[qemu-server.git] / PVE / QemuMigrate.pm

package PVE::QemuMigrate;

use strict;
use warnings;
use PVE::AbstractMigrate;
use IO::File;
use IPC::Open2;
use PVE::INotify;
use PVE::Cluster;
use PVE::Storage;
use PVE::QemuServer;

use base qw(PVE::AbstractMigrate);

sub fork_command_pipe {
    my ($self, $cmd) = @_;

    my $reader = IO::File->new();
    my $writer = IO::File->new();

    my $orig_pid = $$;

    my $cpid;

    eval { $cpid = open2($reader, $writer, @$cmd); };

    my $err = $@;

    # catch exec errors
    if ($orig_pid != $$) {
	$self->log('err', "can't fork command pipe\n");
	POSIX::_exit(1);
	kill('KILL', $$);
    }

    die $err if $err;

    return { writer => $writer, reader => $reader, pid => $cpid };
}

sub finish_command_pipe {
    my ($self, $cmdpipe, $timeout) = @_;

    my $writer = $cmdpipe->{writer};
    my $reader = $cmdpipe->{reader};

    $writer->close();
    $reader->close();

    my $cpid = $cmdpipe->{pid};

    if ($timeout) {
	for (my $i = 0; $i < $timeout; $i++) {
	    return if !PVE::ProcFSTools::check_process_running($cpid);
	    sleep(1);
	}
    }

    $self->log('info', "ssh tunnel still running - terminating now with SIGTERM\n");
    kill(15, $cpid);

    # wait again
    for (my $i = 0; $i < 10; $i++) {
	return if !PVE::ProcFSTools::check_process_running($cpid);
	sleep(1);
    }

    $self->log('info', "ssh tunnel still running - terminating now with SIGKILL\n");
    kill 9, $cpid;
    sleep 1;
}

sub fork_tunnel {
    my ($self, $nodeip, $lport, $rport) = @_;

    my $cmd = [@{$self->{rem_ssh}}, '-L', "$lport:localhost:$rport",
	       'qm', 'mtunnel' ];

    my $tunnel = $self->fork_command_pipe($cmd);

    my $reader = $tunnel->{reader};

    my $helo;
    eval {
	PVE::Tools::run_with_timeout(60, sub { $helo = <$reader>; });
	die "no reply\n" if !$helo;
	die "no quorum on target node\n" if $helo =~ m/^no quorum$/;
	die "got strange reply from mtunnel ('$helo')\n"
	    if $helo !~ m/^tunnel online$/;
    };
    my $err = $@;

    if ($err) {
	$self->finish_command_pipe($tunnel);
	die "can't open migration tunnel - $err";
    }
    return $tunnel;
}

sub finish_tunnel {
    my ($self, $tunnel) = @_;

    my $writer = $tunnel->{writer};

    eval {
	PVE::Tools::run_with_timeout(30, sub {
	    print $writer "quit\n";
	    $writer->flush();
	});
    };
    my $err = $@;

    $self->finish_command_pipe($tunnel, 30);

    die $err if $err;
}

sub lock_vm {
    my ($self, $vmid, $code, @param) = @_;
    
    return PVE::QemuServer::lock_config($vmid, $code, @param);
}

sub prepare {
    my ($self, $vmid) = @_;

    my $online = $self->{opts}->{online};

    $self->{storecfg} = PVE::Storage::config();

    # test is VM exist
    my $conf = $self->{vmconf} = PVE::QemuServer::load_config($vmid);

    PVE::QemuServer::check_lock($conf);

    my $running = 0;
    if (my $pid = PVE::QemuServer::check_running($vmid)) {
	die "cant migrate running VM without --online\n" if !$online;
	$running = $pid;
    }

    if (my $loc_res = PVE::QemuServer::check_local_resources($conf, 1)) {
	if ($self->{running} || !$self->{opts}->{force}) {
	    die "can't migrate VM which uses local devices\n";
	} else {
	    $self->log('info', "migrating VM which uses local devices");
	}
    }

    # activate volumes
    my $vollist = PVE::QemuServer::get_vm_volumes($conf);
    PVE::Storage::activate_volumes($self->{storecfg}, $vollist);

