1 package PVE
::QemuMigrate
;
8 use POSIX
qw( WNOHANG );
9 use Time
::HiRes
qw( usleep );
13 use PVE
::RPCEnvironment
;
15 use PVE
::ReplicationConfig
;
16 use PVE
::ReplicationState
;
21 use PVE
::QemuServer
::CPUConfig
;
22 use PVE
::QemuServer
::Drive
;
23 use PVE
::QemuServer
::Helpers
qw(min_version);
24 use PVE
::QemuServer
::Machine
;
25 use PVE
::QemuServer
::Monitor
qw(mon_cmd);
28 use PVE
::AbstractMigrate
;
29 use base
qw(PVE::AbstractMigrate);
31 sub fork_command_pipe
{
32 my ($self, $cmd) = @_;
34 my $reader = IO
::File-
>new();
35 my $writer = IO
::File-
>new();
41 eval { $cpid = open2
($reader, $writer, @$cmd); };
46 if ($orig_pid != $$) {
47 $self->log('err', "can't fork command pipe\n");
54 return { writer
=> $writer, reader
=> $reader, pid
=> $cpid };
57 sub finish_command_pipe
{
58 my ($self, $cmdpipe, $timeout) = @_;
60 my $cpid = $cmdpipe->{pid
};
61 return if !defined($cpid);
63 my $writer = $cmdpipe->{writer
};
64 my $reader = $cmdpipe->{reader
};
69 my $collect_child_process = sub {
70 my $res = waitpid($cpid, WNOHANG
);
71 if (defined($res) && ($res == $cpid)) {
72 delete $cmdpipe->{cpid
};
80 for (my $i = 0; $i < $timeout; $i++) {
81 return if &$collect_child_process();
86 $self->log('info', "ssh tunnel still running - terminating now with SIGTERM\n");
90 for (my $i = 0; $i < 10; $i++) {
91 return if &$collect_child_process();
95 $self->log('info', "ssh tunnel still running - terminating now with SIGKILL\n");
99 $self->log('err', "ssh tunnel child process (PID $cpid) couldn't be collected\n")
100 if !&$collect_child_process();
104 my ($self, $tunnel, $timeout) = @_;
106 $timeout = 60 if !defined($timeout);
108 my $reader = $tunnel->{reader
};
112 PVE
::Tools
::run_with_timeout
($timeout, sub { $output = <$reader>; });
114 die "reading from tunnel failed: $@\n" if $@;
122 my ($self, $tunnel, $timeout, $command) = @_;
124 $timeout = 60 if !defined($timeout);
126 my $writer = $tunnel->{writer
};
129 PVE
::Tools
::run_with_timeout
($timeout, sub {
130 print $writer "$command\n";
134 die "writing to tunnel failed: $@\n" if $@;
136 if ($tunnel->{version
} && $tunnel->{version
} >= 1) {
137 my $res = eval { $self->read_tunnel($tunnel, 10); };
138 die "no reply to command '$command': $@\n" if $@;
143 die "tunnel replied '$res' to command '$command'\n";
149 my ($self, $ssh_forward_info) = @_;
151 my @localtunnelinfo = ();
152 foreach my $addr (@$ssh_forward_info) {
153 push @localtunnelinfo, '-L', $addr;
156 my $cmd = [@{$self->{rem_ssh
}}, '-o ExitOnForwardFailure=yes', @localtunnelinfo, '/usr/sbin/qm', 'mtunnel' ];
158 my $tunnel = $self->fork_command_pipe($cmd);
161 my $helo = $self->read_tunnel($tunnel, 60);
162 die "no reply\n" if !$helo;
163 die "no quorum on target node\n" if $helo =~ m/^no quorum$/;
164 die "got strange reply from mtunnel ('$helo')\n"
165 if $helo !~ m/^tunnel online$/;
170 my $ver = $self->read_tunnel($tunnel, 10);
171 if ($ver =~ /^ver (\d+)$/) {
172 $tunnel->{version
} = $1;
173 $self->log('info', "ssh tunnel $ver\n");
175 $err = "received invalid tunnel version string '$ver'\n" if !$err;
180 $self->finish_command_pipe($tunnel);
181 die "can't open migration tunnel - $err";
187 my ($self, $tunnel) = @_;
189 eval { $self->write_tunnel($tunnel, 30, 'quit'); };
192 $self->finish_command_pipe($tunnel, 30);
194 if (my $unix_sockets = $tunnel->{unix_sockets
}) {
195 # ssh does not clean up on local host
196 my $cmd = ['rm', '-f', @$unix_sockets];
197 PVE
::Tools
::run_command
($cmd);
199 # .. and just to be sure check on remote side
200 unshift @{$cmd}, @{$self->{rem_ssh
}};
201 PVE
::Tools
::run_command
($cmd);
208 my ($self, $vmid, $code, @param) = @_;
210 return PVE
::QemuConfig-
>lock_config($vmid, $code, @param);
214 my ($self, $vmid) = @_;
216 my $online = $self->{opts
}->{online
};
218 $self->{storecfg
} = PVE
::Storage
::config
();
221 my $conf = $self->{vmconf
} = PVE
::QemuConfig-
>load_config($vmid);
223 PVE
::QemuConfig-
>check_lock($conf);
226 if (my $pid = PVE
::QemuServer
::check_running
($vmid)) {
227 die "can't migrate running VM without --online\n" if !$online;
230 $self->{forcemachine
} = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
232 # To support custom CPU types, we keep QEMU's "-cpu" parameter intact.
233 # Since the parameter itself contains no reference to a custom model,
234 # this makes migration independent of changes to "cpu-models.conf".
