use strict;
use warnings;
-use Net::IP;
use File::Path;
use File::Basename;
use PVE::Tools qw(run_command);
use PVE::Cluster;
use PVE::INotify;
-use PVE::JSONSchema;
+use PVE::JSONSchema qw(get_standard_option);
+use PVE::RPCEnvironment;
use PVE::CLIHandler;
+use PVE::PTY;
+use PVE::API2::ClusterConfig;
use PVE::Corosync;
use base qw(PVE::CLIHandler);
my $basedir = "/etc/pve";
my $clusterconf = "$basedir/corosync.conf";
my $libdir = "/var/lib/pve-cluster";
-my $backupdir = "/var/lib/pve-cluster/backup";
-my $dbfile = "$libdir/config.db";
my $authfile = "/etc/corosync/authkey";
-sub backup_database {
- print "backup old database\n";
-
- mkdir $backupdir;
-
- my $ctime = time();
- my $cmd = [
- ['echo', '.dump'],
- ['sqlite3', $dbfile],
- ['gzip', '-', \ ">${backupdir}/config-${ctime}.sql.gz"],
- ];
-
- run_command($cmd, 'errmsg' => "cannot backup old database\n");
-
- # purge older backup
- my $maxfiles = 10;
-
- my @bklist = ();
- foreach my $fn (<$backupdir/config-*.sql.gz>) {
- if ($fn =~ m!/config-(\d+)\.sql.gz$!) {
- push @bklist, [$fn, $1];
- }
- }
-
- @bklist = sort { $b->[1] <=> $a->[1] } @bklist;
-
- while (scalar (@bklist) >= $maxfiles) {
- my $d = pop @bklist;
- print "delete old backup '$d->[0]'\n";
- unlink $d->[0];
- }
+sub setup_environment {
+ PVE::RPCEnvironment->setup_default_cli_env();
}
__PACKAGE__->register_method ({
return undef;
}});
-__PACKAGE__->register_method ({
- name => 'create',
- path => 'create',
- method => 'PUT',
- description => "Generate new cluster configuration.",
- parameters => {
- additionalProperties => 0,
- properties => {
- clustername => {
- description => "The name of the cluster.",
- type => 'string', format => 'pve-node',
- maxLength => 15,
- },
- nodeid => {
- type => 'integer',
- description => "Node id for this node.",
- minimum => 1,
- optional => 1,
- },
- votes => {
- type => 'integer',
- description => "Number of votes for this node.",
- minimum => 1,
- optional => 1,
- },
- bindnet0_addr => {
- type => 'string', format => 'ip',
- description => "This specifies the network address the corosync ring 0".
- " executive should bind to and defaults to the local IP address of the node.",
- optional => 1,
- },
- ring0_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring0 address of this node.".
- " Defaults to the hostname of the node.",
- optional => 1,
- },
- bindnet1_addr => {
- type => 'string', format => 'ip',
- description => "This specifies the network address the corosync ring 1".
- " executive should bind to and is optional.",
- optional => 1,
- },
- ring1_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring1 address, this".
- " needs an valid bindnet1_addr.",
- optional => 1,
- },
- },
- },
- returns => { type => 'null' },
-
- code => sub {
- my ($param) = @_;
-
- -f $clusterconf && die "cluster config '$clusterconf' already exists\n";
-
- PVE::Cluster::setup_sshd_config(1);
- PVE::Cluster::setup_rootsshconfig();
- PVE::Cluster::setup_ssh_keys();
-
- -f $authfile || __PACKAGE__->keygen({filename => $authfile});
-
- -f $authfile || die "no authentication key available\n";
-
- my $clustername = $param->{clustername};
-
- $param->{nodeid} = 1 if !$param->{nodeid};
-
- $param->{votes} = 1 if !defined($param->{votes});
-
- my $nodename = PVE::INotify::nodename();
-
- my $local_ip_address = PVE::Cluster::remote_node_ip($nodename);
-
- $param->{bindnet0_addr} = $local_ip_address
- if !defined($param->{bindnet0_addr});
-
- $param->{ring0_addr} = $nodename if !defined($param->{ring0_addr});
-
- die "Param bindnet1_addr and ring1_addr are dependend, use both or none!\n"
- if (defined($param->{bindnet1_addr}) != defined($param->{ring1_addr}));
-
- my $bind_is_ipv6 = Net::IP::ip_is_ipv6($param->{bindnet0_addr});
-
- # use string as here-doc format distracts more
- my $interfaces = "interface {\n ringnumber: 0\n" .
