use PVE::HA::Tools ':exit_codes';
use PVE::HA::Env;
use PVE::HA::Config;
+use PVE::HA::FenceConfig;
+use PVE::HA::Resources;
+use PVE::HA::Resources::PVEVM;
+use PVE::HA::Resources::PVECT;
+PVE::HA::Resources::PVEVM->register();
+PVE::HA::Resources::PVECT->register();
+
+PVE::HA::Resources->init();
my $lockdir = "/etc/pve/priv/lock";
return $self->{nodename};
}
+sub hardware {
+ my ($self) = @_;
+
+ die "hardware is for testing and simulation only";
+}
+
sub read_manager_status {
my ($self) = @_;
return PVE::HA::Config::read_crm_commands();
}
-sub service_config_exists {
- my ($self) = @_;
-
- return PVE::HA::Config::resources_config_exists();
-}
-
sub read_service_config {
my ($self) = @_;
return $conf;
}
-sub change_service_location {
+sub read_fence_config {
+ my ($self) = @_;
+
+ return PVE::HA::Config::read_fence_config();
+}
+
+sub fencing_mode {
+ my ($self) = @_;
+
+ my $datacenterconfig = cfs_read_file('datacenter.cfg');
+
+ return 'watchdog' if !$datacenterconfig->{fencing};
+
+ return $datacenterconfig->{fencing};
+}
+
+sub exec_fence_agent {
+ my ($self, $agent, $node, @param) = @_;
+
+ # setup execution environment
+ $ENV{'PATH'} = '/sbin:/bin:/usr/sbin:/usr/bin';
+
+ my $cmd = "$agent " . PVE::HA::FenceConfig::gen_arg_str(@param);
+
+ exec($cmd);
+ exit -1;
+}
+
+# this is only allowed by the master to recover a _fenced_ service
+sub steal_service {
my ($self, $sid, $current_node, $new_node) = @_;
my (undef, $type, $name) = PVE::HA::Tools::parse_sid($sid);
return $self->get_pve_lock("ha_manager_lock");
}
+# release the cluster wide manager lock.
+# when released another CRM may step up and get the lock, thus this should only
+# get called when shutting down/deactivating the current master
+sub release_ha_manager_lock {
+ my ($self) = @_;
+
+ return rmdir("$lockdir/ha_manager_lock");
+}
+
sub get_ha_agent_lock {
my ($self, $node) = @_;
return $self->get_pve_lock("ha_agent_${node}_lock");
}
+# release the respective node agent lock.
+# this should only get called if the nodes LRM gracefully shuts down with
+# all services already cleanly stopped!
+sub release_ha_agent_lock {
+ my ($self) = @_;
+
+ my $node = $self->nodename();
+
+ return rmdir("$lockdir/ha_agent_${node}_lock");
+}
+
sub quorate {
my ($self) = @_;
}
}
-sub upid_wait {
- my ($self, $upid) = @_;
-
- my $task = PVE::Tools::upid_decode($upid);
-
- CORE::sleep(1);
- while (PVE::ProcFSTools::check_process_running($task->{pid}, $task->{pstart})) {
- $self->log('debug', "Task still active, waiting");
- CORE::sleep(1);
- }
-}
-
-sub can_fork {
+sub after_fork {
my ($self) = @_;
- return 1;
-}
-
-sub exec_resource_agent {
- my ($self, $sid, $service_config, $cmd, @params) = @_;
-
- # setup execution environment
-
- $ENV{'PATH'} = '/sbin:/bin:/usr/sbin:/usr/bin';
-
+ # close inherited inotify FD from parent and reopen our own
PVE::INotify::inotify_close();
-
PVE::INotify::inotify_init();
PVE::Cluster::cfs_update();
+}
- my $nodename = $self->{nodename};
-
- my (undef, $service_type, $service_name) = PVE::HA::Tools::parse_sid($sid);
-
- my $plugin = PVE::HA::Resources->lookup($service_type);
- if (!$plugin) {
- $self->log('err', "service type '$service_type' not implemented");
- return EUNKNOWN_SERVICE_TYPE;
- }
-
- if ($service_config->{node} ne $nodename) {
- $self->log('err', "service '$sid' not on this node");
- return EWRONG_NODE;
- }
-
- my $vmid = $service_name;
-
- my $running = $plugin->check_running($vmid);
-
- if ($cmd eq 'started') {
-
- return SUCCESS if $running;
-
- $self->log("info", "starting service $sid");
-
- my $params = {
- node => $nodename,
- vmid => $vmid
- };
-
- $plugin->start($self, $params);
-
- $running = $plugin->check_running($vmid);
-
- if ($running) {
- $self->log("info", "service status $sid started");
- return SUCCESS;
- } else {
- $self->log("warning", "unable to start service $sid");
- return ERROR;
- }
-
- } elsif ($cmd eq 'request_stop' || $cmd eq 'stopped') {
-
- return SUCCESS if !$running;
-
- $self->log("info", "stopping service $sid");
-
- my $timeout = 60; # fixme: make this configurable
-
- my $params = {
- node => $nodename,
- vmid => $vmid,
- timeout => $timeout,
- forceStop => 1,
- };
-
- $plugin->shutdown($self, $params);
-
- $running = $plugin->check_running($vmid);
-
- if (!$running) {
- $self->log("info", "service status $sid stopped");
- return SUCCESS;
- } else {
- $self->log("info", "unable to stop stop service $sid (still running)");
- return ERROR;
- }
-
- } elsif ($cmd eq 'migrate' || $cmd eq 'relocate') {
-
- my $target = $params[0];
- if (!defined($target)) {
- die "$cmd '$sid' failed - missing target\n" if !defined($target);
- return EINVALID_PARAMETER;
- }
-
- if ($service_config->{node} eq $target) {
- # already there
- return SUCCESS;
- }
-
- # we always do (live) migration
- my $params = {
- node => $nodename,
- vmid => $vmid,
- target => $target,
- online => 1,
- };
-
- my $oldconfig = $plugin->config_file($vmid, $nodename);
-
- $plugin->migrate($self, $params);
-
- # something went wrong if old config file is still there
- if (-f $oldconfig) {
- $self->log("err", "service $sid not moved (migration error)");
- return ERROR;
- }
-
- return SUCCESS;
-
- } elsif ($cmd eq 'error') {
-
- if ($running) {
- $self->log("err", "service $sid is in an error state while running");
- } else {
- $self->log("warning", "service $sid is not running and in an error state");
- }
- return SUCCESS; # error always succeeds
+sub get_max_workers {
+ my ($self) = @_;
- }
+ my $datacenterconfig = cfs_read_file('datacenter.cfg');
- $self->log("err", "implement me (cmd '$cmd')");
- return EUNKNOWN_COMMAND;
+ return $datacenterconfig->{max_workers} || 4;
}
1;
projects / pve-ha-manager.git / blobdiff