236 my $cpuconf = PVE
::QemuServer
::CPUConfig
::parse_cpu_conf_basic
($conf->{cpu
});
237 if ($cpuconf && PVE
::QemuServer
::CPUConfig
::is_custom_model
($cpuconf->{cputype
})) {
238 $self->{forcecpu
} = PVE
::QemuServer
::CPUConfig
::get_cpu_from_running_vm
($pid);
243 my $loc_res = PVE
::QemuServer
::check_local_resources
($conf, 1);
244 if (scalar @$loc_res) {
245 if ($self->{running
} || !$self->{opts
}->{force
}) {
246 die "can't migrate VM which uses local devices: " . join(", ", @$loc_res) . "\n";
248 $self->log('info', "migrating VM which uses local devices");
252 my $vollist = PVE
::QemuServer
::get_vm_volumes
($conf);
254 my $need_activate = [];
255 foreach my $volid (@$vollist) {
256 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
258 # check if storage is available on both nodes
259 my $targetsid = PVE
::QemuServer
::map_storage
($self->{opts
}->{storagemap
}, $sid);
261 my $scfg = PVE
::Storage
::storage_check_node
($self->{storecfg
}, $sid);
262 PVE
::Storage
::storage_check_node
($self->{storecfg
}, $targetsid, $self->{node
});
264 if ($scfg->{shared
}) {
265 # PVE::Storage::activate_storage checks this for non-shared storages
266 my $plugin = PVE
::Storage
::Plugin-
>lookup($scfg->{type
});
267 warn "Used shared storage '$sid' is not online on source node!\n"
268 if !$plugin->check_connection($sid, $scfg);
270 # only activate if not shared
271 next if ($volid =~ m/vm-\d+-cloudinit/);
272 push @$need_activate, $volid;
277 PVE
::Storage
::activate_volumes
($self->{storecfg
}, $need_activate);
279 # test ssh connection
280 my $cmd = [ @{$self->{rem_ssh
}}, '/bin/true' ];
281 eval { $self->cmd_quiet($cmd); };
282 die "Can't connect to destination address using public key\n" if $@;
288 my ($self, $vmid) = @_;
290 my $conf = $self->{vmconf
};
292 # local volumes which have been copied
293 # and their old_id => new_id pairs
294 $self->{volumes
} = [];
295 $self->{volume_map
} = {};
297 my $storecfg = $self->{storecfg
};
300 # found local volumes and their origin
301 my $local_volumes = {};
302 my $local_volumes_errors = {};
303 my $other_errors = [];
306 my $log_error = sub {
307 my ($msg, $volid) = @_;
309 if (defined($volid)) {
310 $local_volumes_errors->{$volid} = $msg;
312 push @$other_errors, $msg;
317 my @sids = PVE
::Storage
::storage_ids
($storecfg);
318 foreach my $storeid (@sids) {
319 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
320 next if $scfg->{shared
};
321 next if !PVE
::Storage
::storage_check_enabled
($storecfg, $storeid, undef, 1);
323 # get list from PVE::Storage (for unused volumes)
324 my $dl = PVE
::Storage
::vdisk_list
($storecfg, $storeid, $vmid);
326 next if @{$dl->{$storeid}} == 0;
328 my $targetsid = PVE
::QemuServer
::map_storage
($self->{opts
}->{storagemap
}, $storeid);
329 # check if storage is available on target node
330 PVE
::Storage
::storage_check_node
($storecfg, $targetsid, $self->{node
});
332 # grandfather in existing mismatches
333 if ($targetsid ne $storeid) {
334 my $target_scfg = PVE
::Storage
::storage_config
($storecfg, $targetsid);
335 die "content type 'images' is not available on storage '$targetsid'\n"
336 if !$target_scfg->{content
}->{images
};
339 PVE
::Storage
::foreach_volid
($dl, sub {
340 my ($volid, $sid, $volinfo) = @_;
342 $local_volumes->{$volid}->{ref} = 'storage';
344 # If with_snapshots is not set for storage migrate, it tries to use
345 # a raw+size stream, but on-the-fly conversion from qcow2 to raw+size
346 # back to qcow2 is currently not possible.
347 $local_volumes->{$volid}->{snapshots
} = ($volinfo->{format
} =~ /^(?:qcow2|vmdk)$/);
348 $local_volumes->{$volid}->{format
} = $volinfo->{format
};
352 my $rep_cfg = PVE
::ReplicationConfig-
>new();
353 my $replication_jobcfg = $rep_cfg->find_local_replication_job($vmid, $self->{node
});
354 my $replicatable_volumes = !$replication_jobcfg ?
{}
355 : PVE
::QemuConfig-
>get_replicatable_volumes($storecfg, $vmid, $conf, 0, 1);
357 my $test_volid = sub {
358 my ($volid, $attr) = @_;
360 if ($volid =~ m
|^/|) {
361 return if $attr->{shared
};
362 $local_volumes->{$volid}->{ref} = 'config';
363 die "local file/device\n";
366 my $snaprefs = $attr->{referenced_in_snapshot
};
368 if ($attr->{cdrom
}) {
369 if ($volid eq 'cdrom') {
370 my $msg = "can't migrate local cdrom drive";
371 if (defined($snaprefs) && !$attr->{referenced_in_config
}) {
372 my $snapnames = join(', ', sort keys %$snaprefs);
373 $msg .= " (referenced in snapshot - $snapnames)";
375 &$log_error("$msg\n");
378 return if $volid eq 'none';
381 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
383 my $targetsid = PVE
::QemuServer
::map_storage
($self->{opts
}->{storagemap
}, $sid);
384 # check if storage is available on both nodes
385 my $scfg = PVE
::Storage
::storage_check_node
($storecfg, $sid);
386 PVE
::Storage
::storage_check_node
($storecfg, $targetsid, $self->{node
});
388 return if $scfg->{shared
};
390 $local_volumes->{$volid}->{ref} = $attr->{referenced_in_config
} ?