    # fixme: check if storage is available on both nodes

    # test ssh connection
    my $cmd = [ @{$self->{rem_ssh}}, '/bin/true' ];
    eval { $self->cmd_quiet($cmd); };
    die "Can't connect to destination address using public key\n" if $@;

    return $running;
}

sub sync_disks {
    my ($self, $vmid) = @_;

    $self->log('info', "copying disk images");

    my $conf = $self->{vmconf};

    $self->{volumes} = [];

    my $res = [];

    eval {

	my $volhash = {};
	my $cdromhash = {};

	my @sids = PVE::Storage::storage_ids($self->{storecfg});
        foreach my $storeid (@sids) {
	    my $scfg = PVE::Storage::storage_config($self->{storecfg}, $storeid);
            next if $scfg->{shared};
	    next if !PVE::Storage::storage_check_enabled($self->{storecfg}, $storeid, undef, 1);

            # get list from PVE::Storage (for unused volumes)
            my $dl = PVE::Storage::vdisk_list($self->{storecfg}, $storeid, $vmid);
            PVE::Storage::foreach_volid($dl, sub {
                my ($volid, $sid, $volname) = @_;

                # check if storage is available on target node
                PVE::Storage::storage_check_node($self->{storecfg}, $sid, $self->{node});

                $volhash->{$volid} = 1;
            });
        }

	# and add used,owned/non-shared disks (just to be sure we have all)

	my $sharedvm = 1;
	PVE::QemuServer::foreach_drive($conf, sub {
	    my ($ds, $drive) = @_;

	    my $volid = $drive->{file};
	    return if !$volid;

	    die "cant migrate local file/device '$volid'\n" if $volid =~ m|^/|;

	    if (PVE::QemuServer::drive_is_cdrom($drive)) {
		die "cant migrate local cdrom drive\n" if $volid eq 'cdrom';
		return if $volid eq 'none';
		$cdromhash->{$volid} = 1;
	    }

	    my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);

	    # check if storage is available on both nodes
	    my $scfg = PVE::Storage::storage_check_node($self->{storecfg}, $sid);
	    PVE::Storage::storage_check_node($self->{storecfg}, $sid, $self->{node});

	    return if $scfg->{shared};

	    die "can't migrate local cdrom '$volid'\n" if $cdromhash->{$volid};

	    $sharedvm = 0;

	    my ($path, $owner) = PVE::Storage::path($self->{storecfg}, $volid);

	    die "can't migrate volume '$volid' - owned by other VM (owner = VM $owner)\n"
		if !$owner || ($owner != $self->{vmid});

	    $volhash->{$volid} = 1;
	});

	if ($self->{running} && !$sharedvm) {
	    die "can't do online migration - VM uses local disks\n";
	}

	# do some checks first
	foreach my $volid (keys %$volhash) {
	    my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
	    my $scfg =  PVE::Storage::storage_config($self->{storecfg}, $sid);

	    die "can't migrate '$volid' - storagy type '$scfg->{type}' not supported\n"
		if $scfg->{type} ne 'dir';
	}

	foreach my $volid (keys %$volhash) {
	    my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
	    push @{$self->{volumes}}, $volid;
	    PVE::Storage::storage_migrate($self->{storecfg}, $volid, $self->{nodeip}, $sid);
	}
    };
    die "Failed to sync data - $@" if $@;
}

sub phase1 {
    my ($self, $vmid) = @_;

    $self->log('info', "starting migration of VM $vmid to node '$self->{node}' ($self->{nodeip})");

    my $conf = $self->{vmconf};

    # set migrate lock in config file
    $conf->{lock} = 'migrate';
    PVE::QemuServer::update_config_nolock($vmid, $conf, 1);

    sync_disks($self, $vmid);

};

sub phase1_cleanup {
    my ($self, $vmid, $err) = @_;

    $self->log('info', "aborting phase 1 - cleanup resources");

    my $conf = $self->{vmconf};
    delete $conf->{lock};
    eval { PVE::QemuServer::update_config_nolock($vmid, $conf, 1) };
    if (my $err = $@) {
	$self->log('err', $err);
    }
  
    if ($self->{volumes}) {
	foreach my $volid (@{$self->{volumes}}) {
	    $self->log('err', "found stale volume copy '$volid' on node '$self->{node}'");
	    # fixme: try to remove ?
	}
    }
}

sub phase2 {
    my ($self, $vmid) = @_;

    my $conf = $self->{vmconf};

