fix online migration

PVE/QemuMigrate.pm

@@ -243,6 +243,9 @@ sub migrate {
 	}
 
 	# vm is now owned by other node
+	# Note: there is no VM config file on the local node anymore, so 
+	# we need to pass $nocheck = 1 for vm commands
+
 	my $volids = $rhash->{volumes};
 
 	if ($running) {
@@ -260,10 +263,8 @@ sub migrate {
 		}
 	    }
 
-	    # fixme: ther is no config file, so this will never work
-	    # fixme: use kill(9, $running) to make sure it is stopped
 	    # always stop local VM - no interrupts possible
-	    eval { PVE::QemuServer::vm_stop($session->{vmid}, 1); };
+	    eval { PVE::QemuServer::vm_stop($session->{vmid}, 1, 1); };
 	    if ($@) {
 		logmsg('err', "stopping vm failed - $@");
 		$errors = 1;
@@ -446,7 +447,7 @@ sub phase1 {
 };
 
 sub phase2 {
-    my ($session, $conf, $rhash) = shift;
+    my ($session, $conf, $rhash) = @_;
 
     logmsg('info', "starting VM on remote node '$session->{node}'");
 
@@ -477,12 +478,12 @@ sub phase2 {
 
     my $start = time();
 
-    PVE::QemuServer::vm_monitor_command($session->{vmid}, "migrate -d \"tcp:localhost:$lport\"");
+    PVE::QemuServer::vm_monitor_command($session->{vmid}, "migrate -d \"tcp:localhost:$lport\"", 0, 1);
 
     my $lstat = '';
     while (1) {
 	sleep (2);
-	my $stat = PVE::QemuServer::vm_monitor_command($session->{vmid}, "info migrate", 1);
+	my $stat = PVE::QemuServer::vm_monitor_command($session->{vmid}, "info migrate", 1, 1);
 	if ($stat =~ m/^Migration status: (active|completed|failed|cancelled)$/im) {
 	    my $ms = $1;
 

PVE/QemuServer.pm

@@ -1690,12 +1690,12 @@ sub check_cmdline {
 }
 
 sub check_running {
-    my ($vmid) = @_;
+    my ($vmid, $nocheck) = @_;
 
     my $filename = config_file($vmid);
 
     die "unable to find configuration file for VM $vmid - no such machine\n"
-	if ! -f $filename;
+	if !$nocheck && ! -f $filename;
 
     my $pidfile = pidfile_name($vmid);
 
@@ -1711,8 +1711,11 @@ sub check_running {
 
 	if ($line =~ m/^(\d+)$/) {
 	    my $pid = $1;
-
+	    if (check_cmdline($pidfile, $pid)) {
-	    return $pid if ((-d "/proc/$pid") && check_cmdline ($pidfile, $pid));
+		if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
+		    return $pid;
+		}
+	    }
 	}
     }
 
@@ -2326,6 +2329,7 @@ sub vm_start {
 	    my $cmd = "migrate_set_downtime ${migrate_downtime}";
 	    eval { vm_monitor_command ($vmid, $cmd, 1); };
 	}
+
     });
 }
 
@@ -2362,14 +2366,14 @@ sub __read_avail {
 }
 
 sub vm_monitor_command {
-    my ($vmid, $cmdstr, $nolog) = @_;
+    my ($vmid, $cmdstr, $nolog, $nocheck) = @_;
 
     my $res;
 
     syslog ("info", "VM $vmid monitor command '$cmdstr'") if !$nolog;
 
     eval {
-	die "VM not running\n"  if !check_running ($vmid);
+	die "VM not running\n" if !check_running($vmid, $nocheck);
 
 	my $sname = monitor_socket($vmid);
 
@@ -2477,25 +2481,28 @@ sub vm_shutdown {
     });
 }
 
+# Note: use $nockeck to skip tests if VM configuration file exists.
+# We need that when migration VMs to other nodes (files already moved) 
 sub vm_stop {
-    my ($vmid, $skiplock) = @_;
+    my ($vmid, $skiplock, $nocheck) = @_;
 
     lock_config($vmid, sub {
 
-	my $pid = check_running ($vmid);
+	my $pid = check_running($vmid, $nocheck);
 
 	if (!$pid) {
 	    syslog('info', "VM $vmid already stopped");
 	    return;
 	}
 
+	if (!$nocheck) {
 	    my $conf = load_config($vmid);
-
 	    check_lock($conf) if !$skiplock;
+	}
 
 	syslog ("info", "VM $vmid stopping");
 
-	eval { vm_monitor_command ($vmid, "quit", 1); };
+	eval { vm_monitor_command ($vmid, "quit", 1, $nocheck); };
 
 	my $err = $@;
 
@@ -2504,7 +2511,7 @@ sub vm_stop {
 	    my $timeout = 50; # fixme: how long?
 
 	    my $count = 0;
-	    while (($count < $timeout) && check_running ($vmid)) {
+	    while (($count < $timeout) && check_running($vmid, $nocheck)) {
 		$count++;
 		sleep 1;
 	    }
@@ -2522,7 +2529,7 @@ sub vm_stop {
 	my $timeout = 10;
 
 	my $count = 0;
-	while (($count < $timeout) && check_running ($vmid)) {
+	while (($count < $timeout) && check_running($vmid, $nocheck)) {
 	    $count++;
 	    sleep 1;
 	}