'config' : 'snapshot';
391 $local_volumes->{$volid}->{ref} = 'storage' if $attr->{is_unused
};
393 $local_volumes->{$volid}->{is_vmstate
} = $attr->{is_vmstate
} ?
1 : 0;
395 if ($attr->{cdrom
}) {
396 if ($volid =~ /vm-\d+-cloudinit/) {
397 $local_volumes->{$volid}->{ref} = 'generated';
400 die "local cdrom image\n";
403 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
405 die "owned by other VM (owner = VM $owner)\n"
406 if !$owner || ($owner != $vmid);
408 return if $attr->{is_vmstate
};
410 if (defined($snaprefs)) {
411 $local_volumes->{$volid}->{snapshots
} = 1;
413 # we cannot migrate shapshots on local storage
414 # exceptions: 'zfspool' or 'qcow2' files (on directory storage)
416 die "online storage migration not possible if snapshot exists\n" if $self->{running
};
417 if (!($scfg->{type
} eq 'zfspool' || $local_volumes->{$volid}->{format
} eq 'qcow2')) {
418 die "non-migratable snapshot exists\n";
422 die "referenced by linked clone(s)\n"
423 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
426 PVE
::QemuServer
::foreach_volid
($conf, sub {
427 my ($volid, $attr) = @_;
428 eval { $test_volid->($volid, $attr); };
430 &$log_error($err, $volid);
434 foreach my $vol (sort keys %$local_volumes) {
435 my $type = $replicatable_volumes->{$vol} ?
'local, replicated' : 'local';
436 my $ref = $local_volumes->{$vol}->{ref};
437 if ($ref eq 'storage') {
438 $self->log('info', "found $type disk '$vol' (via storage)\n");
439 } elsif ($ref eq 'config') {
440 &$log_error("can't live migrate attached local disks without with-local-disks option\n", $vol)
441 if $self->{running
} && !$self->{opts
}->{"with-local-disks"};
442 $self->log('info', "found $type disk '$vol' (in current VM config)\n");
443 } elsif ($ref eq 'snapshot') {
444 $self->log('info', "found $type disk '$vol' (referenced by snapshot(s))\n");
445 } elsif ($ref eq 'generated') {
446 $self->log('info', "found generated disk '$vol' (in current VM config)\n");
448 $self->log('info', "found $type disk '$vol'\n");
452 foreach my $vol (sort keys %$local_volumes_errors) {
453 $self->log('warn', "can't migrate local disk '$vol': $local_volumes_errors->{$vol}");
455 foreach my $err (@$other_errors) {
456 $self->log('warn', "$err");
460 die "can't migrate VM - check log\n";
463 # additional checks for local storage
464 foreach my $volid (keys %$local_volumes) {
465 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
466 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
468 my $migratable = $scfg->{type
} =~ /^(?:dir|zfspool|lvmthin|lvm)$/;
470 die "can't migrate '$volid' - storage type '$scfg->{type}' not supported\n"
473 # image is a linked clone on local storage, se we can't migrate.
474 if (my $basename = (PVE
::Storage
::parse_volname
($storecfg, $volid))[3]) {
475 die "can't migrate '$volid' as it's a clone of '$basename'";
479 if ($replication_jobcfg) {
480 if ($self->{running
}) {
482 my $version = PVE
::QemuServer
::kvm_user_version
();
483 if (!min_version
($version, 4, 2)) {
484 die "can't live migrate VM with replicated volumes, pve-qemu to old (< 4.2)!\n"
487 my $live_replicatable_volumes = {};
488 PVE
::QemuConfig-
>foreach_volume($conf, sub {
489 my ($ds, $drive) = @_;
491 my $volid = $drive->{file
};
492 $live_replicatable_volumes->{$ds} = $volid
493 if defined($replicatable_volumes->{$volid});
495 foreach my $drive (keys %$live_replicatable_volumes) {
496 my $volid = $live_replicatable_volumes->{$drive};
498 my $bitmap = "repl_$drive";
500 # start tracking before replication to get full delta + a few duplicates
501 $self->log('info', "$drive: start tracking writes using block-dirty-bitmap '$bitmap'");
502 mon_cmd
($vmid, 'block-dirty-bitmap-add', node
=> "drive-$drive", name
=> $bitmap);
504 # other info comes from target node in phase 2
505 $self->{target_drive
}->{$drive}->{bitmap
} = $bitmap;
508 $self->log('info', "replicating disk images");
510 my $start_time = time();
511 my $logfunc = sub { $self->log('info', shift) };
512 $self->{replicated_volumes
} = PVE
::Replication
::run_replication
(
513 'PVE::QemuConfig', $replication_jobcfg, $start_time, $start_time, $logfunc);
516 # sizes in config have to be accurate for remote node to correctly
517 # allocate disks, rescan to be sure
518 my $volid_hash = PVE
::QemuServer
::scan_volids
($storecfg, $vmid);
519 PVE
::QemuConfig-
>foreach_volume($conf, sub {
520 my ($key, $drive) = @_;