    $self->log('info', "starting VM $vmid on remote node '$self->{node}'");

    my $rport;

    my $nodename = PVE::INotify::nodename();

    ## start on remote node
    my $cmd = [@{$self->{rem_ssh}}, 'qm', 'start',
               $vmid, '--stateuri', 'tcp', '--skiplock', '--migratedfrom', $nodename];

    PVE::Tools::run_command($cmd, outfunc => sub {
	my $line = shift;

	if ($line =~ m/^migration listens on port (\d+)$/) {
	    $rport = $1;
	}
    }, errfunc => sub {});

    die "unable to detect remote migration port\n" if !$rport;

    $self->log('info', "starting migration tunnel");

    ## create tunnel to remote port
    my $lport = PVE::QemuServer::next_migrate_port();
    $self->{tunnel} = $self->fork_tunnel($self->{nodeip}, $lport, $rport);

    $self->log('info', "starting online/live migration on port $lport");
    # start migration

    my $start = time();
    eval {
        PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate", uri => "tcp:localhost:$lport");
    };
    my $merr = $@;

    my $lstat = 0;
    while (1) {
	sleep (2);
	my $stat = PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "query-migrate");
	if ($stat->{status} =~ m/^(active|completed|failed|cancelled)$/im) {
	    $merr = undef;

	    if ($stat->{status} eq 'completed') {
		my $delay = time() - $start;
		if ($delay > 0) {
		    my $mbps = sprintf "%.2f", $conf->{memory}/$delay;
		    $self->log('info', "migration speed: $mbps MB/s");
		}
	    }
	    
	    if ($stat->{status} eq 'failed' || $stat->{status} eq 'cancelled') {
		die "aborting\n"
	    }

	    if ($stat->{status} ne 'active') {
		$self->log('info', "migration status: $stat->{status}");
		last;
	    }

	    if ($stat->{ram}->{transferred} ne $lstat) {
		my $trans = $stat->{ram}->{transferred} || 0;
		my $rem = $stat->{ram}->{remaining} || 0;
		my $total = $stat->{ram}->{total} || 0;

		$self->log('info', "migration status: $stat->{status} (transferred ${trans}, " .
			   "remaining ${rem}), total ${total})");
	    }

	    $lstat = $stat->{ram}->{transferred};

	} else {
	    die $merr if $merr;
	    die "unable to parse migration status '$stat->{status}' - aborting\n";
	}
    }
}

sub phase2_cleanup {
    my ($self, $vmid, $err) = @_;

    $self->log('info', "aborting phase 2 - cleanup resources");

    my $conf = $self->{vmconf};
    delete $conf->{lock};
    eval { PVE::QemuServer::update_config_nolock($vmid, $conf, 1) };
    if (my $err = $@) {
        $self->log('err', $err);
    }

    ## fixme : vm_stop_cleanup on target vm


}

sub phase3 {
    my ($self, $vmid) = @_;
    
    my $volids = $self->{volumes};

    # destroy local copies
    foreach my $volid (@$volids) {
	eval { PVE::Storage::vdisk_free($self->{storecfg}, $volid); };
	if (my $err = $@) {
	    $self->log('err', "removing local copy of '$volid' failed - $err");
	    $self->{errors} = 1;
	    last if $err =~ /^interrupted by signal$/;
	}
    }
}

sub phase3_cleanup {
    my ($self, $vmid, $err) = @_;

    my $conf = $self->{vmconf};

    # move config to remote node
    my $conffile = PVE::QemuServer::config_file($vmid);
    my $newconffile = PVE::QemuServer::config_file($vmid, $self->{node});

    die "Failed to move config to node '$self->{node}' - rename failed: $!\n"
        if !rename($conffile, $newconffile);

    ## now that config file is move, we can resume vm on target if livemigrate
    if ($self->{tunnel}) {

	my $cmd = [@{$self->{rem_ssh}}, 'qm', 'resume', $vmid, '--skiplock'];
	eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
	if (my $err = $@) { 
	    $self->log('err', $err);
	    $self->{errors} = 1;
	}
    }