- " bindnetaddr: $param->{bindnet0_addr}\n }";
-
- my $ring_addresses = "ring0_addr: $param->{ring0_addr}" ;
-
- # allow use of multiple rings (rrp) at cluster creation time
- if ($param->{bindnet1_addr}) {
- die "IPv6 and IPv4 cannot be mixed, use one or the other!\n"
- if Net::IP::ip_is_ipv6($param->{bindnet1_addr}) != $bind_is_ipv6;
-
- $interfaces .= "\n interface {\n ringnumber: 1\n" .
- " bindnetaddr: $param->{bindnet1_addr}\n }\n";
-
- $interfaces .= "rrp_mode: passive\n"; # only passive is stable and tested
-
- $ring_addresses .= "\n ring1_addr: $param->{ring1_addr}";
-
- } elsif($param->{rrp_mode} && $param->{rrp_mode} ne 'none') {
-
- warn "rrp_mode '$param->{rrp_mode}' useless when using only one".
- " ring, using 'none' instead";
- # corosync defaults to none if only one interface is configured
- $param->{rrp_mode} = undef;
+my $foreach_member = sub {
+ my ($code, $noerr) = @_;
+ my $members = PVE::Cluster::get_members();
+ foreach my $node (sort keys %$members) {
+ if (my $ip = $members->{$node}->{ip}) {
+ $code->($node, $ip);
+ } else {
+ die "cannot get the cluster IP for node '$node'.\n" if !$noerr;
+ warn "cannot get the cluster IP for node '$node'.\n";
+ return undef;
}
-
- # No, corosync cannot deduce this on its own
- my $ipversion = $bind_is_ipv6 ? 'ipv6' : 'ipv4';
-
- my $config = <<_EOD;
-totem {
- version: 2
- secauth: on
- cluster_name: $clustername
- config_version: 1
- ip_version: $ipversion
- $interfaces
-}
-
-nodelist {
- node {
- $ring_addresses
- name: $nodename
- nodeid: $param->{nodeid}
- quorum_votes: $param->{votes}
- }
-}
-
-quorum {
- provider: corosync_votequorum
-}
-
-logging {
- to_syslog: yes
- debug: off
-}
-_EOD
-;
- PVE::Tools::file_set_contents($clusterconf, $config);
-
- PVE::Cluster::ssh_merge_keys();
-
- PVE::Cluster::gen_pve_node_files($nodename, $local_ip_address);
-
- PVE::Cluster::ssh_merge_known_hosts($nodename, $local_ip_address, 1);
-
- run_command('systemctl restart pve-cluster'); # restart
-
- run_command('systemctl restart corosync'); # restart
-
- return undef;
-}});
+ }
+};
__PACKAGE__->register_method ({
- name => 'addnode',
- path => 'addnode',
+ name => 'setup_qdevice',
+ path => 'setup_qdevice',
method => 'PUT',
- description => "Adds a node to the cluster configuration.",
+ description => "Setup the use of a QDevice",
parameters => {
- additionalProperties => 0,
+ additionalProperties => 0,
properties => {
- node => PVE::JSONSchema::get_standard_option('pve-node'),
- nodeid => {
- type => 'integer',
- description => "Node id for this node.",
- minimum => 1,
- optional => 1,
+ address => {
+ type => 'string', format => 'ip',
+ description => "Specifies the network address of an external corosync QDevice" ,
},
- votes => {
- type => 'integer',
- description => "Number of votes for this node",
- minimum => 0,
+ network => {
+ type => 'string',
+ format => 'CIDR',
+ description => 'The network which should be used to connect to the external qdevice',
optional => 1,
},
force => {
type => 'boolean',
- description => "Do not throw error if node already exists.",
- optional => 1,
- },
- ring0_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring0 address of this node.".
- " Defaults to nodes hostname.",
- optional => 1,
- },
- ring1_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring1 address, this".
- " needs an valid bindnet1_addr.",
+ description => "Do not throw error on possible dangerous operations.",
optional => 1,
},
},
code => sub {
my ($param) = @_;
- if (!$param->{force} && (-t STDIN || -t STDOUT)) {
- die "error: `addnode` should not get called interactively!\nUse ".
