 349                 if !$plugin->check_connection($sid, $scfg);
 350         }
 351     }
 352
 353     # test ssh connection
 354     my $cmd = [ @{$self->{rem_ssh}}, '/bin/true' ];
 355     eval { $self->cmd_quiet($cmd); };
 356     die "Can't connect to destination address using public key\n" if $@;
 357
 358     return $running;
 359 }
 360
 361 sub scan_local_volumes {
 362     my ($self, $vmid) = @_;
 363
 364     my $conf = $self->{vmconf};
 365
 366     # local volumes which have been copied
 367     # and their old_id => new_id pairs
 368     $self->{volume_map} = {};
 369     $self->{local_volumes} = {};
 370
 371     my $storecfg = $self->{storecfg};
 372     eval {
 373
 374         # found local volumes and their origin
 375         my $local_volumes = $self->{local_volumes};
 376         my $local_volumes_errors = {};
 377         my $other_errors = [];
 378         my $abort = 0;
 379
 380         my $log_error = sub {
 381             my ($msg, $volid) = @_;
 382
 383             if (defined($volid)) {
 384                 $local_volumes_errors->{$volid} = $msg;
 385             } else {
 386                 push @$other_errors, $msg;
 387             }
 388             $abort = 1;
 389         };
 390
 391         my @sids = PVE::Storage::storage_ids($storecfg);
 392         foreach my $storeid (@sids) {
 393             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
 394             next if $scfg->{shared};
 395             next if !PVE::Storage::storage_check_enabled($storecfg, $storeid, undef, 1);
 396
 397             # get list from PVE::Storage (for unused volumes)
 398             my $dl = PVE::Storage::vdisk_list($storecfg, $storeid, $vmid);
 399
 400             next if @{$dl->{$storeid}} == 0;
 401
 402             my $targetsid = PVE::QemuServer::map_storage($self->{opts}->{storagemap}, $storeid);
 403             # check if storage is available on target node
 404             PVE::Storage::storage_check_node($storecfg, $targetsid, $self->{node});
 405
 406             # grandfather in existing mismatches
 407             if ($targetsid ne $storeid) {
 408                 my $target_scfg = PVE::Storage::storage_config($storecfg, $targetsid);
 409                 die "content type 'images' is not available on storage '$targetsid'\n"
 410                     if !$target_scfg->{content}->{images};
 411             }
 412
 413             my $bwlimit = PVE::Storage::get_bandwidth_limit(
 414                 'migration',
 415                 [$targetsid, $storeid],
 416                 $self->{opts}->{bwlimit},
 417             );
 418
 419             PVE::Storage::foreach_volid($dl, sub {
 420                 my ($volid, $sid, $volinfo) = @_;
 421
 422                 $local_volumes->{$volid}->{ref} = 'storage';
 423                 $local_volumes->{$volid}->{size} = $volinfo->{size};
 424                 $local_volumes->{$volid}->{targetsid} = $targetsid;
 425                 $local_volumes->{$volid}->{bwlimit} = $bwlimit;
 426
 427                 # If with_snapshots is not set for storage migrate, it tries to use
 428                 # a raw+size stream, but on-the-fly conversion from qcow2 to raw+size
 429                 # back to qcow2 is currently not possible.
 430                 $local_volumes->{$volid}->{snapshots} = ($volinfo->{format} =~ /^(?:qcow2|vmdk)$/);
 431                 $local_volumes->{$volid}->{format} = $volinfo->{format};
 432             });
 433         }
 434
 435         my $replicatable_volumes = !$self->{replication_jobcfg} ? {}
 436             : PVE::QemuConfig->get_replicatable_volumes($storecfg, $vmid, $conf, 0, 1);
 437         foreach my $volid (keys %{$replicatable_volumes}) {
 438             $local_volumes->{$volid}->{replicated} = 1;
 439         }
 440
 441         my $test_volid = sub {
 442             my ($volid, $attr) = @_;
 443
 444             if ($volid =~ m|^/|) {
 445                 return if $attr->{shared};
 446                 $local_volumes->{$volid}->{ref} = 'config';
 447                 die "local file/device\n";
 448             }
 449
 450             my $snaprefs = $attr->{referenced_in_snapshot};
 451
 452             if ($attr->{cdrom}) {
 453                 if ($volid eq 'cdrom') {
 454                     my $msg = "can't migrate local cdrom drive";
 455                     if (defined($snaprefs) && !$attr->{referenced_in_config}) {
 456                         my $snapnames = join(', ', sort keys %$snaprefs);
 457                         $msg .= " (referenced in snapshot - $snapnames)";
 458                     }
 459                     &$log_error("$msg\n");
 460                     return;
 461                 }
 462                 return if $volid eq 'none';
 463             }
 464
 465             my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
 466
 467             my $targetsid = PVE::QemuServer::map_storage($self->{opts}->{storagemap}, $sid);
 468             # check if storage is available on both nodes
 469             my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
 470             PVE::Storage::storage_check_node($storecfg, $targetsid, $self->{node});
 471
 472             return if $scfg->{shared};
 473
