Bug Fix: my previous attempt to catch failed XEN guests via a state machine

change had some problems.

clientside/tmcc/linux/xen/libvnode_xen.pm

@@ -2298,16 +2298,6 @@ sub vnodeBoot($$$$)
     # stated to do its thing, this state name is treated specially.
     libutil::setState("BOOTING");
 
-    #
-    # But, we find ourselves stuck in BOOTING quite often if the VM
-    # fails to boot far enough to to send in a state transition. We want
-    # to catch this specific hangup, so we will send an intermediate
-    # state that the server side can notice, and watch for how long it
-    # stays in the state.
-    #
-    sleep(1);
-    libutil::setState("VNODEBOOTSTART");
-
     #
     # We are going to watch for a busted control network interface, which
     # happens a lot. There is a problem with the control vif not working,
@@ -2350,6 +2340,15 @@ sub vnodeBoot($$$$)
 	    # Ping returns zero if any packets received.
 	    if (! $?) {
 		TBDebugTimeStamp("Created virtual machine $vnode_id");
+		#
+		# But, we still find ourselves stuck in BOOTING quite
+		# often if the VM fails to boot far enough to to send
+		# in a state transition. We want to catch this
+		# specific hangup, so we will send an intermediate
+		# state that the server side can notice, and watch for
+		# how long it stays in the state.
+		#
+		libutil::setState("VNODEBOOTSTART");
 		return 0;
 	    }
 	    $countdown--;

db/EmulabConstants.pm.in

@@ -141,7 +141,7 @@ use vars qw(@ISA @EXPORT);
 	 TBDB_NODEOPMODE_DELAY
 	 TBDB_NODEOPMODE_BOOTWHAT
 	 TBDB_NODEOPMODE_ANY
-	 TBDB_NODEOPMODE_UNKNOWN
+	 TBDB_NODEOPMODE_UNKNOWN TBDB_NODEOPMODE_NORMALv2
 
 	 TBDB_COMMAND_REBOOT
 	 TBDB_COMMAND_POWEROFF TBDB_COMMAND_POWERON TBDB_COMMAND_POWERCYCLE
@@ -511,6 +511,7 @@ sub TBDB_NODEOPMODE_DELAYING	{ "DELAYING"; }
 sub TBDB_NODEOPMODE_UNKNOWNOS	{ "UNKNOWNOS"; }
 sub TBDB_NODEOPMODE_RELOADING	{ "RELOADING"; }
 sub TBDB_NODEOPMODE_NORMALv1	{ "NORMALv1"; }
+sub TBDB_NODEOPMODE_NORMALv2	{ "NORMALv2"; }
 sub TBDB_NODEOPMODE_MINIMAL	{ "MINIMAL"; }
 sub TBDB_NODEOPMODE_PCVM	{ "PCVM"; }
 sub TBDB_NODEOPMODE_RELOAD	{ "RELOAD"; }

db/Node.pm.in

@@ -928,23 +928,28 @@ sub GetAllocState($$)
 #
 # We do this cause we always want to go to the DB.
 #
-sub GetEventState($$)
+sub GetEventState($$;$)
 {
-    my ($self, $pstate) = @_;
+    my ($self, $pstate, $popmode) = @_;
     my $node_id = $self->node_id();
 
     my $query_result =
-	DBQueryWarn("select eventstate from nodes where node_id='$node_id'");
+	DBQueryWarn("select eventstate,op_mode from nodes ".
+		    "where node_id='$node_id'");
     return -1
 	if (!$query_result || !$query_result->numrows);
 
-    my ($state) = $query_result->fetchrow_array();
+    my ($state,$op_mode) = $query_result->fetchrow_array();
     $state = TBDB_NODESTATE_UNKNOWN
 	if (!defined($state));
+    $op_mode = TBDB_NODEOPMODE_UNKNOWN
+	if (!defined($op_mode));
 
     $self->{'DBROW'}->{'eventstate'} = $state
 	if (defined($self->{'DBROW'}));
     $$pstate = $state;
+    $$popmode= $op_mode
+ 	if (defined($popmode));
     return 0;
 }
 

protogeni/lib/GeniAggregate.pm.in

@@ -1771,9 +1771,9 @@ sub WaitForNodes($$@)
 		}
 		$node->_waitstart($parent->_waitend());
 	    }
-	    my $state;
+	    my ($state,$op_mode);
 
-	    if ($node->GetEventState(\$state)) {
+	    if ($node->GetEventState(\$state, \$op_mode)) {
 		print STDERR "*** Error getting event state for $node_id.\n";
 		$node->_sliver()->SetStatus("failed")
 		    if (defined($node->_sliver()));
@@ -1805,6 +1805,7 @@ sub WaitForNodes($$@)
 	    # XEN VMs, a reboot typically solves it.
 	    #
 	    if ($node->isvirtnode() &&
+		$op_mode eq TBDB_NODEOPMODE_NORMALv2 &&
 		$state eq TBDB_NODESTATE_VNODEBOOTSTART &&
 		$state eq $node->_laststate() &&
 		time() - $node->_laststatestamp() > 180) {