- "`pvecm add <cluster-node>` to add a node to a cluster!\n";
+ die "Node not in a cluster. Aborting.\n"
+ if !PVE::Corosync::check_conf_exists(1);
+
+ my $members = PVE::Cluster::get_members();
+ foreach my $node (sort keys %$members) {
+ die "All nodes must be online! Node $node is offline, aborting.\n"
+ if !$members->{$node}->{online};
}
- PVE::Cluster::check_cfs_quorum();
+ my $conf = PVE::Cluster::cfs_read_file("corosync.conf");
- my $code = sub {
- my $conf = PVE::Cluster::cfs_read_file("corosync.conf");
- my $nodelist = PVE::Corosync::nodelist($conf);
- my $totem_cfg = PVE::Corosync::totem_config($conf);
+ die "QDevice already configured!\n"
+ if defined($conf->{main}->{quorum}->{device}) && !$param->{force};
- my $name = $param->{node};
+ my $network = $param->{network};
- # ensure we do not reuse an address, that can crash the whole cluster!
- my $check_duplicate_addr = sub {
- my $addr = shift;
- return if !defined($addr);
-
- while (my ($k, $v) = each %$nodelist) {
- next if $k eq $name; # allows re-adding a node if force is set
- if ($v->{ring0_addr} eq $addr || ($v->{ring1_addr} && $v->{ring1_addr} eq $addr)) {
- die "corosync: address '$addr' already defined by node '$k'\n";
- }
- }
- };
+ my $model = "net";
+ my $algorithm = 'ffsplit';
+ if (scalar($members) & 1) {
+ if ($param->{force}) {
+ $algorithm = 'lms';
+ } else {
+ die "Clusters with an odd node count are not officially supported!\n";
+ }
+ }
- &$check_duplicate_addr($param->{ring0_addr});
- &$check_duplicate_addr($param->{ring1_addr});
+ my $qnetd_addr = $param->{address};
+ my $base_dir = "/etc/corosync/qdevice/net";
+ my $db_dir_qnetd = "/etc/corosync/qnetd/nssdb";
+ my $db_dir_node = "$base_dir/nssdb";
+ my $ca_export_base = "qnetd-cacert.crt";
+ my $ca_export_file = "$db_dir_qnetd/$ca_export_base";
+ my $crq_file_base = "qdevice-net-node.crq";
+ my $p12_file_base = "qdevice-net-node.p12";
+ my $qdevice_certutil = "corosync-qdevice-net-certutil";
+ my $qnetd_certutil= "corosync-qnetd-certutil";
+ my $clustername = $conf->{main}->{totem}->{cluster_name};
+
+ run_command(['ssh-copy-id', '-i', '/root/.ssh/id_rsa', "root\@$qnetd_addr"]);
+
+ if (-d $db_dir_node) {
+ # FIXME: check on all nodes?!
+ if ($param->{force}) {
+ rmtree $db_dir_node;
+ } else {
+ die "QDevice certificate store already initialised, set force to delete!\n";
+ }
+ }
- $param->{ring0_addr} = $name if !$param->{ring0_addr};
+ my $ssh_cmd = ['ssh', '-o', 'BatchMode=yes', '-lroot'];
+ my $scp_cmd = ['scp', '-o', 'BatchMode=yes'];
+
+ print "\nINFO: initializing qnetd server\n";
+ run_command(
+ [@$ssh_cmd, $qnetd_addr, $qnetd_certutil, "-i"],
+ noerr => 1
+ );
+
+ print "\nINFO: copying CA cert and initializing on all nodes\n";
+ run_command([@$scp_cmd, "root\@\[$qnetd_addr\]:$ca_export_file", "/etc/pve/$ca_export_base"]);
+ $foreach_member->(sub {
+ my ($node, $ip) = @_;
+ my $outsub = sub { print "\nnode '$node': " . shift };
+ run_command(
+ [@$ssh_cmd, $ip, $qdevice_certutil, "-i", "-c", "/etc/pve/$ca_export_base"],
+ noerr => 1, outfunc => \&$outsub
+ );
+ });
+ unlink "/etc/pve/$ca_export_base";
+
+ print "\nINFO: generating cert request\n";
+ run_command([$qdevice_certutil, "-r", "-n", $clustername]);
+
+ print "\nINFO: copying exported cert request to qnetd server\n";
+ run_command([@$scp_cmd, "$db_dir_node/$crq_file_base", "root\@\[$qnetd_addr\]:/tmp"]);
+
+ print "\nINFO: sign and export cluster cert\n";
+ run_command([
+ @$ssh_cmd, $qnetd_addr, $qnetd_certutil, "-s", "-c",
+ "/tmp/$crq_file_base", "-n", "$clustername"
+ ]);
+
+ print "\nINFO: copy exported CRT\n";
+ run_command([
+ @$scp_cmd, "root\@\[$qnetd_addr\]:$db_dir_qnetd/cluster-$clustername.crt",