 474             $local_volumes->{$volid}->{ref} = $attr->{referenced_in_config} ? 'config' : 'snapshot';
 475             $local_volumes->{$volid}->{ref} = 'storage' if $attr->{is_unused};
 476
 477             $local_volumes->{$volid}->{is_vmstate} = $attr->{is_vmstate} ? 1 : 0;
 478
 479             $local_volumes->{$volid}->{drivename} = $attr->{drivename}
 480                 if $attr->{drivename};
 481
 482             if ($attr->{cdrom}) {
 483                 if ($volid =~ /vm-\d+-cloudinit/) {
 484                     $local_volumes->{$volid}->{ref} = 'generated';
 485                     return;
 486                 }
 487                 die "local cdrom image\n";
 488             }
 489
 490             my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
 491
 492             die "owned by other VM (owner = VM $owner)\n"
 493                 if !$owner || ($owner != $vmid);
 494
 495             return if $attr->{is_vmstate};
 496
 497             if (defined($snaprefs)) {
 498                 $local_volumes->{$volid}->{snapshots} = 1;
 499
 500                 # we cannot migrate shapshots on local storage
 501                 # exceptions: 'zfspool' or 'qcow2' files (on directory storage)
 502
 503                 die "online storage migration not possible if snapshot exists\n" if $self->{running};
 504                 if (!($scfg->{type} eq 'zfspool' || $local_volumes->{$volid}->{format} eq 'qcow2')) {
 505                     die "non-migratable snapshot exists\n";
 506                 }
 507             }
 508
 509             die "referenced by linked clone(s)\n"
 510                 if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
 511         };
 512
 513         PVE::QemuServer::foreach_volid($conf, sub {
 514             my ($volid, $attr) = @_;
 515             eval { $test_volid->($volid, $attr); };
 516             if (my $err = $@) {
 517                 &$log_error($err, $volid);
 518             }
 519         });
 520
 521         foreach my $vol (sort keys %$local_volumes) {
 522             my $type = $replicatable_volumes->{$vol} ? 'local, replicated' : 'local';
 523             my $ref = $local_volumes->{$vol}->{ref};
 524             if ($ref eq 'storage') {
 525                 $self->log('info', "found $type disk '$vol' (via storage)\n");
 526             } elsif ($ref eq 'config') {
 527                 &$log_error("can't live migrate attached local disks without with-local-disks option\n", $vol)
 528                     if $self->{running} && !$self->{opts}->{"with-local-disks"};
 529                 $self->log('info', "found $type disk '$vol' (in current VM config)\n");
 530             } elsif ($ref eq 'snapshot') {
 531                 $self->log('info', "found $type disk '$vol' (referenced by snapshot(s))\n");
 532             } elsif ($ref eq 'generated') {
 533                 $self->log('info', "found generated disk '$vol' (in current VM config)\n");
 534             } else {
 535                 $self->log('info', "found $type disk '$vol'\n");
 536             }
 537         }
 538
 539         foreach my $vol (sort keys %$local_volumes_errors) {
 540             $self->log('warn', "can't migrate local disk '$vol': $local_volumes_errors->{$vol}");
 541         }
 542         foreach my $err (@$other_errors) {
 543             $self->log('warn', "$err");
 544         }
 545
 546         if ($abort) {
 547             die "can't migrate VM - check log\n";
 548         }
 549
 550         # additional checks for local storage
 551         foreach my $volid (keys %$local_volumes) {
 552             my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
 553             my $scfg =  PVE::Storage::storage_config($storecfg, $sid);
 554
 555             my $migratable = $scfg->{type} =~ /^(?:dir|zfspool|lvmthin|lvm)$/;
 556
 557             die "can't migrate '$volid' - storage type '$scfg->{type}' not supported\n"
 558                 if !$migratable;
 559
 560             # image is a linked clone on local storage, se we can't migrate.
 561             if (my $basename = (PVE::Storage::parse_volname($storecfg, $volid))[3]) {
 562                 die "can't migrate '$volid' as it's a clone of '$basename'";
 563             }
 564         }
 565
 566         foreach my $volid (sort keys %$local_volumes) {
 567             my $ref = $local_volumes->{$volid}->{ref};
 568             if ($self->{running} && $ref eq 'config') {
 569                 $local_volumes->{$volid}->{migration_mode} = 'online';
 570             } elsif ($self->{running} && $ref eq 'generated') {
 571                 die "can't live migrate VM with local cloudinit disk. use a shared storage instead\n";
 572             } else {
 573                 $local_volumes->{$volid}->{migration_mode} = 'offline';
 574             }
 575         }
 576     };
 577     die "Problem found while scanning volumes - $@" if $@;
 578 }
 579
 580 sub handle_replication {
 581     my ($self, $vmid) = @_;
 582
 583     my $conf = $self->{vmconf};
 584     my $local_volumes = $self->{local_volumes};
 585
 586     return if !$self->{replication_jobcfg};
 587     if ($self->{running}) {
 588
 589         my $version = PVE::QemuServer::kvm_user_version();
