]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuMigrate.pm
fix #2100: ignore cloudinit drive on offline migration
[qemu-server.git] / PVE / QemuMigrate.pm
1 package PVE::QemuMigrate;
2
3 use strict;
4 use warnings;
5 use PVE::AbstractMigrate;
6 use IO::File;
7 use IPC::Open2;
8 use POSIX qw( WNOHANG );
9 use PVE::INotify;
10 use PVE::Tools;
11 use PVE::Cluster;
12 use PVE::Storage;
13 use PVE::QemuServer;
14 use Time::HiRes qw( usleep );
15 use PVE::RPCEnvironment;
16 use PVE::ReplicationConfig;
17 use PVE::ReplicationState;
18 use PVE::Replication;
19
20 use base qw(PVE::AbstractMigrate);
21
22 sub fork_command_pipe {
23 my ($self, $cmd) = @_;
24
25 my $reader = IO::File->new();
26 my $writer = IO::File->new();
27
28 my $orig_pid = $$;
29
30 my $cpid;
31
32 eval { $cpid = open2($reader, $writer, @$cmd); };
33
34 my $err = $@;
35
36 # catch exec errors
37 if ($orig_pid != $$) {
38 $self->log('err', "can't fork command pipe\n");
39 POSIX::_exit(1);
40 kill('KILL', $$);
41 }
42
43 die $err if $err;
44
45 return { writer => $writer, reader => $reader, pid => $cpid };
46 }
47
48 sub finish_command_pipe {
49 my ($self, $cmdpipe, $timeout) = @_;
50
51 my $cpid = $cmdpipe->{pid};
52 return if !defined($cpid);
53
54 my $writer = $cmdpipe->{writer};
55 my $reader = $cmdpipe->{reader};
56
57 $writer->close();
58 $reader->close();
59
60 my $collect_child_process = sub {
61 my $res = waitpid($cpid, WNOHANG);
62 if (defined($res) && ($res == $cpid)) {
63 delete $cmdpipe->{cpid};
64 return 1;
65 } else {
66 return 0;
67 }
68 };
69
70 if ($timeout) {
71 for (my $i = 0; $i < $timeout; $i++) {
72 return if &$collect_child_process();
73 sleep(1);
74 }
75 }
76
77 $self->log('info', "ssh tunnel still running - terminating now with SIGTERM\n");
78 kill(15, $cpid);
79
80 # wait again
81 for (my $i = 0; $i < 10; $i++) {
82 return if &$collect_child_process();
83 sleep(1);
84 }
85
86 $self->log('info', "ssh tunnel still running - terminating now with SIGKILL\n");
87 kill 9, $cpid;
88 sleep 1;
89
90 $self->log('err', "ssh tunnel child process (PID $cpid) couldn't be collected\n")
91 if !&$collect_child_process();
92 }
93
94 sub read_tunnel {
95 my ($self, $tunnel, $timeout) = @_;
96
97 $timeout = 60 if !defined($timeout);
98
99 my $reader = $tunnel->{reader};
100
101 my $output;
102 eval {
103 PVE::Tools::run_with_timeout($timeout, sub { $output = <$reader>; });
104 };
105 die "reading from tunnel failed: $@\n" if $@;
106
107 chomp $output;
108
109 return $output;
110 }
111
112 sub write_tunnel {
113 my ($self, $tunnel, $timeout, $command) = @_;
114
115 $timeout = 60 if !defined($timeout);
116
117 my $writer = $tunnel->{writer};
118
119 eval {
120 PVE::Tools::run_with_timeout($timeout, sub {
121 print $writer "$command\n";
122 $writer->flush();
123 });
124 };
125 die "writing to tunnel failed: $@\n" if $@;
126
127 if ($tunnel->{version} && $tunnel->{version} >= 1) {
128 my $res = eval { $self->read_tunnel($tunnel, 10); };
129 die "no reply to command '$command': $@\n" if $@;
130
131 if ($res eq 'OK') {
132 return;
133 } else {
134 die "tunnel replied '$res' to command '$command'\n";
135 }
136 }
137 }
138
139 sub fork_tunnel {
140 my ($self, $tunnel_addr) = @_;
141
142 my @localtunnelinfo = defined($tunnel_addr) ? ('-L' , $tunnel_addr ) : ();
143
144 my $cmd = [@{$self->{rem_ssh}}, '-o ExitOnForwardFailure=yes', @localtunnelinfo, '/usr/sbin/qm', 'mtunnel' ];
145
146 my $tunnel = $self->fork_command_pipe($cmd);
147
148 eval {
149 my $helo = $self->read_tunnel($tunnel, 60);
150 die "no reply\n" if !$helo;
151 die "no quorum on target node\n" if $helo =~ m/^no quorum$/;
152 die "got strange reply from mtunnel ('$helo')\n"
153 if $helo !~ m/^tunnel online$/;
154 };
155 my $err = $@;
156
157 eval {
158 my $ver = $self->read_tunnel($tunnel, 10);
159 if ($ver =~ /^ver (\d+)$/) {
160 $tunnel->{version} = $1;
161 $self->log('info', "ssh tunnel $ver\n");
162 } else {
163 $err = "received invalid tunnel version string '$ver'\n" if !$err;
164 }
165 };
166
167 if ($err) {
168 $self->finish_command_pipe($tunnel);
169 die "can't open migration tunnel - $err";