+ "$db_dir_node"
+ ]);
+
+ print "\nINFO: import certificate\n";
+ run_command(["$qdevice_certutil", "-M", "-c", "$db_dir_node/cluster-$clustername.crt"]);
+
+ print "\nINFO: copy and import pk12 cert to all nodes\n";
+ run_command([@$scp_cmd, "$db_dir_node/$p12_file_base", "/etc/pve/"]);
+ $foreach_member->(sub {
+ my ($node, $ip) = @_;
+ my $outsub = sub { print "\nnode '$node': " . shift };
+ run_command([
+ @$ssh_cmd, $ip, "$qdevice_certutil", "-m", "-c",
+ "/etc/pve/$p12_file_base"], outfunc => \&$outsub
+ );
+ });
+ unlink "/etc/pve/$p12_file_base";
- die "corosync: using 'ring1_addr' parameter needs a configured ring 1 interface!\n"
- if $param->{ring1_addr} && !defined($totem_cfg->{interface}->{1});
- die "corosync: ring 1 interface configured but 'ring1_addr' parameter not defined!\n"
- if defined($totem_cfg->{interface}->{1}) && !defined($param->{ring1_addr});
+ my $code = sub {
+ my $conf = PVE::Cluster::cfs_read_file("corosync.conf");
+ my $quorum_section = $conf->{main}->{quorum};
- if (defined(my $res = $nodelist->{$name})) {
- $param->{nodeid} = $res->{nodeid} if !$param->{nodeid};
- $param->{votes} = $res->{quorum_votes} if !defined($param->{votes});
+ die "Qdevice already configured, must be removed before setting up new one!\n"
+ if defined($quorum_section->{device}); # must not be forced!
- if ($res->{quorum_votes} == $param->{votes} &&
- $res->{nodeid} == $param->{nodeid}) {
- print "node $name already defined\n";
- if ($param->{force}) {
- exit (0);
- } else {
- exit (-1);
- }
- } else {
- die "can't add existing node\n";
+ my $qdev_section = {
+ model => $model,
+ "$model" => {
+ tls => 'on',
+ host => $qnetd_addr,
+ algorithm => $algorithm,
}
- } elsif (!$param->{nodeid}) {
- my $nodeid = 1;
-
- while(1) {
- my $found = 0;
- foreach my $v (values %$nodelist) {
- if ($v->{nodeid} eq $nodeid) {
- $found = 1;
- $nodeid++;
- last;
- }
- }
- last if !$found;
- };
-
- $param->{nodeid} = $nodeid;
- }
-
- $param->{votes} = 1 if !defined($param->{votes});
-
- PVE::Cluster::gen_local_dirs($name);
-
- eval { PVE::Cluster::ssh_merge_keys(); };
- warn $@ if $@;
-
- $nodelist->{$name} = {
- ring0_addr => $param->{ring0_addr},
- nodeid => $param->{nodeid},
- name => $name,
};
- $nodelist->{$name}->{ring1_addr} = $param->{ring1_addr} if $param->{ring1_addr};
- $nodelist->{$name}->{quorum_votes} = $param->{votes} if $param->{votes};
+ $qdev_section->{votes} = 1 if $algorithm eq 'ffsplit';
+
+ $quorum_section->{device} = $qdev_section;
- PVE::Corosync::update_nodelist($conf, $nodelist);
+ PVE::Corosync::atomic_write_conf($conf);
};
+ print "\nINFO: add QDevice to cluster configuration\n";
PVE::Cluster::cfs_lock_file('corosync.conf', 10, $code);
die $@ if $@;
- exit (0);
- }});
+ $foreach_member->(sub {
+ my ($node, $ip) = @_;
+ my $outsub = sub { print "\nnode '$node': " . shift };
+ print "\nINFO: start and enable corosync qdevice daemon on node '$node'...\n";
+ run_command([@$ssh_cmd, $ip, 'systemctl', 'start', 'corosync-qdevice'], outfunc => \&$outsub);
+ run_command([@$ssh_cmd, $ip, 'systemctl', 'enable', 'corosync-qdevice'], outfunc => \&$outsub);
+ });
+ run_command(['corosync-cfgtool', '-R']); # do cluster wide config reload
+
+ return undef;
+}});
__PACKAGE__->register_method ({
- name => 'delnode',
- path => 'delnode',
- method => 'PUT',
- description => "Removes a node to the cluster configuration.",
+ name => 'remove_qdevice',
+ path => 'remove_qdevice',
+ method => 'DELETE',
+ description => "Remove a configured QDevice",
parameters => {
- additionalProperties => 0,
- properties => {
- node => {
- type => 'string',