 590         if (!min_version($version, 4, 2)) {
 591             die "can't live migrate VM with replicated volumes, pve-qemu to old (< 4.2)!\n"
 592         }
 593
 594         my @live_replicatable_volumes = $self->filter_local_volumes('online', 1);
 595         foreach my $volid (@live_replicatable_volumes) {
 596             my $drive = $local_volumes->{$volid}->{drivename};
 597             die "internal error - no drive for '$volid'\n" if !defined($drive);
 598
 599             my $bitmap = "repl_$drive";
 600
 601             # start tracking before replication to get full delta + a few duplicates
 602             $self->log('info', "$drive: start tracking writes using block-dirty-bitmap '$bitmap'");
 603             mon_cmd($vmid, 'block-dirty-bitmap-add', node => "drive-$drive", name => $bitmap);
 604
 605             # other info comes from target node in phase 2
 606             $self->{target_drive}->{$drive}->{bitmap} = $bitmap;
 607         }
 608     }
 609     $self->log('info', "replicating disk images");
 610
 611     my $start_time = time();
 612     my $logfunc = sub { $self->log('info', shift) };
 613     my $actual_replicated_volumes = PVE::Replication::run_replication(
 614        'PVE::QemuConfig', $self->{replication_jobcfg}, $start_time, $start_time, $logfunc);
 615
 616     # extra safety check
 617     my @replicated_volumes = $self->filter_local_volumes(undef, 1);
 618     foreach my $volid (@replicated_volumes) {
 619         die "expected volume '$volid' to get replicated, but it wasn't\n"
 620             if !$actual_replicated_volumes->{$volid};
 621     }
 622 }
 623
 624 sub config_update_local_disksizes {
 625     my ($self) = @_;
 626
 627     my $conf = $self->{vmconf};
 628     my $local_volumes = $self->{local_volumes};
 629
 630     PVE::QemuConfig->foreach_volume($conf, sub {
 631         my ($key, $drive) = @_;
 632         return if $key eq 'efidisk0'; # skip efidisk, will be handled later
 633
 634         my $volid = $drive->{file};
 635         return if !defined($local_volumes->{$volid}); # only update sizes for local volumes
 636
 637         my ($updated, $msg) = PVE::QemuServer::Drive::update_disksize($drive, $local_volumes->{$volid}->{size});
 638         if (defined($updated)) {
 639             $conf->{$key} = PVE::QemuServer::print_drive($updated);
 640             $self->log('info', "drive '$key': $msg");
 641         }
 642     });
 643
 644     # we want to set the efidisk size in the config to the size of the
 645     # real OVMF_VARS.fd image, else we can create a too big image, which does not work
 646     if (defined($conf->{efidisk0})) {
 647         PVE::QemuServer::update_efidisk_size($conf);
 648     }
 649 }
 650
 651 sub filter_local_volumes {
 652     my ($self, $migration_mode, $replicated) = @_;
 653
 654     my $volumes = $self->{local_volumes};
 655     my @filtered_volids;
 656
 657     foreach my $volid (sort keys %{$volumes}) {
 658         next if defined($migration_mode) && safe_string_ne($volumes->{$volid}->{migration_mode}, $migration_mode);
 659         next if defined($replicated) && safe_boolean_ne($volumes->{$volid}->{replicated}, $replicated);
 660         push @filtered_volids, $volid;
 661     }
 662
 663     return @filtered_volids;
 664 }
 665
 666 sub sync_offline_local_volumes {
 667     my ($self) = @_;
 668
 669     my $local_volumes = $self->{local_volumes};
 670     my @volids = $self->filter_local_volumes('offline', 0);
 671
 672     my $storecfg = $self->{storecfg};
 673     my $opts = $self->{opts};
 674
 675     $self->log('info', "copying local disk images") if scalar(@volids);
 676
 677     foreach my $volid (@volids) {
 678         my $targetsid = $local_volumes->{$volid}->{targetsid};
 679         my $bwlimit = $local_volumes->{$volid}->{bwlimit};
 680         $bwlimit = $bwlimit * 1024 if defined($bwlimit); # storage_migrate uses bps
 681
 682         my $storage_migrate_opts = {
 683             'ratelimit_bps' => $bwlimit,
 684             'insecure' => $opts->{migration_type} eq 'insecure',
 685             'with_snapshots' => $local_volumes->{$volid}->{snapshots},
 686             'allow_rename' => !$local_volumes->{$volid}->{is_vmstate},
 687         };
 688
 689         my $logfunc = sub { $self->log('info', $_[0]); };
 690         my $new_volid = eval {
 691             PVE::Storage::storage_migrate($storecfg, $volid, $self->{ssh_info},
 692                                           $targetsid, $storage_migrate_opts, $logfunc);
 693         };
 694         if (my $err = $@) {
 695             die "storage migration for '$volid' to storage '$targetsid' failed - $err\n";
 696         }
 697
 698         $self->{volume_map}->{$volid} = $new_volid;
 699         $self->log('info', "volume '$volid' is '$new_volid' on the target\n");
 700
 701         eval { PVE::Storage::deactivate_volumes($storecfg, [$volid]); };
 702         if (my $err = $@) {
 703             $self->log('warn', $err);
 704         }
 705     }
 706 }
 707
 708 sub cleanup_remotedisks {
 709     my ($self) = @_;
 710
 711     my $local_volumes = $self->{local_volumes};
 712
 713     foreach my $volid (values %{$self->{volume_map}}) {
 714         # don't clean up replicated disks!
