write lrm mode into lrm status file

LRM is normally in 'active' mode, but can be set to 'reboot', 'shutdown' or 'restart'.
We use this to freeze services, so that we can safely reboot a node, or restart
the LRM.

diff --git a/src/PVE/HA/LRM.pm b/src/PVE/HA/LRM.pm
index da30cc8aa024d1b1aec7eb9eec250ecfd9e480f0..99041017b626dec96bc7427202d0fa31b2a6b3b4 100644 (file)
--- a/src/PVE/HA/LRM.pm
+++ b/src/PVE/HA/LRM.pm
@@ -30,9 +30,13 @@ sub new {
        workers => {},
        results => {},
        shutdown_request => 0,
+       # mode can be: active, reboot, shutdown, restart
+       mode => 'active',
     }, $class;
 
     $self->set_local_status({ state =>         'wait_for_agent_lock' });   
+
+    $self->update_lrm_status();
     
     return $self;
 }
@@ -41,6 +45,13 @@ sub shutdown_request {
     my ($self) = @_;
 
     $self->{shutdown_request} = 1;
+
+    $self->{mode} = 'restart'; # fixme: detect shutdown/reboot
+
+    eval { $self->update_lrm_status(); };
+    if (my $err = $@) {
+       $self->log('err', "unable to update lrm status file");
+    }
 }
 
 sub get_local_status {
@@ -68,6 +79,17 @@ sub set_local_status {
     $self->{status} = $new;
 }
 
+sub update_lrm_status {
+    my ($self) = @_;
+
+    my $lrm_status = { 
+       mode => $self->{mode},
+       results => $self->{results},
+    };
+    
+    $self->{haenv}->write_lrm_status($lrm_status);
+}
+
 sub get_protected_ha_agent_lock {
     my ($self) = @_;
 
@@ -117,6 +139,7 @@ sub active_service_count {
        my $req_state = $sd->{state};
        next if !defined($req_state);
        next if $req_state eq 'stopped';
+       next if $req_state eq 'freeze';
 
        $count++;
     }
@@ -275,6 +298,7 @@ sub manage_resources {
        next if $sd->{node} ne $nodename;
        my $req_state = $sd->{state};
        next if !defined($req_state);
+       next if $req_state eq 'freeze';
        eval {
            $self->queue_resource_command($sid, $sd->{uid}, $req_state, $sd->{target});
        };
@@ -422,7 +446,7 @@ sub resource_command_finished {
     }
     $self->{results} = $results;
 
-    $haenv->write_lrm_status($results);
+    $self->update_lrm_status();
 }
 
 1;

diff --git a/src/PVE/HA/Manager.pm b/src/PVE/HA/Manager.pm
index 1c653de5d4a1f85bad57f4fad31a099cc5788428..61e87400b2aa6cd2c75e00d19bc9fd7142296d3e 100644 (file)
--- a/src/PVE/HA/Manager.pm
+++ b/src/PVE/HA/Manager.pm
@@ -140,6 +140,7 @@ my $valid_service_states = {
     fence => 1,
     migrate => 1,
     relocate => 1,
+    freeze => 1,
     error => 1,
 };
 
@@ -159,7 +160,7 @@ sub recompute_online_node_usage {
        my $state = $sd->{state};
        if (defined($online_node_usage->{$sd->{node}})) {
            if (($state eq 'started') || ($state eq 'request_stop') || 
-               ($state eq 'fence') || ($state eq 'error')) {
+               ($state eq 'fence') || ($state eq 'freeze') || ($state eq 'error')) {
                $online_node_usage->{$sd->{node}}++;
            } elsif (($state eq 'migrate') || ($state eq 'relocate')) {
                $online_node_usage->{$sd->{target}}++;
@@ -210,24 +211,26 @@ my $change_service_state = sub {
     $haenv->log('info', "service '$sid': state changed from '${old_state}' to '${new_state}' $text_state");
 };
 
-# read LRM status for all active nodes 
+# read LRM status for all nodes 
 sub read_lrm_status {
     my ($self) = @_;
 
-    my $nodes = $self->{ns}->list_online_nodes();
+    my $nodes = $self->{ns}->list_nodes();
     my $haenv = $self->{haenv};
 
-    my $res = {};
-
+    my $results = {};
+    my $modes = {};
     foreach my $node (@$nodes) {
-       my $ls = $haenv->read_lrm_status($node);
-       foreach my $uid (keys %$ls) {
-           next if $res->{$uid}; # should not happen
-           $res->{$uid} = $ls->{$uid};
+       my $lrm_status = $haenv->read_lrm_status($node);
+       $modes->{$node} = $lrm_status->{mode} || 'unknown';
+       foreach my $uid (keys %{$lrm_status->{results}}) {
+           next if $results->{$uid}; # should not happen
+           $results->{$uid} = $lrm_status->{results}->{$uid};
        }
     }
 
-    return $res;
+    
+    return ($results, $modes);
 }
 
 # read new crm commands and save them into crm master status
@@ -277,7 +280,7 @@ sub manage {
        return;
     }
 
-    my $lrm_status = $self->read_lrm_status();
+    my ($lrm_results, $lrm_modes) = $self->read_lrm_status();
 
     my $sc = $haenv->read_service_config();
 
@@ -312,7 +315,7 @@ sub manage {
            my $sd = $ss->{$sid};
            my $cd = $sc->{$sid} || { state => 'disabled' };
 
-           my $lrm_res = $sd->{uid} ? $lrm_status->{$sd->{uid}} : undef;
+           my $lrm_res = $sd->{uid} ? $lrm_results->{$sd->{uid}} : undef;
 
            my $last_state = $sd->{state};
 
@@ -336,6 +339,14 @@ sub manage {
 
                $self->next_state_request_stop($sid, $cd, $sd, $lrm_res);
 
+           } elsif ($last_state eq 'freeze') {
+
+               my $lrm_mode = $sd->{node} ? $lrm_modes->{$sd->{node}} : undef;
+               $lrm_mode = 'unknown'if !$lrm_mode;
+               # unfreeze
+               &$change_service_state($self, $sid, 'started') 
+                   if $lrm_mode eq 'active';
+
            } elsif ($last_state eq 'error') {
 
                # fixme: 
@@ -345,6 +356,14 @@ sub manage {
                die "unknown service state '$last_state'";
            }
 
+
+           my $lrm_mode = $sd->{node} ? $lrm_modes->{$sd->{node}} : undef;
+           $lrm_mode = 'unknown'if !$lrm_mode;
+           if (($sd->{state} eq 'started' || $sd->{state} eq 'stopped' ||
+                $sd->{state} eq 'request_stop') && ($lrm_mode ne 'active')) {
+               &$change_service_state($self, $sid, 'freeze');
+           }
+                   
            $repeat = 1 if $sd->{state} ne $last_state;
        }
 

diff --git a/src/PVE/HA/NodeStatus.pm b/src/PVE/HA/NodeStatus.pm
index 285cf931ed4c6fc3efb6e9be5d5e324848f5bc01..fe8c0ef413186d1ea9608abef41390e523680ea6 100644 (file)
--- a/src/PVE/HA/NodeStatus.pm
+++ b/src/PVE/HA/NodeStatus.pm
@@ -60,6 +60,12 @@ sub node_is_offline_delayed {
     return ($ctime - $last_online) >= $delay;
 }
 
+sub list_nodes {
+    my ($self) = @_;
+
+    return [sort keys %{$self->{status}}];
+}
+
 sub list_online_nodes {
     my ($self) = @_;