170 }
171 return $tunnel;
172 }
173
174 sub finish_tunnel {
175 my ($self, $tunnel) = @_;
176
177 eval { $self->write_tunnel($tunnel, 30, 'quit'); };
178 my $err = $@;
179
180 $self->finish_command_pipe($tunnel, 30);
181
182 if ($tunnel->{sock_addr}) {
183 # ssh does not clean up on local host
184 my $cmd = ['rm', '-f', $tunnel->{sock_addr}]; #
185 PVE::Tools::run_command($cmd);
186
187 # .. and just to be sure check on remote side
188 unshift @{$cmd}, @{$self->{rem_ssh}};
189 PVE::Tools::run_command($cmd);
190 }
191
192 die $err if $err;
193 }
194
195 sub lock_vm {
196 my ($self, $vmid, $code, @param) = @_;
197
198 return PVE::QemuConfig->lock_config($vmid, $code, @param);
199 }
200
201 sub prepare {
202 my ($self, $vmid) = @_;
203
204 my $online = $self->{opts}->{online};
205
206 $self->{storecfg} = PVE::Storage::config();
207
208 # test if VM exists
209 my $conf = $self->{vmconf} = PVE::QemuConfig->load_config($vmid);
210
211 PVE::QemuConfig->check_lock($conf);
212
213 my $running = 0;
214 if (my $pid = PVE::QemuServer::check_running($vmid)) {
215 die "can't migrate running VM without --online\n" if !$online;
216 $running = $pid;
217
218 $self->{forcemachine} = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
219
220 }
221
222 if (my $loc_res = PVE::QemuServer::check_local_resources($conf, 1)) {
223 if ($self->{running} || !$self->{opts}->{force}) {
224 die "can't migrate VM which uses local devices\n";
225 } else {
226 $self->log('info', "migrating VM which uses local devices");
227 }
228 }
229
230 my $vollist = PVE::QemuServer::get_vm_volumes($conf);
231
232 my $need_activate = [];
233 foreach my $volid (@$vollist) {
234 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
235
236 # check if storage is available on both nodes
237 my $targetsid = $self->{opts}->{targetstorage} // $sid;
238
239 my $scfg = PVE::Storage::storage_check_node($self->{storecfg}, $sid);
240 PVE::Storage::storage_check_node($self->{storecfg}, $targetsid, $self->{node});
241
242 if ($scfg->{shared}) {
243 # PVE::Storage::activate_storage checks this for non-shared storages
244 my $plugin = PVE::Storage::Plugin->lookup($scfg->{type});
245 warn "Used shared storage '$sid' is not online on source node!\n"
246 if !$plugin->check_connection($sid, $scfg);
247 } else {
248 # only activate if not shared
249 next if ($volid =~ m/vm-\d+-cloudinit/);
250 push @$need_activate, $volid;
251 }
252 }
253
254 # activate volumes
255 PVE::Storage::activate_volumes($self->{storecfg}, $need_activate);
256
257 # test ssh connection
258 my $cmd = [ @{$self->{rem_ssh}}, '/bin/true' ];
259 eval { $self->cmd_quiet($cmd); };
260 die "Can't connect to destination address using public key\n" if $@;
261
262 return $running;
263 }
264
265 sub sync_disks {
266 my ($self, $vmid) = @_;
267
268 my $conf = $self->{vmconf};
269
270 # local volumes which have been copied
271 $self->{volumes} = [];
272
273 my $override_targetsid = $self->{opts}->{targetstorage};
274
275 eval {
276
277 # found local volumes and their origin
278 my $local_volumes = {};
279 my $local_volumes_errors = {};
280 my $other_errors = [];
281 my $abort = 0;
282
283 my $sharedvm = 1;
284
285 my $log_error = sub {
286 my ($msg, $volid) = @_;
287
288 if (defined($volid)) {
289 $local_volumes_errors->{$volid} = $msg;
290 } else {
291 push @$other_errors, $msg;
292 }
293 $abort = 1;
294 };
295
296 my @sids = PVE::Storage::storage_ids($self->{storecfg});
297 foreach my $storeid (@sids) {
298 my $scfg = PVE::Storage::storage_config($self->{storecfg}, $storeid);
299 next if $scfg->{shared};
300 next if !PVE::Storage::storage_check_enabled($self->{storecfg}, $storeid, undef, 1);
301
302 # get list from PVE::Storage (for unused volumes)
303 my $dl = PVE::Storage::vdisk_list($self->{storecfg}, $storeid, $vmid);
304
305 next if @{$dl->{$storeid}} == 0;
306
307 my $targetsid = $override_targetsid // $storeid;
308
309 # check if storage is available on target node
310 PVE::Storage::storage_check_node($self->{storecfg}, $targetsid, $self->{node});
311 $sharedvm = 0; # there is a non-shared disk
312
