}
sub select_service_node {
- my ($groups, $online_node_usage, $service_conf, $current_node, $try_next, $tried_nodes) = @_;
+ my ($groups, $online_node_usage, $service_conf, $current_node, $try_next, $tried_nodes, $maintenance_fallback) = @_;
my $group = get_service_group($groups, $online_node_usage, $service_conf);
} keys %{$pri_groups->{$top_pri}};
my $found;
+ my $found_maintenace_fallback;
for (my $i = scalar(@nodes) - 1; $i >= 0; $i--) {
my $node = $nodes[$i];
if ($node eq $current_node) {
$found = $i;
- last;
}
+ if (defined($maintenance_fallback) && $node eq $maintenance_fallback) {
+ $found_maintenace_fallback = $i;
+ }
+ }
+
+ if (defined($found_maintenace_fallback)) {
+ return $nodes[$found_maintenace_fallback];
}
if ($try_next) {
($state eq 'fence') || ($state eq 'freeze') || ($state eq 'error')) {
$online_node_usage->{$sd->{node}}++;
} elsif (($state eq 'migrate') || ($state eq 'relocate')) {
+ # count it for both, source and target as load is put on both
+ $online_node_usage->{$sd->{node}}++;
$online_node_usage->{$sd->{target}}++;
} elsif ($state eq 'stopped') {
# do nothing
my $old_state = $sd->{state};
my $old_node = $sd->{node};
my $old_failed_nodes = $sd->{failed_nodes};
+ my $old_maintenance_node = $sd->{maintenance_node};
die "no state change" if $old_state eq $new_state; # just to be sure
$sd->{state} = $new_state;
$sd->{node} = $old_node;
$sd->{failed_nodes} = $old_failed_nodes if defined($old_failed_nodes);
+ $sd->{maintenance_node} = $old_maintenance_node if defined($old_maintenance_node);
my $text_state = '';
foreach my $k (sort keys %params) {
&$fence_recovery_cleanup($self, $sid, $fenced_node);
$haenv->steal_service($sid, $sd->{node}, $recovery_node);
+ $self->{online_node_usage}->{$recovery_node}++;
# $sd *is normally read-only*, fencing is the exception
$cd->{node} = $sd->{node} = $recovery_node;
my ($haenv, $ms, $ns, $ss) = ($self->{haenv}, $self->{ms}, $self->{ns}, $self->{ss});
- $ns->update($haenv->get_node_info());
+ my ($node_info) = $haenv->get_node_info();
+ my ($lrm_results, $lrm_modes) = $self->read_lrm_status();
+
+ $ns->update($node_info, $lrm_modes);
- if (!$ns->node_is_online($haenv->nodename())) {
+ if (!$ns->node_is_operational($haenv->nodename())) {
$haenv->log('info', "master seems offline");
return;
}
- my ($lrm_results, $lrm_modes) = $self->read_lrm_status();
-
my $sc = $haenv->read_service_config();
$self->{groups} = $haenv->read_group_config(); # update
if ($ns->node_is_offline_delayed($sd->{node})) {
&$change_service_state($self, $sid, 'fence');
}
- return;
+ if ($ns->get_node_state($sd->{node}) ne 'maintenance') {
+ return;
+ } else {
+ # save current node as fallback for when it comes out of
+ # maintenance
+ $sd->{maintenance_node} = $sd->{node};
+ }
}
if ($cd->{state} eq 'disabled' || $cd->{state} eq 'stopped') {
}
}
- my $node = select_service_node($self->{groups}, $self->{online_node_usage},
- $cd, $sd->{node}, $try_next, $sd->{failed_nodes});
+ my $node = select_service_node(
+ $self->{groups},
+ $self->{online_node_usage},
+ $cd,
+ $sd->{node},
+ $try_next,
+ $sd->{failed_nodes},
+ $sd->{maintenance_node},
+ );
if ($node && ($sd->{node} ne $node)) {
+ $self->{online_node_usage}->{$node}++;
+
+ if (defined(my $fallback = $sd->{maintenance_node})) {
+ if ($node eq $fallback) {
+ $haenv->log('info', "moving service '$sid' back to '$fallback', node came back from maintenance.");
+ delete $sd->{maintenance_node};
+ } elsif ($sd->{node} ne $fallback) {
+ $haenv->log('info', "dropping maintenance fallback node '$fallback' for '$sid'");
+ delete $sd->{maintenance_node};
+ }
+ }
+
if ($cd->{type} eq 'vm') {
$haenv->log('info', "migrate service '$sid' to node '$node' (running)");
&$change_service_state($self, $sid, 'migrate', node => $sd->{node}, target => $node);