 715         next if $local_volumes->{$volid}->{replicated};
 716
 717         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
 718
 719         my $cmd = [@{$self->{rem_ssh}}, 'pvesm', 'free', "$storeid:$volname"];
 720
 721         eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
 722         if (my $err = $@) {
 723             $self->log('err', $err);
 724             $self->{errors} = 1;
 725         }
 726     }
 727 }
 728
 729 sub cleanup_bitmaps {
 730     my ($self) = @_;
 731     foreach my $drive (keys %{$self->{target_drive}}) {
 732         my $bitmap = $self->{target_drive}->{$drive}->{bitmap};
 733         next if !$bitmap;
 734         $self->log('info', "$drive: removing block-dirty-bitmap '$bitmap'");
 735         mon_cmd($self->{vmid}, 'block-dirty-bitmap-remove', node => "drive-$drive", name => $bitmap);
 736     }
 737 }
 738
 739 sub phase1 {
 740     my ($self, $vmid) = @_;
 741
 742     $self->log('info', "starting migration of VM $vmid to node '$self->{node}' ($self->{nodeip})");
 743
 744     my $conf = $self->{vmconf};
 745
 746     # set migrate lock in config file
 747     $conf->{lock} = 'migrate';
 748     PVE::QemuConfig->write_config($vmid, $conf);
 749
 750     $self->scan_local_volumes($vmid);
 751
 752     # fix disk sizes to match their actual size and write changes,
 753     # so that the target allocates the correct volumes
 754     $self->config_update_local_disksizes();
 755     PVE::QemuConfig->write_config($vmid, $conf);
 756
 757     $self->handle_replication($vmid);
 758
 759     $self->sync_offline_local_volumes();
 760 };
 761
 762 sub phase1_cleanup {
 763     my ($self, $vmid, $err) = @_;
 764
 765     $self->log('info', "aborting phase 1 - cleanup resources");
 766
 767     my $conf = $self->{vmconf};
 768     delete $conf->{lock};
 769     eval { PVE::QemuConfig->write_config($vmid, $conf) };
 770     if (my $err = $@) {
 771         $self->log('err', $err);
 772     }
 773
 774     eval { $self->cleanup_remotedisks() };
 775     if (my $err = $@) {
 776         $self->log('err', $err);
 777     }
 778
 779     eval { $self->cleanup_bitmaps() };
 780     if (my $err =$@) {
 781         $self->log('err', $err);
 782     }
 783 }
 784
 785 sub phase2 {
 786     my ($self, $vmid) = @_;
 787
 788     my $conf = $self->{vmconf};
 789     my $local_volumes = $self->{local_volumes};
 790     my @online_local_volumes = $self->filter_local_volumes('online');
 791
 792     $self->{storage_migration} = 1 if scalar(@online_local_volumes);
 793
 794     $self->log('info', "starting VM $vmid on remote node '$self->{node}'");
 795
 796     my $raddr;
 797     my $rport;
 798     my $ruri; # the whole migration dst. URI (protocol:address[:port])
 799     my $nodename = PVE::INotify::nodename();
 800
 801     ## start on remote node
 802     my $cmd = [@{$self->{rem_ssh}}];
 803
 804     my $spice_ticket;
 805     if (PVE::QemuServer::vga_conf_has_spice($conf->{vga})) {
 806         my $res = mon_cmd($vmid, 'query-spice');
 807         $spice_ticket = $res->{ticket};
 808     }
 809
 810     push @$cmd , 'qm', 'start', $vmid, '--skiplock', '--migratedfrom', $nodename;
 811
 812     my $migration_type = $self->{opts}->{migration_type};
 813
 814     push @$cmd, '--migration_type', $migration_type;
 815
 816     push @$cmd, '--migration_network', $self->{opts}->{migration_network}
 817       if $self->{opts}->{migration_network};
 818
 819     if ($migration_type eq 'insecure') {
 820         push @$cmd, '--stateuri', 'tcp';
 821     } else {
 822         push @$cmd, '--stateuri', 'unix';
 823     }
 824
 825     if ($self->{forcemachine}) {
 826         push @$cmd, '--machine', $self->{forcemachine};
 827     }
 828
 829     if ($self->{forcecpu}) {
 830         push @$cmd, '--force-cpu', $self->{forcecpu};
 831     }
 832
 833     if ($self->{storage_migration}) {
 834         push @$cmd, '--targetstorage', ($self->{opts}->{targetstorage} // '1');
 835     }
 836
 837     my $spice_port;
 838     my $unix_socket_info = {};
 839     # version > 0 for unix socket support
 840     my $nbd_protocol_version = 1;
 841     # TODO change to 'spice_ticket: <ticket>\n' in 7.0
 842     my $input = $spice_ticket ? "$spice_ticket\n" : "\n";
 843     $input .= "nbd_protocol_version: $nbd_protocol_version\n";
 844
 845     my @online_replicated_volumes = $self->filter_local_volumes('online', 1);
 846     foreach my $volid (@online_replicated_volumes) {
 847         $input .= "replicated_volume: $volid\n";
 848     }
 849
 850     my $handle_storage_migration_listens = sub {
 851         my ($drive_key, $drivestr, $nbd_uri) = @_;
 852
 853         $self->{stopnbd} = 1;
 854         $self->{target_drive}->{$drive_key}->{drivestr} = $drivestr;
 855         $self->{target_drive}->{$drive_key}->{nbd_uri} = $nbd_uri;
 856
 857         my $source_drive = PVE::QemuServer::parse_drive($drive_key, $conf->{$drive_key});
 858         my $target_drive = PVE::QemuServer::parse_drive($drive_key, $drivestr);
 859         my $source_volid = $source_drive->{file};
 860         my $target_volid = $target_drive->{file};