313 PVE::Storage::foreach_volid($dl, sub {
314 my ($volid, $sid, $volname) = @_;
315
316 $local_volumes->{$volid}->{ref} = 'storage';
317 });
318 }
319
320 my $test_volid = sub {
321 my ($volid, $attr) = @_;
322
323 if ($volid =~ m|^/|) {
324 return if $attr->{shared};
325 $local_volumes->{$volid}->{ref} = 'config';
326 die "local file/device\n";
327 }
328
329 my $snaprefs = $attr->{referenced_in_snapshot};
330
331 if ($attr->{cdrom}) {
332 if ($volid eq 'cdrom') {
333 my $msg = "can't migrate local cdrom drive";
334 if (defined($snaprefs) && !$attr->{referenced_in_config}) {
335 my $snapnames = join(', ', sort keys %$snaprefs);
336 $msg .= " (referenced in snapshot - $snapnames)";
337 }
338 &$log_error("$msg\n");
339 return;
340 }
341 return if $volid eq 'none';
342 }
343
344 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
345
346 my $targetsid = $override_targetsid // $sid;
347 # check if storage is available on both nodes
348 my $scfg = PVE::Storage::storage_check_node($self->{storecfg}, $sid);
349 PVE::Storage::storage_check_node($self->{storecfg}, $targetsid, $self->{node});
350
351 return if $scfg->{shared};
352
353 $sharedvm = 0;
354
355 $local_volumes->{$volid}->{ref} = $attr->{referenced_in_config} ? 'config' : 'snapshot';
356
357 if ($attr->{cdrom}) {
358 if ($volid =~ /vm-\d+-cloudinit/) {
359 $local_volumes->{$volid}->{ref} = 'generated';
360 return;
361 }
362 die "local cdrom image\n";
363 }
364
365 my ($path, $owner) = PVE::Storage::path($self->{storecfg}, $volid);
366
367 die "owned by other VM (owner = VM $owner)\n"
368 if !$owner || ($owner != $self->{vmid});
369
370 my $format = PVE::QemuServer::qemu_img_format($scfg, $volname);
371 $local_volumes->{$volid}->{snapshots} = defined($snaprefs) || ($format =~ /^(?:qcow2|vmdk)$/);
372 if (defined($snaprefs)) {
373 # we cannot migrate shapshots on local storage
374 # exceptions: 'zfspool' or 'qcow2' files (on directory storage)
375
376 die "online storage migration not possible if snapshot exists\n" if $self->{running};
377 if (!($scfg->{type} eq 'zfspool' || $format eq 'qcow2')) {
378 die "non-migratable snapshot exists\n";
379 }
380 }
381
382 die "referenced by linked clone(s)\n"
383 if PVE::Storage::volume_is_base_and_used($self->{storecfg}, $volid);
384 };
385
386 PVE::QemuServer::foreach_volid($conf, sub {
387 my ($volid, $attr) = @_;
388 eval { $test_volid->($volid, $attr); };
389 if (my $err = $@) {
390 &$log_error($err, $volid);
391 }
392 });
393
394 foreach my $vol (sort keys %$local_volumes) {
395 my $ref = $local_volumes->{$vol}->{ref};
396 if ($ref eq 'storage') {
397 $self->log('info', "found local disk '$vol' (via storage)\n");
398 } elsif ($ref eq 'config') {
399 &$log_error("can't live migrate attached local disks without with-local-disks option\n", $vol)
400 if $self->{running} && !$self->{opts}->{"with-local-disks"};
401 $self->log('info', "found local disk '$vol' (in current VM config)\n");
402 } elsif ($ref eq 'snapshot') {
403 $self->log('info', "found local disk '$vol' (referenced by snapshot(s))\n");
404 } elsif ($ref eq 'generated') {
405 $self->log('info', "found generated disk '$vol' (in current VM config)\n");
406 } else {
407 $self->log('info', "found local disk '$vol'\n");
408 }
409 }
410
411 foreach my $vol (sort keys %$local_volumes_errors) {
412 $self->log('warn', "can't migrate local disk '$vol': $local_volumes_errors->{$vol}");
413 }
414 foreach my $err (@$other_errors) {
415 $self->log('warn', "$err");
416 }
417
418 if ($abort) {
419 die "can't migrate VM - check log\n";
420 }
421
422 # additional checks for local storage
423 foreach my $volid (keys %$local_volumes) {
424 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
425 my $scfg = PVE::Storage::storage_config($self->{storecfg}, $sid);
426
427 my $migratable = ($scfg->{type} eq 'dir') || ($scfg->{type} eq 'zfspool') ||
428 ($scfg->{type} eq 'lvmthin') || ($scfg->{type} eq 'lvm');
429
430 die "can't migrate '$volid' - storage type '$scfg->{type}' not supported\n"
431 if !$migratable;
432
433 # image is a linked clone on local storage, se we can't migrate.