521 return if $key eq 'efidisk0'; # skip efidisk, will be handled later
523 my $volid = $drive->{file
};
524 return if !defined($local_volumes->{$volid}); # only update sizes for local volumes
526 my ($updated, $old_size, $new_size) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volid_hash);
527 if (defined($updated)) {
528 $conf->{$key} = PVE
::QemuServer
::print_drive
($updated);
529 $self->log('info', "size of disk '$updated->{file}' ($key) updated from $old_size to $new_size\n");
533 # we want to set the efidisk size in the config to the size of the
534 # real OVMF_VARS.fd image, else we can create a too big image, which does not work
535 if (defined($conf->{efidisk0
})) {
536 PVE
::QemuServer
::update_efidisk_size
($conf);
539 $self->log('info', "copying local disk images") if scalar(%$local_volumes);
541 foreach my $volid (keys %$local_volumes) {
542 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
543 my $targetsid = PVE
::QemuServer
::map_storage
($self->{opts
}->{storagemap
}, $sid);
544 my $ref = $local_volumes->{$volid}->{ref};
545 if ($self->{running
} && $ref eq 'config') {
546 push @{$self->{online_local_volumes
}}, $volid;
547 } elsif ($ref eq 'generated') {
548 die "can't live migrate VM with local cloudinit disk. use a shared storage instead\n" if $self->{running
};
549 # skip all generated volumes but queue them for deletion in phase3_cleanup
550 push @{$self->{volumes
}}, $volid;
553 next if $self->{replicated_volumes
}->{$volid};
554 push @{$self->{volumes
}}, $volid;
555 my $opts = $self->{opts
};
556 # use 'migrate' limit for transfer to other node
557 my $bwlimit = PVE
::Storage
::get_bandwidth_limit
('migration', [$targetsid, $sid], $opts->{bwlimit
});
558 # JSONSchema and get_bandwidth_limit use kbps - storage_migrate bps
559 $bwlimit = $bwlimit * 1024 if defined($bwlimit);
561 my $storage_migrate_opts = {
562 'bwlimit' => $bwlimit,
563 'insecure' => $opts->{migration_type
} eq 'insecure',
564 'with_snapshots' => $local_volumes->{$volid}->{snapshots
},
565 'allow_rename' => !$local_volumes->{$volid}->{is_vmstate
},
568 my $logfunc = sub { $self->log('info', $_[0]); };
569 my $new_volid = eval {
570 PVE
::Storage
::storage_migrate
($storecfg, $volid, $self->{ssh_info
},
571 $targetsid, $storage_migrate_opts, $logfunc);
574 die "storage migration for '$volid' to storage '$targetsid' failed - $err\n";
577 $self->{volume_map
}->{$volid} = $new_volid;
578 $self->log('info', "volume '$volid' is '$new_volid' on the target\n");
582 die "Failed to sync data - $@" if $@;
585 sub cleanup_remotedisks
{
588 foreach my $target_drive (keys %{$self->{target_drive
}}) {
589 my $drivestr = $self->{target_drive
}->{$target_drive}->{drivestr
};
590 next if !defined($drivestr);
592 my $drive = PVE
::QemuServer
::parse_drive
($target_drive, $drivestr);
594 # don't clean up replicated disks!
595 next if defined($self->{replicated_volumes
}->{$drive->{file
}});
597 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
599 my $cmd = [@{$self->{rem_ssh
}}, 'pvesm', 'free', "$storeid:$volname"];
601 eval{ PVE
::Tools
::run_command
($cmd, outfunc
=> sub {}, errfunc
=> sub {}) };
603 $self->log('err', $err);
609 sub cleanup_bitmaps
{
611 foreach my $drive (keys %{$self->{target_drive
}}) {
612 my $bitmap = $self->{target_drive
}->{$drive}->{bitmap
};
614 $self->log('info', "$drive: removing block-dirty-bitmap '$bitmap'");
615 mon_cmd
($self->{vmid
}, 'block-dirty-bitmap-remove', node
=> "drive-$drive", name
=> $bitmap);
620 my ($self, $vmid) = @_;
622 $self->log('info', "starting migration of VM $vmid to node '$self->{node}' ($self->{nodeip})");
624 my $conf = $self->{vmconf
};
626 # set migrate lock in config file
627 $conf->{lock} = 'migrate';
628 PVE
::QemuConfig-
>write_config($vmid, $conf);
630 sync_disks
($self, $vmid);
632 # sync_disks fixes disk sizes to match their actual size, write changes so
633 # target allocates correct volumes
634 PVE
::QemuConfig-
>write_config($vmid, $conf);
638 my ($self, $vmid, $err) = @_;
640 $self->log('info', "aborting phase 1 - cleanup resources");
642 my $conf = $self->{vmconf
};
643 delete $conf->{lock};
644 eval { PVE
::QemuConfig-
>write_config($vmid, $conf) };
646 $self->log('err', $err);
649 if ($self->{volumes
}) {
650 foreach my $volid (@{$self->{volumes
}}) {
651 $self->log('err', "found stale volume copy '$volid' on node '$self->{node}'");
652 # fixme: try to remove ?