    # always stop local VM
    eval { PVE::QemuServer::vm_stop($self->{storecfg}, $vmid, 1, 1); };
    if (my $err = $@) {
	$self->log('err', "stopping vm failed - $err");
	$self->{errors} = 1;
    }

    if ($self->{tunnel}) {
	eval { finish_tunnel($self, $self->{tunnel});  };
	if (my $err = $@) {
	    $self->log('err', $err);
	    $self->{errors} = 1;
	}
    }

    # always deactivate volumes - avoid lvm LVs to be active on several nodes
    eval {
	my $vollist = PVE::QemuServer::get_vm_volumes($conf);
	PVE::Storage::deactivate_volumes($self->{storecfg}, $vollist);
    };
    if (my $err = $@) {
	$self->log('err', $err);
	$self->{errors} = 1;
    }

    # clear migrate lock
    my $cmd = [ @{$self->{rem_ssh}}, 'qm', 'unlock', $vmid ];
    $self->cmd_logerr($cmd, errmsg => "failed to clear migrate lock");
}

sub final_cleanup {
    my ($self, $vmid) = @_;

    # nothing to do
}

1;
Commit	Line	Data
3ea94c60	1	package PVE::QemuMigrate;
1ef75254	2
1e3baf05	3	use strict;
3ea94c60	4	use warnings;
16e903f2	5	use PVE::AbstractMigrate;
3ea94c60	6	use IO::File;
1e3baf05	7	use IPC::Open2;
3ea94c60 DM	8	use PVE::INotify;
3ea94c60 DM	9	use PVE::Cluster;
1e3baf05	10	use PVE::Storage;
3ea94c60	11	use PVE::QemuServer;
1e3baf05	12
16e903f2	13	use base qw(PVE::AbstractMigrate);
1e3baf05	14
1ef75254	15	sub fork_command_pipe {
46a84fd4	16	my ($self, $cmd) = @_;
19672434	17
1ef75254 DM	18	my $reader = IO::File->new();
	19	my $writer = IO::File->new();
	20
	21	my $orig_pid = $$;
	22
	23	my $cpid;
	24
	25	eval { $cpid = open2($reader, $writer, @$cmd); };
	26
	27	my $err = $@;
	28
	29	# catch exec errors
	30	if ($orig_pid != $$) {
46a84fd4	31	$self->log('err', "can't fork command pipe\n");
19672434 DM	32	POSIX::_exit(1);
19672434 DM	33	kill('KILL', $$);
1ef75254 DM	34	}
	35
	36	die $err if $err;
	37
	38	return { writer => $writer, reader => $reader, pid => $cpid };
	39	}
	40
19672434	41	sub finish_command_pipe {
97439670	42	my ($self, $cmdpipe, $timeout) = @_;
1ef75254 DM	43
	44	my $writer = $cmdpipe->{writer};
	45	my $reader = $cmdpipe->{reader};
	46
	47	$writer->close();
	48	$reader->close();
	49
	50	my $cpid = $cmdpipe->{pid};
	51
97439670 DM	52	if ($timeout) {
	53	for (my $i = 0; $i < $timeout; $i++) {
	54	return if !PVE::ProcFSTools::check_process_running($cpid);
	55	sleep(1);
	56	}
	57	}
	58
	59	$self->log('info', "ssh tunnel still running - terminating now with SIGTERM\n");
	60	kill(15, $cpid);
1ef75254	61
97439670 DM	62	# wait again
	63	for (my $i = 0; $i < 10; $i++) {
	64	return if !PVE::ProcFSTools::check_process_running($cpid);
	65	sleep(1);
	66	}
	67
	68	$self->log('info', "ssh tunnel still running - terminating now with SIGKILL\n");
	69	kill 9, $cpid;
	70	sleep 1;
1ef75254 DM	71	}
1ef75254 DM	72
1e3baf05	73	sub fork_tunnel {
16e903f2	74	my ($self, $nodeip, $lport, $rport) = @_;
1e3baf05	75
16e903f2	76	my $cmd = [@{$self->{rem_ssh}}, '-L', "$lport:localhost:$rport",
1e3baf05	77	'qm', 'mtunnel' ];
19672434	78
46a84fd4	79	my $tunnel = $self->fork_command_pipe($cmd);
1e3baf05 DM	80
	81	my $reader = $tunnel->{reader};
	82
	83	my $helo;
19672434	84	eval {