434 if (my $basename = (PVE::Storage::parse_volname($self->{storecfg}, $volid))[3]) {
435 die "can't migrate '$volid' as it's a clone of '$basename'";
436 }
437 }
438
439 my $rep_volumes;
440
441 $self->log('info', "copying disk images");
442
443 my $rep_cfg = PVE::ReplicationConfig->new();
444
445 if (my $jobcfg = $rep_cfg->find_local_replication_job($vmid, $self->{node})) {
446 die "can't live migrate VM with replicated volumes\n" if $self->{running};
447 my $start_time = time();
448 my $logfunc = sub { my ($msg) = @_; $self->log('info', $msg); };
449 $rep_volumes = PVE::Replication::run_replication(
450 'PVE::QemuConfig', $jobcfg, $start_time, $start_time, $logfunc);
451 $self->{replicated_volumes} = $rep_volumes;
452 }
453
454 foreach my $volid (keys %$local_volumes) {
455 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
456 my $targetsid = $override_targetsid // $sid;
457 my $ref = $local_volumes->{$volid}->{ref};
458 if ($self->{running} && $ref eq 'config') {
459 push @{$self->{online_local_volumes}}, $volid;
460 } elsif ($ref eq 'generated') {
461 # skip all generated volumes but queue them for deletion in phase3_cleanup
462 push @{$self->{volumes}}, $volid;
463 next;
464 } else {
465 next if $rep_volumes->{$volid};
466 push @{$self->{volumes}}, $volid;
467 my $insecure = $self->{opts}->{migration_type} eq 'insecure';
468 my $with_snapshots = $local_volumes->{$volid}->{snapshots};
469 PVE::Storage::storage_migrate($self->{storecfg}, $volid, $self->{ssh_info}, $targetsid,
470 undef, undef, undef, undef, $insecure, $with_snapshots);
471 }
472 }
473 };
474 die "Failed to sync data - $@" if $@;
475 }
476
477 sub cleanup_remotedisks {
478 my ($self) = @_;
479
480 foreach my $target_drive (keys %{$self->{target_drive}}) {
481
482 my $drive = PVE::QemuServer::parse_drive($target_drive, $self->{target_drive}->{$target_drive}->{volid});
483 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
484
485 my $cmd = [@{$self->{rem_ssh}}, 'pvesm', 'free', "$storeid:$volname"];
486
487 eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
488 if (my $err = $@) {
489 $self->log('err', $err);
490 $self->{errors} = 1;
491 }
492 }
493 }
494
495 sub phase1 {
496 my ($self, $vmid) = @_;
497
498 $self->log('info', "starting migration of VM $vmid to node '$self->{node}' ($self->{nodeip})");
499
500 my $conf = $self->{vmconf};
501
502 # set migrate lock in config file
503 $conf->{lock} = 'migrate';
504 PVE::QemuConfig->write_config($vmid, $conf);
505
506 sync_disks($self, $vmid);
507
508 };
509
510 sub phase1_cleanup {
511 my ($self, $vmid, $err) = @_;
512
513 $self->log('info', "aborting phase 1 - cleanup resources");
514
515 my $conf = $self->{vmconf};
516 delete $conf->{lock};
517 eval { PVE::QemuConfig->write_config($vmid, $conf) };
518 if (my $err = $@) {
519 $self->log('err', $err);
520 }
521
522 if ($self->{volumes}) {
523 foreach my $volid (@{$self->{volumes}}) {
524 $self->log('err', "found stale volume copy '$volid' on node '$self->{node}'");
525 # fixme: try to remove ?