656 eval { $self->cleanup_bitmaps() };
658 $self->log('err', $err);
664 my ($self, $vmid) = @_;
666 my $conf = $self->{vmconf
};
668 $self->log('info', "starting VM $vmid on remote node '$self->{node}'");
672 my $ruri; # the whole migration dst. URI (protocol:address[:port])
673 my $nodename = PVE
::INotify
::nodename
();
675 ## start on remote node
676 my $cmd = [@{$self->{rem_ssh
}}];
679 if (PVE
::QemuServer
::vga_conf_has_spice
($conf->{vga
})) {
680 my $res = mon_cmd
($vmid, 'query-spice');
681 $spice_ticket = $res->{ticket
};
684 push @$cmd , 'qm', 'start', $vmid, '--skiplock', '--migratedfrom', $nodename;
686 my $migration_type = $self->{opts
}->{migration_type
};
688 push @$cmd, '--migration_type', $migration_type;
690 push @$cmd, '--migration_network', $self->{opts
}->{migration_network
}
691 if $self->{opts
}->{migration_network
};
693 if ($migration_type eq 'insecure') {
694 push @$cmd, '--stateuri', 'tcp';
696 push @$cmd, '--stateuri', 'unix';
699 if ($self->{forcemachine
}) {
700 push @$cmd, '--machine', $self->{forcemachine
};
703 if ($self->{forcecpu
}) {
704 push @$cmd, '--force-cpu', $self->{forcecpu
};
707 if ($self->{online_local_volumes
}) {
708 push @$cmd, '--targetstorage', ($self->{opts
}->{targetstorage
} // '1');
712 my $unix_socket_info = {};
713 # version > 0 for unix socket support
714 my $nbd_protocol_version = 1;
715 # TODO change to 'spice_ticket: <ticket>\n' in 7.0
716 my $input = $spice_ticket ?
"$spice_ticket\n" : "\n";
717 $input .= "nbd_protocol_version: $nbd_protocol_version\n";
719 my $number_of_online_replicated_volumes = 0;
721 # prevent auto-vivification
722 if ($self->{online_local_volumes
}) {
723 foreach my $volid (@{$self->{online_local_volumes
}}) {
724 next if !$self->{replicated_volumes
}->{$volid};
725 $number_of_online_replicated_volumes++;
726 $input .= "replicated_volume: $volid\n";
730 my $target_replicated_volumes = {};
732 # Note: We try to keep $spice_ticket secret (do not pass via command line parameter)
733 # instead we pipe it through STDIN
734 my $exitcode = PVE
::Tools
::run_command
($cmd, input
=> $input, outfunc
=> sub {
737 if ($line =~ m/^migration listens on tcp:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+)$/) {
740 $ruri = "tcp:$raddr:$rport";
742 elsif ($line =~ m!^migration listens on unix:(/run/qemu-server/(\d+)\.migrate)$!) {
744 die "Destination UNIX sockets VMID does not match source VMID" if $vmid ne $2;
745 $ruri = "unix:$raddr";
747 elsif ($line =~ m/^migration listens on port (\d+)$/) {
748 $raddr = "localhost";
750 $ruri = "tcp:$raddr:$rport";
752 elsif ($line =~ m/^spice listens on port (\d+)$/) {
753 $spice_port = int($1);
755 elsif ($line =~ m/^storage migration listens on nbd:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+):exportname=(\S+) volume:(\S+)$/) {
757 my $nbd_uri = "nbd:$1:$2:exportname=$3";
758 my $targetdrive = $3;
759 $targetdrive =~ s/drive-//g;
761 $self->{stopnbd
} = 1;
762 $self->{target_drive
}->{$targetdrive}->{drivestr
} = $drivestr;
763 $self->{target_drive
}->{$targetdrive}->{nbd_uri
} = $nbd_uri;
764 } elsif ($line =~ m!^storage migration listens on nbd:unix:(/run/qemu-server/(\d+)_nbd\.migrate):exportname=(\S+) volume:(\S+)$!) {
766 die "Destination UNIX socket's VMID does not match source VMID" if $vmid ne $2;
767 my $nbd_unix_addr = $1;
768 my $nbd_uri = "nbd:unix:$nbd_unix_addr:exportname=$3";
769 my $targetdrive = $3;
770 $targetdrive =~ s/drive-//g;
772 $self->{stopnbd
} = 1;
773 $self->{target_drive
}->{$targetdrive}->{drivestr
} = $drivestr;
774 $self->{target_drive
}->{$targetdrive}->{nbd_uri
} = $nbd_uri;
775 $unix_socket_info->{$nbd_unix_addr} = 1;
776 } elsif ($line =~ m/^re-using replicated volume: (\S+) - (.*)$/) {
779 $target_replicated_volumes->{$volid} = $drive;
780 } elsif ($line =~ m/^QEMU: (.*)$/) {
781 $self->log('info', "[$self->{node}] $1\n");
785 $self->log('info', "[$self->{node}] $line");
788 die "remote command failed with exit code $exitcode\n" if $exitcode;
790 die "unable to detect remote migration address\n" if !$raddr;
792 if (scalar(keys %$target_replicated_volumes) != $number_of_online_replicated_volumes) {
793 die "number of replicated disks on source and target node do not match - target node too old?\n"
796 $self->log('info', "start remote tunnel");
798 if ($migration_type eq 'secure') {
800 if ($ruri =~ /^unix:/) {
801 my $ssh_forward_info = ["$raddr:$raddr"];
802 $unix_socket_info->{$raddr} = 1;
804 my $unix_sockets = [ keys %$unix_socket_info ];
805 for my $sock (@$unix_sockets) {
806 push @$ssh_forward_info, "$sock:$sock";
810 $self->{tunnel