17eed025	85	PVE::Tools::run_with_timeout(60, sub { $helo = <$reader>; });
1e3baf05	86	die "no reply\n" if !$helo;
1ef75254	87	die "no quorum on target node\n" if $helo =~ m/^no quorum$/;
19672434	88	die "got strange reply from mtunnel ('$helo')\n"
1e3baf05 DM	89	if $helo !~ m/^tunnel online$/;
	90	};
	91	my $err = $@;
	92
	93	if ($err) {
46a84fd4	94	$self->finish_command_pipe($tunnel);
1e3baf05 DM	95	die "can't open migration tunnel - $err";
	96	}
	97	return $tunnel;
	98	}
	99
19672434	100	sub finish_tunnel {
16e903f2	101	my ($self, $tunnel) = @_;
1e3baf05 DM	102
	103	my $writer = $tunnel->{writer};
	104
19672434	105	eval {
17eed025	106	PVE::Tools::run_with_timeout(30, sub {
1e3baf05 DM	107	print $writer "quit\n";
1e3baf05 DM	108	$writer->flush();
19672434	109	});
1e3baf05 DM	110	};
1e3baf05 DM	111	my $err = $@;
19672434	112
97439670	113	$self->finish_command_pipe($tunnel, 30);
19672434	114
1e3baf05 DM	115	die $err if $err;
	116	}
	117
16e903f2 DM	118	sub lock_vm {
16e903f2 DM	119	my ($self, $vmid, $code, @param) = @_;
3ea94c60	120
16e903f2 DM	121	return PVE::QemuServer::lock_config($vmid, $code, @param);
16e903f2 DM	122	}
ff1a2432	123
16e903f2 DM	124	sub prepare {
16e903f2 DM	125	my ($self, $vmid) = @_;
ff1a2432	126
16e903f2	127	my $online = $self->{opts}->{online};
3ea94c60	128
16e903f2	129	$self->{storecfg} = PVE::Storage::config();
3ea94c60	130
16e903f2 DM	131	# test is VM exist
16e903f2 DM	132	my $conf = $self->{vmconf} = PVE::QemuServer::load_config($vmid);
3ea94c60	133
16e903f2	134	PVE::QemuServer::check_lock($conf);
3ea94c60	135
16e903f2 DM	136	my $running = 0;
	137	if (my $pid = PVE::QemuServer::check_running($vmid)) {
	138	die "cant migrate running VM without --online\n" if !$online;
	139	$running = $pid;
3ea94c60 DM	140	}
3ea94c60 DM	141
16e903f2 DM	142	if (my $loc_res = PVE::QemuServer::check_local_resources($conf, 1)) {
	143	if ($self->{running} \|\| !$self->{opts}->{force}) {
	144	die "can't migrate VM which uses local devices\n";
	145	} else {
	146	$self->log('info', "migrating VM which uses local devices");
	147	}
3ea94c60 DM	148	}
3ea94c60 DM	149
ff1a2432 DM	150	# activate volumes
ff1a2432 DM	151	my $vollist = PVE::QemuServer::get_vm_volumes($conf);
16e903f2 DM	152	PVE::Storage::activate_volumes($self->{storecfg}, $vollist);
	153
	154	# fixme: check if storage is available on both nodes
3ea94c60 DM	155
3ea94c60 DM	156	# test ssh connection
16e903f2 DM	157	my $cmd = [ @{$self->{rem_ssh}}, '/bin/true' ];
16e903f2 DM	158	eval { $self->cmd_quiet($cmd); };
3ea94c60	159	die "Can't connect to destination address using public key\n" if $@;
ff1a2432	160
16e903f2	161	return $running;
3ea94c60 DM	162	}
	163
	164	sub sync_disks {
16e903f2 DM	165	my ($self, $vmid) = @_;
	166
	167	$self->log('info', "copying disk images");
3ea94c60	168
16e903f2 DM	169	my $conf = $self->{vmconf};
	170
	171	$self->{volumes} = [];
3ea94c60 DM	172
	173	my $res = [];
	174
	175	eval {
	176
	177	my $volhash = {};
	178	my $cdromhash = {};
	179
522c8f97 DM	180	my @sids = PVE::Storage::storage_ids($self->{storecfg});
	181	foreach my $storeid (@sids) {
	182	my $scfg = PVE::Storage::storage_config($self->{storecfg}, $storeid);