526 }
527 }
528 }
529
530 sub phase2 {
531 my ($self, $vmid) = @_;
532
533 my $conf = $self->{vmconf};
534
535 $self->log('info', "starting VM $vmid on remote node '$self->{node}'");
536
537 my $raddr;
538 my $rport;
539 my $ruri; # the whole migration dst. URI (protocol:address[:port])
540 my $nodename = PVE::INotify::nodename();
541
542 ## start on remote node
543 my $cmd = [@{$self->{rem_ssh}}];
544
545 my $spice_ticket;
546 if (PVE::QemuServer::vga_conf_has_spice($conf->{vga})) {
547 my $res = PVE::QemuServer::vm_mon_cmd($vmid, 'query-spice');
548 $spice_ticket = $res->{ticket};
549 }
550
551 push @$cmd , 'qm', 'start', $vmid, '--skiplock', '--migratedfrom', $nodename;
552
553 my $migration_type = $self->{opts}->{migration_type};
554
555 push @$cmd, '--migration_type', $migration_type;
556
557 push @$cmd, '--migration_network', $self->{opts}->{migration_network}
558 if $self->{opts}->{migration_network};
559
560 if ($migration_type eq 'insecure') {
561 push @$cmd, '--stateuri', 'tcp';
562 } else {
563 push @$cmd, '--stateuri', 'unix';
564 }
565
566 if ($self->{forcemachine}) {
567 push @$cmd, '--machine', $self->{forcemachine};
568 }
569
570 if ($self->{online_local_volumes}) {
571 push @$cmd, '--targetstorage', ($self->{opts}->{targetstorage} // '1');
572 }
573
574 my $spice_port;
575
576 # Note: We try to keep $spice_ticket secret (do not pass via command line parameter)
577 # instead we pipe it through STDIN
578 PVE::Tools::run_command($cmd, input => $spice_ticket, outfunc => sub {
579 my $line = shift;
580
581 if ($line =~ m/^migration listens on tcp:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+)$/) {
582 $raddr = $1;
583 $rport = int($2);
584 $ruri = "tcp:$raddr:$rport";
585 }
586 elsif ($line =~ m!^migration listens on unix:(/run/qemu-server/(\d+)\.migrate)$!) {
587 $raddr = $1;
588 die "Destination UNIX sockets VMID does not match source VMID" if $vmid ne $2;
589 $ruri = "unix:$raddr";
590 }
591 elsif ($line =~ m/^migration listens on port (\d+)$/) {
592 $raddr = "localhost";
593 $rport = int($1);
594 $ruri = "tcp:$raddr:$rport";
595 }
596 elsif ($line =~ m/^spice listens on port (\d+)$/) {
597 $spice_port = int($1);
598 }
599 elsif ($line =~ m/^storage migration listens on nbd:(localhost|[\d\.]+|\[[\d\.:a-fA-F]+\]):(\d+):exportname=(\S+) volume:(\S+)$/) {
600 my $volid = $4;
601 my $nbd_uri = "nbd:$1:$2:exportname=$3";
602 my $targetdrive = $3;
603 $targetdrive =~ s/drive-//g;
604
605 $self->{target_drive}->{$targetdrive}->{volid} = $volid;
606 $self->{target_drive}->{$targetdrive}->{nbd_uri} = $nbd_uri;
607
608 }
609 }, errfunc => sub {
610 my $line = shift;
611 $self->log('info', $line);
612 });
613
614 die "unable to detect remote migration address\n" if !$raddr;
615
616 $self->log('info', "start remote tunnel");
617
618 if ($migration_type eq 'secure') {
619
620 if ($ruri =~ /^unix:/) {
621 unlink $raddr;
622 $self->{tunnel} = $self->fork_tunnel("$raddr:$raddr");
623 $self->{tunnel}->{sock_addr} = $raddr;
624
625 my $unix_socket_try = 0; # wait for the socket to become ready
626 while (! -S $raddr) {
627 $unix_socket_try++;
628 if ($unix_socket_try > 100) {
629 $self->{errors} = 1;
630 $self->finish_tunnel($self->{tunnel});
631 die "Timeout, migration socket $ruri did not get ready";
632 }
633
634 usleep(50000);
635 }
636
637 } elsif ($ruri =~ /^tcp:/) {
638 my $tunnel_addr;
639 if ($raddr eq "localhost") {
640 # for backwards compatibility with older qemu-server versions
641 my $pfamily = PVE::Tools::get_host_address_family($nodename);
642 my $lport = PVE::Tools::next_migrate_port($pfamily);
643 $tunnel_addr = "$lport:localhost:$rport";
644 }
645
646 $self->{tunnel} = $self->fork_tunnel($tunnel_addr);
647
648 } else {
649 die "unsupported protocol in migration URI: $ruri\n";
650 }
651 } else {
652 #fork tunnel for insecure migration, to send faster commands like resume
653 $self->{tunnel} = $self->fork_tunnel();
654 }
655
656 my $start = time();
657
658 if (defined($self->{online_local_volumes})) {
659 $self->{storage_migration} = 1;
660 $self->{storage_migration_jobs} = {};
661 $self->log('info', "starting storage migration");
662
663 die "The number of local disks does not match between the source and the destination.\n"
664 if (scalar(keys %{$self->{target_drive}}) != scalar @{$self->{online_local_volumes}});
665 foreach my $drive (keys %{$self->{target_drive}}){
666 my $nbd_uri = $self->{target_drive}->{$drive}->{nbd_uri};
667 $self->log('info', "$drive: start migration to $nbd_uri");
668 PVE::QemuServer::qemu_drive_mirror($vmid, $drive, $nbd_uri, $vmid, undef, $self->{storage_migration_jobs}, 1);
669 }
670 }
671
672 $self->log('info', "starting online/live migration on $ruri");
673 $self->{livemigration} = 1;
674
675 # load_defaults
676 my $defaults = PVE::QemuServer::load_defaults();