} = $self->fork_tunnel($ssh_forward_info);
812 my $unix_socket_try = 0; # wait for the socket to become ready
813 while ($unix_socket_try <= 100) {
816 foreach my $sock (@$unix_sockets) {
822 if ($available == @$unix_sockets) {
828 if ($unix_socket_try > 100) {
830 $self->finish_tunnel($self->{tunnel
});
831 die "Timeout, migration socket $ruri did not get ready";
833 $self->{tunnel
}->{unix_sockets
} = $unix_sockets if (@$unix_sockets);
835 } elsif ($ruri =~ /^tcp:/) {
836 my $ssh_forward_info = [];
837 if ($raddr eq "localhost") {
838 # for backwards compatibility with older qemu-server versions
839 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
840 my $lport = PVE
::Tools
::next_migrate_port
($pfamily);
841 push @$ssh_forward_info, "$lport:localhost:$rport";
844 $self->{tunnel
} = $self->fork_tunnel($ssh_forward_info);
847 die "unsupported protocol in migration URI: $ruri\n";
850 #fork tunnel for insecure migration, to send faster commands like resume
851 $self->{tunnel
} = $self->fork_tunnel();
855 my $opt_bwlimit = $self->{opts
}->{bwlimit
};
857 if (defined($self->{online_local_volumes
})) {
858 $self->{storage_migration
} = 1;
859 $self->{storage_migration_jobs
} = {};
860 $self->log('info', "starting storage migration");
862 die "The number of local disks does not match between the source and the destination.\n"
863 if (scalar(keys %{$self->{target_drive
}}) != scalar @{$self->{online_local_volumes
}});
864 foreach my $drive (keys %{$self->{target_drive
}}){
865 my $target = $self->{target_drive
}->{$drive};
866 my $nbd_uri = $target->{nbd_uri
};
868 my $source_drive = PVE
::QemuServer
::parse_drive
($drive, $conf->{$drive});
869 my $target_drive = PVE
::QemuServer
::parse_drive
($drive, $target->{drivestr
});
871 my $source_volid = $source_drive->{file
};
872 my $target_volid = $target_drive->{file
};
874 my $source_sid = PVE
::Storage
::Plugin
::parse_volume_id
($source_volid);
875 my $target_sid = PVE
::Storage
::Plugin
::parse_volume_id
($target_volid);
877 my $bwlimit = PVE
::Storage
::get_bandwidth_limit
('migration', [$source_sid, $target_sid], $opt_bwlimit);
878 my $bitmap = $target->{bitmap
};
880 $self->log('info', "$drive: start migration to $nbd_uri");
881 PVE
::QemuServer
::qemu_drive_mirror
($vmid, $drive, $nbd_uri, $vmid, undef, $self->{storage_migration_jobs
}, 'skip', undef, $bwlimit, $bitmap);
883 $self->{volume_map
}->{$source_volid} = $target_volid;
884 $self->log('info', "volume '$source_volid' is '$target_volid' on the target\n");
888 $self->log('info', "starting online/live migration on $ruri");
889 $self->{livemigration
} = 1;
892 my $defaults = PVE
::QemuServer
::load_defaults
();
894 $self->log('info', "set migration_caps");
896 PVE
::QemuServer
::set_migration_caps
($vmid);
900 my $qemu_migrate_params = {};
902 # migrate speed can be set via bwlimit (datacenter.cfg and API) and via the
903 # migrate_speed parameter in qm.conf - take the lower of the two.
904 my $bwlimit = PVE
::Storage
::get_bandwidth_limit
('migration', undef, $opt_bwlimit) // 0;
905 my $migrate_speed = $conf->{migrate_speed
} // $bwlimit;
906 # migrate_speed is in MB/s, bwlimit in KB/s
907 $migrate_speed *= 1024;
909 $migrate_speed = ($bwlimit < $migrate_speed) ?
$bwlimit : $migrate_speed;
911 # always set migrate speed (overwrite kvm default of 32m) we set a very high
912 # default of 8192m which is basically unlimited
913 $migrate_speed ||= ($defaults->{migrate_speed
} || 8192) * 1024;
915 # qmp takes migrate_speed in B/s.
916 $migrate_speed *= 1024;
917 $self->log('info', "migration speed limit: $migrate_speed B/s");
918 $qemu_migrate_params->{'max-bandwidth'} = int($migrate_speed);
920 my $migrate_downtime = $defaults->{migrate_downtime
};
921 $migrate_downtime = $conf->{migrate_downtime
} if defined($conf->{migrate_downtime
});
922 # migrate-set-parameters expects limit in ms
923 $migrate_downtime *= 1000;
924 $self->log('info', "migration downtime limit: $migrate_downtime ms");
925 $qemu_migrate_params->{'downtime-limit'} = int($migrate_downtime);
927 # set cachesize to 10% of the total memory
928 my $memory = $conf->{memory
} || $defaults->{memory
};
929 my $cachesize = int($memory * 1048576 / 10);
930 $cachesize = round_powerof2
($cachesize);
932 $self->log('info', "migration cachesize: $cachesize B");
933 $qemu_migrate_params->{'xbzrle-cache-size'} = int($cachesize);
935 $self->log('info', "set migration parameters");
937 mon_cmd
($vmid, "migrate-set-parameters", %{$qemu_migrate_params});
939 $self->log('info', "migrate-set-parameters error: $@") if $@;
941 if (PVE
::QemuServer
::vga_conf_has_spice