 861
 862         $self->{volume_map}->{$source_volid} = $target_volid;
 863         $self->log('info', "volume '$source_volid' is '$target_volid' on the target\n");
 864     };
 865
 866     my $target_replicated_volumes = {};
 867
 868     # Note: We try to keep $spice_ticket secret (do not pass via command line parameter)
 869     # instead we pipe it through STDIN
 870     my $exitcode = PVE::Tools::run_command($cmd, input => $input, outfunc => sub {
 871         my $line = shift;
 872
 873         if ($line =~ m/^migration listens on tcp:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+)$/) {
 874             $raddr = $1;
 875             $rport = int($2);
 876             $ruri = "tcp:$raddr:$rport";
 877         }
 878         elsif ($line =~ m!^migration listens on unix:(/run/qemu-server/(\d+)\.migrate)$!) {
 879             $raddr = $1;
 880             die "Destination UNIX sockets VMID does not match source VMID" if $vmid ne $2;
 881             $ruri = "unix:$raddr";
 882         }
 883         elsif ($line =~ m/^migration listens on port (\d+)$/) {
 884             $raddr = "localhost";
 885             $rport = int($1);
 886             $ruri = "tcp:$raddr:$rport";
 887         }
 888         elsif ($line =~ m/^spice listens on port (\d+)$/) {
 889             $spice_port = int($1);
 890         }
 891         elsif ($line =~ m/^storage migration listens on nbd:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+):exportname=(\S+) volume:(\S+)$/) {
 892             my $drivestr = $4;
 893             my $nbd_uri = "nbd:$1:$2:exportname=$3";
 894             my $targetdrive = $3;
 895             $targetdrive =~ s/drive-//g;
 896
 897             $handle_storage_migration_listens->($targetdrive, $drivestr, $nbd_uri);
 898         } elsif ($line =~ m!^storage migration listens on nbd:unix:(/run/qemu-server/(\d+)_nbd\.migrate):exportname=(\S+) volume:(\S+)$!) {
 899             my $drivestr = $4;
 900             die "Destination UNIX socket's VMID does not match source VMID" if $vmid ne $2;
 901             my $nbd_unix_addr = $1;
 902             my $nbd_uri = "nbd:unix:$nbd_unix_addr:exportname=$3";
 903             my $targetdrive = $3;
 904             $targetdrive =~ s/drive-//g;
 905
 906             $handle_storage_migration_listens->($targetdrive, $drivestr, $nbd_uri);
 907             $unix_socket_info->{$nbd_unix_addr} = 1;
 908         } elsif ($line =~ m/^re-using replicated volume: (\S+) - (.*)$/) {
 909             my $drive = $1;
 910             my $volid = $2;
 911             $target_replicated_volumes->{$volid} = $drive;
 912         } elsif ($line =~ m/^QEMU: (.*)$/) {
 913             $self->log('info', "[$self->{node}] $1\n");
 914         }
 915     }, errfunc => sub {
 916         my $line = shift;
 917         $self->log('info', "[$self->{node}] $line");
 918     }, noerr => 1);
 919
 920     die "remote command failed with exit code $exitcode\n" if $exitcode;
 921
 922     die "unable to detect remote migration address\n" if !$raddr;
 923
 924     if (scalar(keys %$target_replicated_volumes) != scalar(@online_replicated_volumes)) {
 925         die "number of replicated disks on source and target node do not match - target node too old?\n"
 926     }
 927
 928     $self->log('info', "start remote tunnel");
 929     $self->start_remote_tunnel($raddr, $rport, $ruri, $unix_socket_info);
 930
 931     my $start = time();
 932
 933     if ($self->{storage_migration}) {
 934         $self->{storage_migration_jobs} = {};
 935         $self->log('info', "starting storage migration");
 936
 937         die "The number of local disks does not match between the source and the destination.\n"
 938             if (scalar(keys %{$self->{target_drive}}) != scalar(@online_local_volumes));
 939         foreach my $drive (keys %{$self->{target_drive}}){
 940             my $target = $self->{target_drive}->{$drive};
 941             my $nbd_uri = $target->{nbd_uri};
 942
 943             my $source_drive = PVE::QemuServer::parse_drive($drive, $conf->{$drive});
 944             my $source_volid = $source_drive->{file};
 945
 946             my $bwlimit = $local_volumes->{$source_volid}->{bwlimit};
 947             my $bitmap = $target->{bitmap};
 948
 949             $self->log('info', "$drive: start migration to $nbd_uri");
 950             PVE::QemuServer::qemu_drive_mirror($vmid, $drive, $nbd_uri, $vmid, undef, $self->{storage_migration_jobs}, 'skip', undef, $bwlimit, $bitmap);
 951         }
 952     }
 953
 954     $self->log('info', "starting online/live migration on $ruri");
 955     $self->{livemigration} = 1;
 956
 957     # load_defaults
 958     my $defaults = PVE::QemuServer::load_defaults();
 959
 960     $self->log('info', "set migration_caps");
 961     eval {
 962         PVE::QemuServer::set_migration_caps($vmid);
 963     };
 964     warn $@ if $@;
 965
 966     my $qemu_migrate_params = {};
 967
 968     # migrate speed can be set via bwlimit (datacenter.cfg and API) and via the
 969     # migrate_speed parameter in qm.conf - take the lower of the two.