677
678 # always set migrate speed (overwrite kvm default of 32m)
679 # we set a very hight default of 8192m which is basically unlimited
680 my $migrate_speed = $defaults->{migrate_speed} || 8192;
681 $migrate_speed = $conf->{migrate_speed} || $migrate_speed;
682 $migrate_speed = $migrate_speed * 1048576;
683 $self->log('info', "migrate_set_speed: $migrate_speed");
684 eval {
685 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate_set_speed", value => int($migrate_speed));
686 };
687 $self->log('info', "migrate_set_speed error: $@") if $@;
688
689 my $migrate_downtime = $defaults->{migrate_downtime};
690 $migrate_downtime = $conf->{migrate_downtime} if defined($conf->{migrate_downtime});
691 if (defined($migrate_downtime)) {
692 $self->log('info', "migrate_set_downtime: $migrate_downtime");
693 eval {
694 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate_set_downtime", value => int($migrate_downtime*100)/100);
695 };
696 $self->log('info', "migrate_set_downtime error: $@") if $@;
697 }
698
699 $self->log('info', "set migration_caps");
700 eval {
701 PVE::QemuServer::set_migration_caps($vmid);
702 };
703 warn $@ if $@;
704
705 # set cachesize to 10% of the total memory
706 my $memory = $conf->{memory} || $defaults->{memory};
707 my $cachesize = int($memory * 1048576 / 10);
708 $cachesize = round_powerof2($cachesize);
709
710 $self->log('info', "set cachesize: $cachesize");
711 eval {
712 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate-set-cache-size", value => int($cachesize));
713 };
714 $self->log('info', "migrate-set-cache-size error: $@") if $@;
715
716 if (PVE::QemuServer::vga_conf_has_spice($conf->{vga})) {
717 my $rpcenv = PVE::RPCEnvironment::get();
718 my $authuser = $rpcenv->get_user();
719
720 my (undef, $proxyticket) = PVE::AccessControl::assemble_spice_ticket($authuser, $vmid, $self->{node});
721
722 my $filename = "/etc/pve/nodes/$self->{node}/pve-ssl.pem";
723 my $subject = PVE::AccessControl::read_x509_subject_spice($filename);
724
725 $self->log('info', "spice client_migrate_info");
726
727 eval {
728 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "client_migrate_info", protocol => 'spice',
729 hostname => $proxyticket, 'tls-port' => $spice_port,
730 'cert-subject' => $subject);
731 };
732 $self->log('info', "client_migrate_info error: $@") if $@;
733
734 }
735
736 $self->log('info', "start migrate command to $ruri");
737 eval {
738 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate", uri => $ruri);
739 };
740 my $merr = $@;
741 $self->log('info', "migrate uri => $ruri failed: $merr") if $merr;
742
743 my $lstat = 0;
744 my $usleep = 1000000;
745 my $i = 0;
746 my $err_count = 0;
747 my $lastrem = undef;
748 my $downtimecounter = 0;
749 while (1) {
750 $i++;
751 my $avglstat = $lstat/$i if $lstat;
752
753 usleep($usleep);
754 my $stat;
755 eval {
756 $stat = PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "query-migrate");
757 };
758 if (my $err = $@) {
759 $err_count++;
760 warn "query migrate failed: $err\n";
761 $self->log('info', "query migrate failed: $err");
762 if ($err_count <= 5) {
763 usleep(1000000);
764 next;
765 }
766 die "too many query migrate failures - aborting\n";
767 }
768
769 if (defined($stat->{status}) && $stat->{status} =~ m/^(setup)$/im) {
770 sleep(1);
771 next;
772 }
773
774 if (defined($stat->{status}) && $stat->{status} =~ m/^(active|completed|failed|cancelled)$/im) {
775 $merr = undef;
776 $err_count = 0;
777 if ($stat->{status} eq 'completed') {
778 my $delay = time() - $start;
779 if ($delay > 0) {
780 my $mbps = sprintf "%.2f", $memory / $delay;
781 my $downtime = $stat->{downtime} || 0;
782 $self->log('info', "migration speed: $mbps MB/s - downtime $downtime ms");
783 }
784 }
785
786 if ($stat->{status} eq 'failed' || $stat->{status} eq 'cancelled') {
787 $self->log('info', "migration status error: $stat->{status}");
788 die "aborting\n"
789 }
790
791 if ($stat->{status} ne 'active') {
792 $self->log('info', "migration status: $stat->{status}");
793 last;
794 }
795
796 if ($stat->{ram}->{transferred} ne $lstat) {
797 my $trans = $stat->{ram}->{transferred} || 0;
798 my $rem = $stat->{ram}->{remaining} || 0;
799 my $total = $stat->{ram}->{total} || 0;
800 my $xbzrlecachesize = $stat->{"xbzrle-cache"}->{"cache-size"} || 0;
801 my $xbzrlebytes = $stat->{"xbzrle-cache"}->{"bytes"} || 0;
802 my $xbzrlepages = $stat->{"xbzrle-cache"}->{"pages"} || 0;
803 my $xbzrlecachemiss = $stat->{"xbzrle-cache"}->{"cache-miss"} || 0;
804 my $xbzrleoverflow = $stat->{"xbzrle-cache"}->{"overflow"} || 0;
805 # reduce sleep if remainig memory is lower than the average transfer speed
806 $usleep = 100000 if $avglstat && $rem < $avglstat;
807
808 $self->log('info', "migration status: $stat->{status} (transferred ${trans}, " .
