This commit fixes a bug where the drbd kernel module build was being killed mid-compile, leaving DBRD unusable.

* Created System->wait_on_dnf() which was plucked from anvil-daemon, and now also called in scancore and anvil-safe-start.
* Updated scancore and anvil-safe-start to check on start that DRBD's kernel module is available (and build if not).

Signed-off-by: digimer <mkelly@alteeve.ca>

Anvil/Tools/DRBD.pm

@@ -2251,6 +2251,8 @@ sub get_status
 		if ($is_local)
 		{
 			# Try rebuilding the module.
+			$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "message_0328"});
+			
 			my $problem = $anvil->DRBD->_initialize_kmod({debug => 2});
 			$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { problem => $problem }});
 			
@@ -2264,7 +2266,7 @@ sub get_status
 				}});
 				if ($output =~ /modprobe: FATAL: Module drbd not found/i)
 				{
-					$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "error_0415", variables => { 
+					$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "error_0415", variables => { 
 						output      => $output,
 						return_code => $anvil->data->{drbd}{status}{$host}{return_code},
 					}});

Anvil/Tools/System.pm

@@ -46,6 +46,7 @@ my $THIS_FILE = "System.pm";
 # stop_daemon
 # stty_echo
 # update_hosts
+# wait_on_dnf
 # _check_anvil_conf
 # _load_firewalld_zones
 # _load_specific_firewalld_zone
@@ -5389,6 +5390,60 @@ sub update_hosts
 	return(0);
 }
 
+=head2 wait_on_dnf
+
+This method checks to see if 'dnf' is running and, if so, won't return until it finishes. This is useful when holding off doing certain tasks, like building kernel modules, while an OS update is under way.
+
+This method takes no parameters.
+
+=cut
+sub wait_on_dnf
+{
+	my $self      = shift;
+	my $parameter = shift;
+	my $anvil     = $self->parent;
+	my $debug     = defined $parameter->{debug} ? $parameter->{debug} : 3;
+	$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => $debug, key => "log_0125", variables => { method => "System->wait_on_dnf()" }});
+	
+	my $next_log = time - 1;
+	my $waiting  = 1;
+	$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { 
+		next_log => $next_log, 
+		waiting  => $waiting,
+	}});
+	while ($waiting)
+	{
+		my $pids = $anvil->System->pids({program_name => $anvil->data->{path}{exe}{dnf}, debug => $debug});
+		my $dnf_instances = @{$pids};
+		$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { dnf_instances => $dnf_instances }});
+		
+		if ($dnf_instances)
+		{
+			if (time > $next_log)
+			{
+				my $say_pids = "";
+				foreach my $pid (@{$pids})
+				{
+					$say_pids .= $pid.", ";
+				}
+				$say_pids =~ s/, $//;
+				$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "message_0325", variables => { pids => $say_pids }});
+				
+				$next_log = time + 60;
+				$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { next_log => $next_log }});
+			}
+			sleep 10;
+		}
+		else
+		{
+			$waiting = 0;
+			$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { waiting => $waiting }});
+		}
+	}
+	
+	return(0);
+}
+
 # =head3
 # 
 # Private Functions;

share/words.xml

@@ -2926,6 +2926,7 @@ Proceed? [y/N]</key>
 		<key name="message_0325">[ Note ] - It looks like 'dnf' (pid(s): [#!variable!pids!#]) is running, holding our start up until it's done (in case the system is being updated now).</key>
 		<key name="message_0326">This daemon just started. Holding off starting jobs for another: [#!variable!will_start_in!#] second(s).</key>
 		<key name="message_0327">[ Note ] - It looks like 'anvil-version-changes' (pid(s): [#!variable!pids!#]) is running, holding off on power action until it's done (in case the system is being updated now or kernel modules are being built).</key>
+		<key name="message_0328">[ Note ] - The DRBD (replicated storage) kernel module appears to not exist. This is normal after an OS update, will try building the kernel module now. Please be patient.</key>
 