	183	next if $scfg->{shared};
373ea579 DM	184	next if !PVE::Storage::storage_check_enabled($self->{storecfg}, $storeid, undef, 1);
373ea579 DM	185
80b2cbd1 AD	186	# get list from PVE::Storage (for unused volumes)
	187	my $dl = PVE::Storage::vdisk_list($self->{storecfg}, $storeid, $vmid);
	188	PVE::Storage::foreach_volid($dl, sub {
	189	my ($volid, $sid, $volname) = @_;
	190
373ea579	191	# check if storage is available on target node
80b2cbd1 AD	192	PVE::Storage::storage_check_node($self->{storecfg}, $sid, $self->{node});
	193
	194	$volhash->{$volid} = 1;
	195	});
	196	}
3ea94c60 DM	197
	198	# and add used,owned/non-shared disks (just to be sure we have all)
	199
	200	my $sharedvm = 1;
	201	PVE::QemuServer::foreach_drive($conf, sub {
	202	my ($ds, $drive) = @_;
	203
	204	my $volid = $drive->{file};
	205	return if !$volid;
	206
	207	die "cant migrate local file/device '$volid'\n" if $volid =~ m\|^/\|;
	208
	209	if (PVE::QemuServer::drive_is_cdrom($drive)) {
	210	die "cant migrate local cdrom drive\n" if $volid eq 'cdrom';
	211	return if $volid eq 'none';
	212	$cdromhash->{$volid} = 1;
	213	}
	214
	215	my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
	216
16e903f2 DM	217	# check if storage is available on both nodes
	218	my $scfg = PVE::Storage::storage_check_node($self->{storecfg}, $sid);
	219	PVE::Storage::storage_check_node($self->{storecfg}, $sid, $self->{node});
3ea94c60 DM	220
	221	return if $scfg->{shared};
	222
	223	die "can't migrate local cdrom '$volid'\n" if $cdromhash->{$volid};
	224
	225	$sharedvm = 0;
	226
16e903f2	227	my ($path, $owner) = PVE::Storage::path($self->{storecfg}, $volid);
3ea94c60 DM	228
3ea94c60 DM	229	die "can't migrate volume '$volid' - owned by other VM (owner = VM $owner)\n"
16e903f2	230	if !$owner \|\| ($owner != $self->{vmid});
3ea94c60 DM	231
	232	$volhash->{$volid} = 1;
	233	});
	234
16e903f2	235	if ($self->{running} && !$sharedvm) {
3ea94c60 DM	236	die "can't do online migration - VM uses local disks\n";
	237	}
	238
	239	# do some checks first
	240	foreach my $volid (keys %$volhash) {
	241	my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
16e903f2	242	my $scfg = PVE::Storage::storage_config($self->{storecfg}, $sid);
3ea94c60 DM	243
	244	die "can't migrate '$volid' - storagy type '$scfg->{type}' not supported\n"
	245	if $scfg->{type} ne 'dir';
	246	}
	247
	248	foreach my $volid (keys %$volhash) {
	249	my ($sid, $volname) = PVE::Storage::parse_volume_id($volid);
16e903f2 DM	250	push @{$self->{volumes}}, $volid;
16e903f2 DM	251	PVE::Storage::storage_migrate($self->{storecfg}, $volid, $self->{nodeip}, $sid);
3ea94c60 DM	252	}
	253	};
	254	die "Failed to sync data - $@" if $@;
	255	}
	256
1e3baf05	257	sub phase1 {
16e903f2	258	my ($self, $vmid) = @_;
1e3baf05	259
16e903f2	260	$self->log('info', "starting migration of VM $vmid to node '$self->{node}' ($self->{nodeip})");
1e3baf05	261
16e903f2	262	my $conf = $self->{vmconf};
1e3baf05 DM	263
1e3baf05 DM	264	# set migrate lock in config file
1858638f DM	265	$conf->{lock} = 'migrate';
1858638f DM	266	PVE::QemuServer::update_config_nolock($vmid, $conf, 1);