809 "remaining ${rem}), total ${total})");
810
811 if (${xbzrlecachesize}) {
812 $self->log('info', "migration xbzrle cachesize: ${xbzrlecachesize} transferred ${xbzrlebytes} pages ${xbzrlepages} cachemiss ${xbzrlecachemiss} overflow ${xbzrleoverflow}");
813 }
814
815 if (($lastrem && $rem > $lastrem ) || ($rem == 0)) {
816 $downtimecounter++;
817 }
818 $lastrem = $rem;
819
820 if ($downtimecounter > 5) {
821 $downtimecounter = 0;
822 $migrate_downtime *= 2;
823 $self->log('info', "migrate_set_downtime: $migrate_downtime");
824 eval {
825 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate_set_downtime", value => int($migrate_downtime*100)/100);
826 };
827 $self->log('info', "migrate_set_downtime error: $@") if $@;
828 }
829
830 }
831
832
833 $lstat = $stat->{ram}->{transferred};
834
835 } else {
836 die $merr if $merr;
837 die "unable to parse migration status '$stat->{status}' - aborting\n";
838 }
839 }
840 }
841
842 sub phase2_cleanup {
843 my ($self, $vmid, $err) = @_;
844
845 return if !$self->{errors};
846 $self->{phase2errors} = 1;
847
848 $self->log('info', "aborting phase 2 - cleanup resources");
849
850 $self->log('info', "migrate_cancel");
851 eval {
852 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate_cancel");
853 };
854 $self->log('info', "migrate_cancel error: $@") if $@;
855
856 my $conf = $self->{vmconf};
857 delete $conf->{lock};
858 eval { PVE::QemuConfig->write_config($vmid, $conf) };
859 if (my $err = $@) {
860 $self->log('err', $err);
861 }
862
863 # cleanup ressources on target host
864 if ($self->{storage_migration}) {
865
866 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $self->{storage_migration_jobs}) };
867 if (my $err = $@) {
868 $self->log('err', $err);
869 }
870
871 eval { PVE::QemuMigrate::cleanup_remotedisks($self) };
872 if (my $err = $@) {
873 $self->log('err', $err);
874 }
875 }
876
877 my $nodename = PVE::INotify::nodename();
878
879 my $cmd = [@{$self->{rem_ssh}}, 'qm', 'stop', $vmid, '--skiplock', '--migratedfrom', $nodename];
880 eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
881 if (my $err = $@) {
882 $self->log('err', $err);
883 $self->{errors} = 1;
884 }
885
886 if ($self->{tunnel}) {
887 eval { finish_tunnel($self, $self->{tunnel}); };
888 if (my $err = $@) {
889 $self->log('err', $err);
890 $self->{errors} = 1;
891 }
892 }
893 }
894
895 sub phase3 {
896 my ($self, $vmid) = @_;
897
898 my $volids = $self->{volumes};
899 return if $self->{phase2errors};
900
901 # destroy local copies
902 foreach my $volid (@$volids) {
903 eval { PVE::Storage::vdisk_free($self->{storecfg}, $volid); };
904 if (my $err = $@) {
905 $self->log('err', "removing local copy of '$volid' failed - $err");
906 $self->{errors} = 1;
907 last if $err =~ /^interrupted by signal$/;
908 }
909 }
910 }
911
912 sub phase3_cleanup {
913 my ($self, $vmid, $err) = @_;
914
915 my $conf = $self->{vmconf};
916 return if $self->{phase2errors};
917
918 my $tunnel = $self->{tunnel};
919
920 if ($self->{storage_migration}) {
921 # finish block-job
922 eval { PVE::QemuServer::qemu_drive_mirror_monitor($vmid, undef, $self->{storage_migration_jobs}); };
923
924 if (my $err = $@) {
925 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $self->{storage_migration_jobs}) };
926 eval { PVE::QemuMigrate::cleanup_remotedisks($self) };
927 die "Failed to completed storage migration\n";
928 } else {
929 foreach my $target_drive (keys %{$self->{target_drive}}) {
930 my $drive = PVE::QemuServer::parse_drive($target_drive, $self->{target_drive}->{$target_drive}->{volid});
931 $conf->{$target_drive} = PVE::QemuServer::print_drive($vmid, $drive);
932 PVE::QemuConfig->write_config($vmid, $conf);
933 }
934 }
935 }
936
937 # transfer replication state before move config