 		<!-- Translate names (protocols, etc) -->
 		<key name="name_0001">Normal Password</key>	<!-- none in mail-server -->

tools/anvil-daemon

@@ -68,7 +68,7 @@ if (($< != 0) && ($> != 0))
 $anvil->System->_check_anvil_conf();
 
 # If dnf is running, hold.
-wait_on_dnf($anvil);
+$anvil->System->wait_on_dnf();
 
 # Connect to the database(s). If we have no connections, we'll proceed anyway as one of the 'run_once' tasks
 # is to setup the database server.
@@ -252,49 +252,6 @@ $anvil->nice_exit({exit_code => 0});
 # Functions                                                                                                 #
 #############################################################################################################
 
-# This checks to see if dnf is running when this daemon starts (as happens during an OS update) and hold 
-# until the dnf update is complete.
-sub wait_on_dnf
-{
-	my ($anvil) = @_;
-	
-	my $next_log = time - 1;
-	my $waiting  = 1;
-	$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { 
-		next_log => $next_log, 
-		waiting  => $waiting,
-	}});
-	while ($waiting)
-	{
-		my $pids = $anvil->System->pids({program_name => $anvil->data->{path}{exe}{dnf}});
-		my $dnf_instances = @{$pids};
-		$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { dnf_instances => $dnf_instances }});
-		
-		if ($dnf_instances)
-		{
-			if (time > $next_log)
-			{
-				my $say_pids = "";
-				foreach my $pid (@{$pids})
-				{
-					$say_pids .= $pid.", ";
-				}
-				$say_pids =~ s/, $//;
-				$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "message_0325", variables => { pids => $say_pids }});
-				
-				$next_log = time + 60;
-				$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { next_log => $next_log }});
-			}
-			sleep 10;
-		}
-		else
-		{
-			$waiting = 0;
-			$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { waiting => $waiting }});
-		}
-	}
-}
-
 # If we're using too much ram, send an alert and exit.
 sub check_ram
 {

tools/anvil-safe-start

@@ -53,6 +53,9 @@ if (($< != 0) && ($> != 0))
 	$anvil->nice_exit({exit_code => 1});
 }
 
+# If dnf is running, hold.
+$anvil->System->wait_on_dnf();
+
 # Connect to the database(s). If we have no connections, we'll proceed anyway as one of the 'run_once' tasks
 # is to setup the database server.
 $anvil->Database->connect();
@@ -585,6 +588,13 @@ sub prerun_checks
 		host_type => $host_type, 
 	}});
 	
+	if (($host_type eq "node") or ($host_type eq "dr"))
+	{
+		# Call DRBD->get_status because, if we're just starting up and the kernel module needs to be
+		# built, do it before we start calling scan agents.
+		$anvil->DRBD->get_status({debug => 2});
+	}
+	
 	if ($host_type ne "node")
 	{
 		# We're done.

tools/scancore

@@ -67,6 +67,9 @@ $anvil->data->{scancore} = {
 
 $anvil->Storage->read_config();
 
+# If dnf is running, hold.
+$anvil->System->wait_on_dnf();
+
 # Read switches
 $anvil->Get->switches({list => ["purge", "run-once"], man => $THIS_FILE});
 $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => $anvil->data->{switches}});
@@ -431,10 +434,12 @@ sub startup_tasks
 	
 	# This used to call anvil-safe-start, which isn't done here anymore.
 	my $host_type = $anvil->Get->host_type();
-	$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { host_type => $host_type }});
-	if ($host_type eq "node")
+	$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { host_type => $host_type }});
+	if (($host_type eq "node") or ($host_type eq "dr"))
 	{
-		# For future use.
+		# Call DRBD->get_status because, if we're just starting up and the kernel module needs to be
+		# built, do it before we start calling scan agents.
+		$anvil->DRBD->get_status({debug => 2});
 	}
 	elsif ($host_type eq "striker")
 	{