1e3baf05	267
16e903f2	268	sync_disks($self, $vmid);
1ef75254	269
1e3baf05 DM	270	};
1e3baf05 DM	271
16e903f2 DM	272	sub phase1_cleanup {
	273	my ($self, $vmid, $err) = @_;
	274
	275	$self->log('info', "aborting phase 1 - cleanup resources");
	276
1858638f DM	277	my $conf = $self->{vmconf};
	278	delete $conf->{lock};
	279	eval { PVE::QemuServer::update_config_nolock($vmid, $conf, 1) };
16e903f2 DM	280	if (my $err = $@) {
	281	$self->log('err', $err);
	282	}
	283
	284	if ($self->{volumes}) {
	285	foreach my $volid (@{$self->{volumes}}) {
	286	$self->log('err', "found stale volume copy '$volid' on node '$self->{node}'");
	287	# fixme: try to remove ?
	288	}
	289	}
	290	}
	291
1e3baf05	292	sub phase2 {
16e903f2	293	my ($self, $vmid) = @_;
1e3baf05	294
16e903f2 DM	295	my $conf = $self->{vmconf};
16e903f2 DM	296
46a84fd4	297	$self->log('info', "starting VM $vmid on remote node '$self->{node}'");
1e3baf05 DM	298
	299	my $rport;
	300
7e8dcf2c AD	301	my $nodename = PVE::INotify::nodename();
7e8dcf2c AD	302
19672434	303	## start on remote node
7e8dcf2c AD	304	my $cmd = [@{$self->{rem_ssh}}, 'qm', 'start',
7e8dcf2c AD	305	$vmid, '--stateuri', 'tcp', '--skiplock', '--migratedfrom', $nodename];
1e3baf05	306
72afda82	307	PVE::Tools::run_command($cmd, outfunc => sub {
1e3baf05 DM	308	my $line = shift;
	309
	310	if ($line =~ m/^migration listens on port (\d+)$/) {
	311	$rport = $1;
	312	}
72afda82	313	}, errfunc => sub {});
1e3baf05 DM	314
	315	die "unable to detect remote migration port\n" if !$rport;
	316
16e903f2	317	$self->log('info', "starting migration tunnel");
1ef75254	318
1e3baf05	319	## create tunnel to remote port
1ef75254	320	my $lport = PVE::QemuServer::next_migrate_port();
16e903f2	321	$self->{tunnel} = $self->fork_tunnel($self->{nodeip}, $lport, $rport);
1e3baf05	322
d68afb26	323	$self->log('info', "starting online/live migration on port $lport");
1e3baf05 DM	324	# start migration
	325
	326	my $start = time();
5a7835f5 AD	327	eval {
	328	PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "migrate", uri => "tcp:localhost:$lport");
	329	};
	330	my $merr = $@;
1e3baf05	331
a05b47a8	332	my $lstat = 0;
1e3baf05 DM	333	while (1) {
1e3baf05 DM	334	sleep (2);
5a7835f5	335	my $stat = PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "query-migrate");
5a7835f5	336	if ($stat->{status} =~ m/^(active\|completed\|failed\|cancelled)$/im) {
d68afb26	337	$merr = undef;
1e3baf05	338
5a7835f5	339	if ($stat->{status} eq 'completed') {
1e3baf05 DM	340	my $delay = time() - $start;
	341	if ($delay > 0) {
	342	my $mbps = sprintf "%.2f", $conf->{memory}/$delay;
16e903f2	343	$self->log('info', "migration speed: $mbps MB/s");
1e3baf05 DM	344	}
1e3baf05 DM	345	}
16e903f2	346
5a7835f5	347	if ($stat->{status} eq 'failed' \|\| $stat->{status} eq 'cancelled') {
1e3baf05 DM	348	die "aborting\n"
	349	}
	350
a05b47a8 DM	351	if ($stat->{status} ne 'active') {
	352	$self->log('info', "migration status: $stat->{status}");
	353	last;
	354	}
	355
	356	if ($stat->{ram}->{transferred} ne $lstat) {
	357	my $trans = $stat->{ram}->{transferred} \|\| 0;
	358	my $rem = $stat->{ram}->{remaining} \|\| 0;