 970     my $bwlimit = PVE::Storage::get_bandwidth_limit('migration', undef, $self->{opts}->{bwlimit}) // 0;
 971     my $migrate_speed = $conf->{migrate_speed} // 0;
 972     # migrate_speed is in MB/s, bwlimit in KB/s
 973     $migrate_speed *= 1024;
 974
 975     if ($bwlimit && $migrate_speed) {
 976         $migrate_speed = ($bwlimit < $migrate_speed) ? $bwlimit : $migrate_speed;
 977     } else {
 978         $migrate_speed ||= $bwlimit;
 979     }
 980
 981     # always set migrate speed (overwrite kvm default of 32m) we set a very high
 982     # default of 8192m which is basically unlimited
 983     $migrate_speed ||= ($defaults->{migrate_speed} || 8192) * 1024;
 984
 985     # qmp takes migrate_speed in B/s.
 986     $migrate_speed *= 1024;
 987     $self->log('info', "migration speed limit: $migrate_speed B/s");
 988     $qemu_migrate_params->{'max-bandwidth'} = int($migrate_speed);
 989
 990     my $migrate_downtime = $defaults->{migrate_downtime};
 991     $migrate_downtime = $conf->{migrate_downtime} if defined($conf->{migrate_downtime});
 992     # migrate-set-parameters expects limit in ms
 993     $migrate_downtime *= 1000;
 994     $self->log('info', "migration downtime limit: $migrate_downtime ms");
 995     $qemu_migrate_params->{'downtime-limit'} = int($migrate_downtime);
 996
 997     # set cachesize to 10% of the total memory
 998     my $memory =  $conf->{memory} || $defaults->{memory};
 999     my $cachesize = int($memory * 1048576 / 10);
1000     $cachesize = round_powerof2($cachesize);
1001
1002     $self->log('info', "migration cachesize: $cachesize B");
1003     $qemu_migrate_params->{'xbzrle-cache-size'} = int($cachesize);
1004
1005     $self->log('info', "set migration parameters");
1006     eval {
1007         mon_cmd($vmid, "migrate-set-parameters", %{$qemu_migrate_params});
1008     };
1009     $self->log('info', "migrate-set-parameters error: $@") if $@;
1010
1011     if (PVE::QemuServer::vga_conf_has_spice($conf->{vga})) {
1012         my $rpcenv = PVE::RPCEnvironment::get();
1013         my $authuser = $rpcenv->get_user();
1014
1015         my (undef, $proxyticket) = PVE::AccessControl::assemble_spice_ticket($authuser, $vmid, $self->{node});
1016
1017         my $filename = "/etc/pve/nodes/$self->{node}/pve-ssl.pem";
1018         my $subject =  PVE::AccessControl::read_x509_subject_spice($filename);
1019
1020         $self->log('info', "spice client_migrate_info");
1021
1022         eval {
1023             mon_cmd($vmid, "client_migrate_info", protocol => 'spice',
1024                                                 hostname => $proxyticket, 'port' => 0, 'tls-port' => $spice_port,
1025                                                 'cert-subject' => $subject);
1026         };
1027         $self->log('info', "client_migrate_info error: $@") if $@;
1028
1029     }
1030
1031     $self->log('info', "start migrate command to $ruri");
1032     eval {
1033         mon_cmd($vmid, "migrate", uri => $ruri);
1034     };
1035     my $merr = $@;
1036     $self->log('info', "migrate uri => $ruri failed: $merr") if $merr;
1037
1038     my $lstat = 0;
1039     my $usleep = 1000000;
1040     my $i = 0;
1041     my $err_count = 0;
1042     my $lastrem = undef;
1043     my $downtimecounter = 0;
1044     while (1) {
1045         $i++;
1046         my $avglstat = $lstat ? $lstat / $i : 0;
1047
1048         usleep($usleep);
1049         my $stat;
1050         eval {
1051             $stat = mon_cmd($vmid, "query-migrate");
1052         };
1053         if (my $err = $@) {
1054             $err_count++;
1055             warn "query migrate failed: $err\n";
1056             $self->log('info', "query migrate failed: $err");
1057             if ($err_count <= 5) {
1058                 usleep(1000000);
1059                 next;
1060             }
1061             die "too many query migrate failures - aborting\n";
1062         }
1063
1064         if (defined($stat->{status}) && $stat->{status} =~ m/^(setup)$/im) {
1065             sleep(1);
1066             next;
1067         }
1068
1069         if (defined($stat->{status}) && $stat->{status} =~ m/^(active|completed|failed|cancelled)$/im) {
1070             $merr = undef;
1071             $err_count = 0;
1072             if ($stat->{status} eq 'completed') {
1073                 my $delay = time() - $start;
1074                 if ($delay > 0) {
1075                     my $mbps = sprintf "%.2f", $memory / $delay;
1076                     my $downtime = $stat->{downtime} || 0;
1077                     $self->log('info', "migration speed: $mbps MB/s - downtime $downtime ms");
1078                 }
1079             }
1080
1081             if ($stat->{status} eq 'failed' || $stat->{status} eq 'cancelled') {
1082                 $self->log('info', "migration status error: $stat->{status}");
1083                 die "aborting\n"
1084             }
1085
1086             if ($stat->{status} ne 'active') {
1087                 $self->log('info', "migration status: $stat->{status}");
1088                 last;
1089             }
1090
1091             if ($stat->{ram}->{transferred} ne $lstat) {
1092                 my $trans = $stat->{ram}->{transferred} || 0;
1093                 my $rem = $stat->{ram}->{remaining} || 0;
1094                 my $total = $stat->{ram}->{total} || 0;
1095                 my $xbzrlecachesize = $stat->{"xbzrle-cache"}->{"cache-size"} || 0;
1096                 my $xbzrlebytes = $stat->{"xbzrle-cache"}->{"bytes"} || 0;
1097                 my $xbzrlepages = $stat->{"xbzrle-cache"}->{"pages"} || 0;
1098                 my $xbzrlecachemiss = $stat->{"xbzrle-cache"}->{"cache-miss"} || 0;
1099                 my $xbzrleoverflow = $stat->{"xbzrle-cache"}->{"overflow"} || 0;
1100                 # reduce sleep if remainig memory is lower than the average transfer speed
1101                 $usleep = 100000 if $avglstat && $rem < $avglstat;
1102
1103                 $self->log('info', "migration status: $stat->{status} (transferred ${trans}, " .