($conf->{vga
})) {
942 my $rpcenv = PVE
::RPCEnvironment
::get
();
943 my $authuser = $rpcenv->get_user();
945 my (undef, $proxyticket) = PVE
::AccessControl
::assemble_spice_ticket
($authuser, $vmid, $self->{node
});
947 my $filename = "/etc/pve/nodes/$self->{node}/pve-ssl.pem";
948 my $subject = PVE
::AccessControl
::read_x509_subject_spice
($filename);
950 $self->log('info', "spice client_migrate_info");
953 mon_cmd
($vmid, "client_migrate_info", protocol
=> 'spice',
954 hostname
=> $proxyticket, 'port' => 0, 'tls-port' => $spice_port,
955 'cert-subject' => $subject);
957 $self->log('info', "client_migrate_info error: $@") if $@;
961 $self->log('info', "start migrate command to $ruri");
963 mon_cmd
($vmid, "migrate", uri
=> $ruri);
966 $self->log('info', "migrate uri => $ruri failed: $merr") if $merr;
969 my $usleep = 1000000;
973 my $downtimecounter = 0;
976 my $avglstat = $lstat/$i if $lstat;
981 $stat = mon_cmd
($vmid, "query-migrate");
985 warn "query migrate failed: $err\n";
986 $self->log('info', "query migrate failed: $err");
987 if ($err_count <= 5) {
991 die "too many query migrate failures - aborting\n";
994 if (defined($stat->{status
}) && $stat->{status
} =~ m/^(setup)$/im) {
999 if (defined($stat->{status
}) && $stat->{status
} =~ m/^(active|completed|failed|cancelled)$/im) {
1002 if ($stat->{status
} eq 'completed') {
1003 my $delay = time() - $start;
1005 my $mbps = sprintf "%.2f", $memory / $delay;
1006 my $downtime = $stat->{downtime
} || 0;
1007 $self->log('info', "migration speed: $mbps MB/s - downtime $downtime ms");
1011 if ($stat->{status
} eq 'failed' || $stat->{status
} eq 'cancelled') {
1012 $self->log('info', "migration status error: $stat->{status}");
1016 if ($stat->{status
} ne 'active') {
1017 $self->log('info', "migration status: $stat->{status}");
1021 if ($stat->{ram
}->{transferred
} ne $lstat) {
1022 my $trans = $stat->{ram
}->{transferred
} || 0;
1023 my $rem = $stat->{ram
}->{remaining
} || 0;
1024 my $total = $stat->{ram
}->{total
} || 0;
1025 my $xbzrlecachesize = $stat->{"xbzrle-cache"}->{"cache-size"} || 0;
1026 my $xbzrlebytes = $stat->{"xbzrle-cache"}->{"bytes"} || 0;
1027 my $xbzrlepages = $stat->{"xbzrle-cache"}->{"pages"} || 0;
1028 my $xbzrlecachemiss = $stat->{"xbzrle-cache"}->{"cache-miss"} || 0;
1029 my $xbzrleoverflow = $stat->{"xbzrle-cache"}->{"overflow"} || 0;
1030 # reduce sleep if remainig memory is lower than the average transfer speed
1031 $usleep = 100000 if $avglstat && $rem < $avglstat;
1033 $self->log('info', "migration status: $stat->{status} (transferred ${trans}, " .
1034 "remaining ${rem}), total ${total})");
1036 if (${xbzrlecachesize
}) {
1037 $self->log('info', "migration xbzrle cachesize: ${xbzrlecachesize} transferred ${xbzrlebytes} pages ${xbzrlepages} cachemiss ${xbzrlecachemiss} overflow ${xbzrleoverflow}");
1040 if (($lastrem && $rem > $lastrem ) || ($rem == 0)) {
1045 if ($downtimecounter > 5) {
1046 $downtimecounter = 0;
1047 $migrate_downtime *= 2;
1048 $self->log('info', "auto-increased downtime to continue migration: $migrate_downtime ms");
1050 # migrate-set-parameters does not touch values not
1051 # specified, so this only changes downtime-limit
1052 mon_cmd
($vmid, "migrate-set-parameters", 'downtime-limit' => int($migrate_downtime));
1054 $self->log('info', "migrate-set-parameters error: $@") if $@;
1060 $lstat = $stat->{ram
}->{transferred
};
1064 die "unable to parse migration status '$stat->{status}' - aborting\n";
1069 sub phase2_cleanup
{
1070 my ($self, $vmid, $err) = @_;
1072 return if !$self->{errors
};
1073 $self->{phase2errors
} = 1;
1075 $self->log('info', "aborting phase 2 - cleanup resources");
1077 $self->log('info', "migrate_cancel");
1079 mon_cmd
($vmid, "migrate_cancel");
1081 $self->log('info', "migrate_cancel error: $@") if $@;
1083 my $conf = $self->{vmconf
};
1084 delete $conf->{lock};
1085 eval { PVE
::QemuConfig-
>write_config($vmid, $conf) };
1087 $self->log('err', $err);
1090 # cleanup ressources on target host
1091 if ($self->{storage_migration
}) {
1092 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $self->{storage_migration_jobs
}) };
1094 $self->log('err', $err);
1098 eval { $self->cleanup_bitmaps() };
1100 $self->log('err', $err);
1103 my $nodename = PVE
::INotify
::nodename
();
1105 my $cmd = [@{$self->{rem_ssh
}}, 'qm', 'stop', $vmid, '--skiplock', '--migratedfrom', $nodename];
1106 eval{ PVE
::Tools
::run_command
($cmd, outfunc
=> sub {}, errfunc
=> sub {}) };
1108 $self->log('err', $err);
1109 $self->{errors
} = 1;
1112 # cleanup after stopping, otherwise disks might be in-use by target VM!