	359	my $total = $stat->{ram}->{total} \|\| 0;
	360
	361	$self->log('info', "migration status: $stat->{status} (transferred ${trans}, " .
	362	"remaining ${rem}), total ${total})");
	363	}
	364
	365	$lstat = $stat->{ram}->{transferred};
	366
1e3baf05	367	} else {
d68afb26	368	die $merr if $merr;
5a7835f5	369	die "unable to parse migration status '$stat->{status}' - aborting\n";
1e3baf05	370	}
a05b47a8	371	}
1e3baf05	372	}
16e903f2	373
c04b5b04 AD	374	sub phase2_cleanup {
	375	my ($self, $vmid, $err) = @_;
	376
	377	$self->log('info', "aborting phase 2 - cleanup resources");
	378
	379	my $conf = $self->{vmconf};
	380	delete $conf->{lock};
	381	eval { PVE::QemuServer::update_config_nolock($vmid, $conf, 1) };
	382	if (my $err = $@) {
	383	$self->log('err', $err);
	384	}
	385
	386	## fixme : vm_stop_cleanup on target vm
	387
	388
	389	}
	390
16e903f2 DM	391	sub phase3 {
	392	my ($self, $vmid) = @_;
	393
	394	my $volids = $self->{volumes};
	395
	396	# destroy local copies
	397	foreach my $volid (@$volids) {
	398	eval { PVE::Storage::vdisk_free($self->{storecfg}, $volid); };
	399	if (my $err = $@) {
	400	$self->log('err', "removing local copy of '$volid' failed - $err");
	401	$self->{errors} = 1;
	402	last if $err =~ /^interrupted by signal$/;
	403	}
	404	}
16e903f2 DM	405	}
	406
	407	sub phase3_cleanup {
	408	my ($self, $vmid, $err) = @_;
	409
	410	my $conf = $self->{vmconf};
	411
b8d20802 AD	412	# move config to remote node
	413	my $conffile = PVE::QemuServer::config_file($vmid);
	414	my $newconffile = PVE::QemuServer::config_file($vmid, $self->{node});
	415
	416	die "Failed to move config to node '$self->{node}' - rename failed: $!\n"
	417	if !rename($conffile, $newconffile);
	418
b67900f1 AD	419	## now that config file is move, we can resume vm on target if livemigrate
	420	if ($self->{tunnel}) {
	421
	422	my $cmd = [@{$self->{rem_ssh}}, 'qm', 'resume', $vmid, '--skiplock'];
	423	eval{ PVE::Tools::run_command($cmd, outfunc => sub {}, errfunc => sub {}) };
	424	if (my $err = $@) {
	425	$self->log('err', $err);
	426	$self->{errors} = 1;
	427	}
	428	}
	429
	430
16e903f2 DM	431	# always stop local VM
	432	eval { PVE::QemuServer::vm_stop($self->{storecfg}, $vmid, 1, 1); };
	433	if (my $err = $@) {
	434	$self->log('err', "stopping vm failed - $err");
	435	$self->{errors} = 1;
	436	}
	437
97439670 DM	438	if ($self->{tunnel}) {
	439	eval { finish_tunnel($self, $self->{tunnel}); };
	440	if (my $err = $@) {
	441	$self->log('err', $err);
	442	$self->{errors} = 1;
	443	}
	444	}
	445
16e903f2 DM	446	# always deactivate volumes - avoid lvm LVs to be active on several nodes
	447	eval {
	448	my $vollist = PVE::QemuServer::get_vm_volumes($conf);
	449	PVE::Storage::deactivate_volumes($self->{storecfg}, $vollist);
	450	};
	451	if (my $err = $@) {
	452	$self->log('err', $err);
	453	$self->{errors} = 1;
	454	}
	455
	456	# clear migrate lock
	457	my $cmd = [ @{$self->{rem_ssh}}, 'qm', 'unlock', $vmid ];
	458	$self->cmd_logerr($cmd, errmsg => "failed to clear migrate lock");
	459	}
	460
	461	sub final_cleanup {
	462	my ($self, $vmid) = @_;
	463
	464	# nothing to do
	465	}
	466
	467	1;