1104                            "remaining ${rem}), total ${total})");
1105
1106                 if (${xbzrlecachesize}) {
1107                     $self->log('info', "migration xbzrle cachesize: ${xbzrlecachesize} transferred ${xbzrlebytes} pages ${xbzrlepages} cachemiss ${xbzrlecachemiss} overflow ${xbzrleoverflow}");
1108                 }
1109
1110                 if (($lastrem  && $rem > $lastrem ) || ($rem == 0)) {
1111                     $downtimecounter++;
1112                 }
1113                 $lastrem = $rem;
1114
1115                 if ($downtimecounter > 5) {
1116                     $downtimecounter = 0;
1117                     $migrate_downtime *= 2;
1118                     $self->log('info', "auto-increased downtime to continue migration: $migrate_downtime ms");
1119                     eval {
1120                         # migrate-set-parameters does not touch values not
1121                         # specified, so this only changes downtime-limit
1122                         mon_cmd($vmid, "migrate-set-parameters", 'downtime-limit' => int($migrate_downtime));
1123                     };
1124                     $self->log('info', "migrate-set-parameters error: $@") if $@;
1125                 }
1126
1127             }
1128
1129
1130             $lstat = $stat->{ram}->{transferred};
1131
1132         } else {
1133             die $merr if $merr;
1134             die "unable to parse migration status '$stat->{status}' - aborting\n";
1135         }
1136     }
1137
1138     if ($self->{storage_migration}) {
1139         # finish block-job with block-job-cancel, to disconnect source VM from NBD
1140         # to avoid it trying to re-establish it. We are in blockjob ready state,
1141         # thus, this command changes to it to blockjob complete (see qapi docs)
1142         eval { PVE::QemuServer::qemu_drive_mirror_monitor($vmid, undef, $self->{storage_migration_jobs}, 'cancel'); };
1143         if (my $err = $@) {
1144             die "Failed to complete storage migration: $err\n";
1145         }
1146     }
1147 }
1148
1149 sub phase2_cleanup {
1150     my ($self, $vmid, $err) = @_;
1151
1152     return if !$self->{errors};
1153     $self->{phase2errors} = 1;
1154
1155     $self->log('info', "aborting phase 2 - cleanup resources");
1156
1157     $self->log('info', "migrate_cancel");
1158     eval {
1159         mon_cmd($vmid, "migrate_cancel");
1160     };
1161     $self->log('info', "migrate_cancel error: $@") if $@;
1162
1163     my $conf = $self->{vmconf};
1164     delete $conf->{lock};
1165     eval { PVE::QemuConfig->write_config($vmid, $conf) };
1166     if (my $err = $@) {
1167         $self->log('err', $err);
1168     }
1169
1170     # cleanup ressources on target host
1171     if ($self->{storage_migration}) {
1172         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $self->{storage_migration_jobs}) };
1173         if (my $err = $@) {
1174             $self->log('err', $err);
1175         }
1176     }
1177
1178     eval { $self->cleanup_bitmaps() };
1179     if (my $err =$@) {
1180         $self->log('err', $err);
1181     }
1182
1183     my $nodename = PVE::INotify::nodename();
1184
1185     my $cmd = [@{$self->{rem_ssh}}, 'qm', 'stop', $vmid, '--skiplock', '--migratedfrom', $nodename];
1186     eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
1187     if (my $err = $@) {
1188         $self->log('err', $err);
1189         $self->{errors} = 1;
1190     }
1191
1192     # cleanup after stopping, otherwise disks might be in-use by target VM!