1113 eval { PVE
::QemuMigrate
::cleanup_remotedisks
($self) };
1115 $self->log('err', $err);
1119 if ($self->{tunnel
}) {
1120 eval { finish_tunnel
($self, $self->{tunnel
}); };
1122 $self->log('err', $err);
1123 $self->{errors
} = 1;
1129 my ($self, $vmid) = @_;
1131 my $volids = $self->{volumes
};
1132 return if $self->{phase2errors
};
1134 # destroy local copies
1135 foreach my $volid (@$volids) {
1136 eval { PVE
::Storage
::vdisk_free
($self->{storecfg
}, $volid); };
1138 $self->log('err', "removing local copy of '$volid' failed - $err");
1139 $self->{errors
} = 1;
1140 last if $err =~ /^interrupted by signal$/;
1145 sub phase3_cleanup
{
1146 my ($self, $vmid, $err) = @_;
1148 my $conf = $self->{vmconf
};
1149 return if $self->{phase2errors
};
1151 my $tunnel = $self->{tunnel
};
1153 if ($self->{storage_migration
}) {
1154 # finish block-job with block-job-cancel, to disconnect source VM from NBD
1155 # to avoid it trying to re-establish it. We are in blockjob ready state,
1156 # thus, this command changes to it to blockjob complete (see qapi docs)
1157 eval { PVE
::QemuServer
::qemu_drive_mirror_monitor
($vmid, undef, $self->{storage_migration_jobs
}, 'cancel'); };
1160 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $self->{storage_migration_jobs
}) };
1161 eval { PVE
::QemuMigrate
::cleanup_remotedisks
($self) };
1162 die "Failed to complete storage migration: $err\n";
1166 if ($self->{volume_map
}) {
1167 my $target_drives = $self->{target_drive
};
1169 # FIXME: for NBD storage migration we now only update the volid, and
1170 # not the full drivestr from the target node. Workaround that until we
1171 # got some real rescan, to avoid things like wrong format in the drive
1172 delete $conf->{$_} for keys %$target_drives;
1173 PVE
::QemuConfig-
>update_volume_ids($conf, $self->{volume_map
});
1175 for my $drive (keys %$target_drives) {
1176 $conf->{$drive} = $target_drives->{$drive}->{drivestr
};
1178 PVE
::QemuConfig-
>write_config($vmid, $conf);
1181 # transfer replication state before move config
1182 $self->transfer_replication_state() if $self->{replicated_volumes
};
1184 # move config to remote node
1185 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
1186 my $newconffile = PVE
::QemuConfig-
>config_file($vmid, $self->{node
});
1188 die "Failed to move config to node '$self->{node}' - rename failed: $!\n"
1189 if !rename($conffile, $newconffile);
1191 $self->switch_replication_job_target() if $self->{replicated_volumes
};
1193 if ($self->{livemigration
}) {
1194 if ($self->{stopnbd
}) {
1195 $self->log('info', "stopping NBD storage migration server on target.");
1196 # stop nbd server on remote vm - requirement for resume since 2.9
1197 my $cmd = [@{$self->{rem_ssh
}}, 'qm', 'nbdstop', $vmid];
1199 eval{ PVE
::Tools
::run_command
($cmd, outfunc
=> sub {}, errfunc
=> sub {}) };
1201 $self->log('err', $err);
1202 $self->{errors
} = 1;
1206 # config moved and nbd server stopped - now we can resume vm on target
1207 if ($tunnel && $tunnel->{version
} && $tunnel->{version
} >= 1) {
1209 $self->write_tunnel($tunnel, 30, "resume $vmid");
1212 $self->log('err', $err);
1213 $self->{errors
} = 1;
1216 my $cmd = [@{$self->{rem_ssh
}}, 'qm', 'resume', $vmid, '--skiplock', '--nocheck'];
1219 $self->log('err', $line);
1221 eval { PVE
::Tools
::run_command
($cmd, outfunc
=> sub {}, errfunc
=> $logf); };
1223 $self->log('err', $err);
1224 $self->{errors
} = 1;
1228 if ($self->{storage_migration
} && PVE
::QemuServer
::parse_guest_agent
($conf)->{fstrim_cloned_disks
} && $self->{running
}) {
1229 my $cmd = [@{$self->{rem_ssh
}}, 'qm', 'guest', 'cmd', $vmid, 'fstrim'];
1230 eval{ PVE
::Tools
::run_command
($cmd, outfunc
=> sub {}, errfunc
=> sub {}) };
1234 # close tunnel on successful migration, on error phase2_cleanup closed it
1236 eval { finish_tunnel
($self, $tunnel); };
1238 $self->log('err', $err);
1239 $self->{errors
} = 1;
1245 if (PVE
::QemuServer
::vga_conf_has_spice
($conf->{vga
}) && $self->{running
}) {
1246 $self->log('info', "Waiting for spice server migration");
1248 my $res = mon_cmd
($vmid, 'query-spice');
1249 last if int($res->{'migrated'}) == 1;
1250 last if $timer > 50;
1257 # always stop local VM
1258 eval { PVE
::QemuServer
::vm_stop
($self->{storecfg
}, $vmid, 1, 1); };
1260 $self->log('err', "stopping vm failed - $err");
1261 $self->{errors
} = 1;
1264 # always deactivate volumes - avoid lvm LVs to be active on several nodes
1266 my $vollist = PVE
::QemuServer
::get_vm_volumes
($conf);
1267 PVE
::Storage
::deactivate_volumes
($self->{storecfg
}, $vollist);
1270 $self->log('err', $err);
1271 $self->{errors
} = 1;
1274 if($self->{storage_migration
}) {
1275 # destroy local copies
1276 my $volids = $self->{online_local_volumes
};
1278 foreach my $volid (@$volids) {
1279 # keep replicated volumes!
1280 next if $self->{replicated_volumes
}->{$volid};
1282 eval { PVE
::Storage
::vdisk_free
($self->{storecfg
}, $volid); };
1284 $self->log('err', "removing local copy of '$volid' failed - $err");
1285 $self->{errors
} = 1;
1286 last if $err =~ /^interrupted by signal$/;
1292 # clear migrate lock
1293 my $cmd = [ @{$self->{rem_ssh
}}, 'qm', 'unlock', $vmid ];
1294 $self->cmd_logerr($cmd, errmsg
=> "failed to clear migrate lock");
1298 my ($self, $vmid) = @_;
1303 sub round_powerof2
{
1304 return 1 if $_[0] < 2;
1305 return 2 << int(log($_[0]-1)/log(2));