938 $self->transfer_replication_state() if $self->{replicated_volumes};
939
940 # move config to remote node
941 my $conffile = PVE::QemuConfig->config_file($vmid);
942 my $newconffile = PVE::QemuConfig->config_file($vmid, $self->{node});
943
944 die "Failed to move config to node '$self->{node}' - rename failed: $!\n"
945 if !rename($conffile, $newconffile);
946
947 $self->switch_replication_job_target() if $self->{replicated_volumes};
948
949 if ($self->{livemigration}) {
950 if ($self->{storage_migration}) {
951 # stop nbd server on remote vm - requirement for resume since 2.9
952 my $cmd = [@{$self->{rem_ssh}}, 'qm', 'nbdstop', $vmid];
953
954 eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
955 if (my $err = $@) {
956 $self->log('err', $err);
957 $self->{errors} = 1;
958 }
959 }
960
961 # config moved and nbd server stopped - now we can resume vm on target
962 if ($tunnel && $tunnel->{version} && $tunnel->{version} >= 1) {
963 eval {
964 $self->write_tunnel($tunnel, 30, "resume $vmid");
965 };
966 if (my $err = $@) {
967 $self->log('err', $err);
968 $self->{errors} = 1;
969 }
970 } else {
971 my $cmd = [@{$self->{rem_ssh}}, 'qm', 'resume', $vmid, '--skiplock', '--nocheck'];
972 my $logf = sub {
973 my $line = shift;
974 $self->log('err', $line);
975 };
976 eval { PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => $logf); };
977 if (my $err = $@) {
978 $self->log('err', $err);
979 $self->{errors} = 1;
980 }
981 }
982
983 if ($self->{storage_migration} && PVE::QemuServer::parse_guest_agent($conf)->{fstrim_cloned_disks} && $self->{running}) {
984 my $cmd = [@{$self->{rem_ssh}}, 'qm', 'guest', 'cmd', $vmid, 'fstrim'];
985 eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
986 }
987 }
988
989 # close tunnel on successful migration, on error phase2_cleanup closed it
990 if ($tunnel) {
991 eval { finish_tunnel($self, $tunnel); };
992 if (my $err = $@) {
993 $self->log('err', $err);
994 $self->{errors} = 1;
995 }
996 }
997
998 eval {
999 my $timer = 0;
1000 if (PVE::QemuServer::vga_conf_has_spice($conf->{vga}) && $self->{running}) {
1001 $self->log('info', "Waiting for spice server migration");
1002 while (1) {
1003 my $res = PVE::QemuServer::vm_mon_cmd_nocheck($vmid, 'query-spice');
1004 last if int($res->{'migrated'}) == 1;
1005 last if $timer > 50;
1006 $timer ++;
1007 usleep(200000);
1008 }
1009 }
1010 };
1011
1012 # always stop local VM
1013 eval { PVE::QemuServer::vm_stop($self->{storecfg}, $vmid, 1, 1); };
1014 if (my $err = $@) {
1015 $self->log('err', "stopping vm failed - $err");
1016 $self->{errors} = 1;
1017 }
1018
1019 # always deactivate volumes - avoid lvm LVs to be active on several nodes
1020 eval {
1021 my $vollist = PVE::QemuServer::get_vm_volumes($conf);
1022 PVE::Storage::deactivate_volumes($self->{storecfg}, $vollist);
1023 };
1024 if (my $err = $@) {
1025 $self->log('err', $err);
1026 $self->{errors} = 1;
1027 }
1028
1029 if($self->{storage_migration}) {
1030 # destroy local copies
1031 my $volids = $self->{online_local_volumes};
1032
1033 foreach my $volid (@$volids) {
1034 eval { PVE::Storage::vdisk_free($self->{storecfg}, $volid); };
1035 if (my $err = $@) {
1036 $self->log('err', "removing local copy of '$volid' failed - $err");
1037 $self->{errors} = 1;
1038 last if $err =~ /^interrupted by signal$/;
1039 }
1040 }
1041
1042 }
1043
1044 # clear migrate lock
1045 my $cmd = [ @{$self->{rem_ssh}}, 'qm', 'unlock', $vmid ];
1046 $self->cmd_logerr($cmd, errmsg => "failed to clear migrate lock");
1047 }
1048
1049 sub final_cleanup {
1050 my ($self, $vmid) = @_;
1051
1052 # nothing to do
1053 }
1054
1055 sub round_powerof2 {
1056 return 1 if $_[0] < 2;
1057 return 2 << int(log($_[0]-1)/log(2));
1058 }
1059
1060 1;