1193     eval { PVE::QemuMigrate::cleanup_remotedisks($self) };
1194     if (my $err = $@) {
1195         $self->log('err', $err);
1196     }
1197
1198
1199     if ($self->{tunnel}) {
1200         eval { finish_tunnel($self, $self->{tunnel});  };
1201         if (my $err = $@) {
1202             $self->log('err', $err);
1203             $self->{errors} = 1;
1204         }
1205     }
1206 }
1207
1208 sub phase3 {
1209     my ($self, $vmid) = @_;
1210
1211     return;
1212 }
1213
1214 sub phase3_cleanup {
1215     my ($self, $vmid, $err) = @_;
1216
1217     my $conf = $self->{vmconf};
1218     return if $self->{phase2errors};
1219
1220     my $tunnel = $self->{tunnel};
1221
1222     if ($self->{volume_map}) {
1223         my $target_drives = $self->{target_drive};
1224
1225         # FIXME: for NBD storage migration we now only update the volid, and
1226         # not the full drivestr from the target node. Workaround that until we
1227         # got some real rescan, to avoid things like wrong format in the drive
1228         delete $conf->{$_} for keys %$target_drives;
1229         PVE::QemuConfig->update_volume_ids($conf, $self->{volume_map});
1230
1231         for my $drive (keys %$target_drives) {
1232             $conf->{$drive} = $target_drives->{$drive}->{drivestr};
1233         }
1234         PVE::QemuConfig->write_config($vmid, $conf);
1235     }
1236
1237     # transfer replication state before move config
1238     $self->transfer_replication_state() if $self->{is_replicated};
1239     PVE::QemuConfig->move_config_to_node($vmid, $self->{node});
1240     $self->switch_replication_job_target() if $self->{is_replicated};
1241
1242     if ($self->{livemigration}) {
1243         if ($self->{stopnbd}) {
1244             $self->log('info', "stopping NBD storage migration server on target.");
1245             # stop nbd server on remote vm - requirement for resume since 2.9
1246             my $cmd = [@{$self->{rem_ssh}}, 'qm', 'nbdstop', $vmid];
1247
1248             eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
1249             if (my $err = $@) {
1250                 $self->log('err', $err);
1251                 $self->{errors} = 1;
1252             }
1253         }
1254
1255         # config moved and nbd server stopped - now we can resume vm on target
1256         if ($tunnel && $tunnel->{version} && $tunnel->{version} >= 1) {
1257             eval {
1258                 $self->write_tunnel($tunnel, 30, "resume $vmid");
1259             };
1260             if (my $err = $@) {
1261                 $self->log('err', $err);
1262                 $self->{errors} = 1;
1263             }
1264         } else {
1265             my $cmd = [@{$self->{rem_ssh}}, 'qm', 'resume', $vmid, '--skiplock', '--nocheck'];
1266             my $logf = sub {
1267                 my $line = shift;
1268                 $self->log('err', $line);
1269             };
1270             eval { PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => $logf); };
1271             if (my $err = $@) {
1272                 $self->log('err', $err);
1273                 $self->{errors} = 1;
1274             }
1275         }
1276
1277         if ($self->{storage_migration} && PVE::QemuServer::parse_guest_agent($conf)->{fstrim_cloned_disks} && $self->{running}) {
1278             my $cmd = [@{$self->{rem_ssh}}, 'qm', 'guest', 'cmd', $vmid, 'fstrim'];
1279             eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
1280         }
1281     }
1282
1283     # close tunnel on successful migration, on error phase2_cleanup closed it
1284     if ($tunnel) {
1285         eval { finish_tunnel($self, $tunnel);  };
1286         if (my $err = $@) {
1287             $self->log('err', $err);
1288             $self->{errors} = 1;
1289         }
1290     }
1291
1292     eval {
1293         my $timer = 0;
1294         if (PVE::QemuServer::vga_conf_has_spice($conf->{vga}) && $self->{running}) {
1295             $self->log('info', "Waiting for spice server migration");
1296             while (1) {
1297                 my $res = mon_cmd($vmid, 'query-spice');
1298                 last if int($res->{'migrated'}) == 1;
1299                 last if $timer > 50;
1300                 $timer ++;
1301                 usleep(200000);
1302             }
1303         }
1304     };
1305
1306     # always stop local VM
1307     eval { PVE::QemuServer::vm_stop($self->{storecfg}, $vmid, 1, 1); };
1308     if (my $err = $@) {
1309         $self->log('err', "stopping vm failed - $err");
1310         $self->{errors} = 1;
1311     }
1312
1313     # always deactivate volumes - avoid lvm LVs to be active on several nodes
1314     eval {
1315         my $vollist = PVE::QemuServer::get_vm_volumes($conf);
1316         PVE::Storage::deactivate_volumes($self->{storecfg}, $vollist);
1317     };
1318     if (my $err = $@) {
1319         $self->log('err', $err);
1320         $self->{errors} = 1;
1321     }
1322
1323     my @not_replicated_volumes = $self->filter_local_volumes(undef, 0);
1324
1325     # destroy local copies
1326     foreach my $volid (@not_replicated_volumes) {
1327         eval { PVE::Storage::vdisk_free($self->{storecfg}, $volid); };
1328         if (my $err = $@) {
1329             $self->log('err', "removing local copy of '$volid' failed - $err");
1330             $self->{errors} = 1;
1331             last if $err =~ /^interrupted by signal$/;
1332         }
1333     }
1334
1335     # clear migrate lock
1336     my $cmd = [ @{$self->{rem_ssh}}, 'qm', 'unlock', $vmid ];
1337     $self->cmd_logerr($cmd, errmsg => "failed to clear migrate lock");
1338 }
1339
1340 sub final_cleanup {
1341     my ($self, $vmid) = @_;
1342
1343     # nothing to do
1344 }
1345
1346 sub round_powerof2 {
1347     return 1 if $_[0] < 2;
1348     return 2 << int(log($_[0]-1)/log(2));
1349 }
1350
1351 1;