- description => "Hostname or IP of the corosync ring0 address of this node.",
- },
- },
+ additionalProperties => 0,
+ properties => {},
},
returns => { type => 'null' },
code => sub {
my ($param) = @_;
- PVE::Cluster::check_cfs_quorum();
+ die "Node not in a cluster. Aborting.\n"
+ if !PVE::Corosync::check_conf_exists(1);
+
+ my $members = PVE::Cluster::get_members();
+ foreach my $node (sort keys %$members) {
+ die "All nodes must be online! Node $node is offline, aborting.\n"
+ if !$members->{$node}->{online};
+ }
+
+ my $ssh_cmd = ['ssh', '-o', 'BatchMode=yes', '-lroot'];
my $code = sub {
my $conf = PVE::Cluster::cfs_read_file("corosync.conf");
- my $nodelist = PVE::Corosync::nodelist($conf);
-
- my $node;
- my $nodeid;
-
- foreach my $tmp_node (keys %$nodelist) {
- my $d = $nodelist->{$tmp_node};
- my $ring0_addr = $d->{ring0_addr};
- my $ring1_addr = $d->{ring1_addr};
- if (($tmp_node eq $param->{node}) ||
- (defined($ring0_addr) && ($ring0_addr eq $param->{node})) ||
- (defined($ring1_addr) && ($ring1_addr eq $param->{node}))) {
- $node = $tmp_node;
- $nodeid = $d->{nodeid};
- last;
- }
- }
+ my $quorum_section = $conf->{main}->{quorum};
- die "Node/IP: $param->{node} is not a known host of the cluster.\n"
- if !defined($node);
+ die "No QDevice configured!\n" if !defined($quorum_section->{device});
- my $our_nodename = PVE::INotify::nodename();
- die "Cannot delete myself from cluster!\n" if $node eq $our_nodename;
+ delete $quorum_section->{device};
- delete $nodelist->{$node};
+ PVE::Corosync::atomic_write_conf($conf);
- PVE::Corosync::update_nodelist($conf, $nodelist);
-
- run_command(['corosync-cfgtool','-k', $nodeid]) if defined($nodeid);
+ # cleanup qdev state (cert storage)
+ my $qdev_state_dir = "/etc/corosync/qdevice";
+ $foreach_member->(sub {
+ my (undef, $ip) = @_;
+ run_command([@$ssh_cmd, $ip, '--', 'rm', '-rf', $qdev_state_dir]);
+ });
};
PVE::Cluster::cfs_lock_file('corosync.conf', 10, $code);
die $@ if $@;
+ $foreach_member->(sub {
+ my (undef, $ip) = @_;
+ run_command([@$ssh_cmd, $ip, 'systemctl', 'stop', 'corosync-qdevice']);
+ run_command([@$ssh_cmd, $ip, 'systemctl', 'disable', 'corosync-qdevice']);
+ });
+
+ run_command(['corosync-cfgtool', '-R']);
+
+ print "\nRemoved Qdevice.\n";
+
return undef;
- }});
+}});
__PACKAGE__->register_method ({
name => 'add',
type => 'string',
description => "Hostname (or IP) of an existing cluster member."
},
- nodeid => {
- type => 'integer',
- description => "Node id for this node.",
- minimum => 1,
- optional => 1,
- },
+ nodeid => get_standard_option('corosync-nodeid'),
votes => {
type => 'integer',
description => "Number of votes for this node",
description => "Do not throw error if node already exists.",
optional => 1,
},
- ring0_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring0 address of this node.".
- " Defaults to nodes hostname.",
+ ring0_addr => get_standard_option('corosync-ring0-addr'),
+ ring1_addr => get_standard_option('corosync-ring1-addr'),
+ fingerprint => get_standard_option('fingerprint-sha256', {
optional => 1,
- },
- ring1_addr => {
- type => 'string', format => 'address',
- description => "Hostname (or IP) of the corosync ring1 address, this".
- " needs an valid configured ring 1 interface in the cluster.",
+ }),
+ 'use_ssh' => {
+ type => 'boolean',
+ description => "Always use SSH to join, even if peer may do it over API.",
optional => 1,
},
},
my $nodename = PVE::INotify::nodename();
- PVE::Cluster::setup_sshd_config(1);
- PVE::Cluster::setup_rootsshconfig();
- PVE::Cluster::setup_ssh_keys();
-
my $host = $param->{hostname};
+ my $local_ip_address = PVE::Cluster::remote_node_ip($nodename);
- my ($errors, $warnings) = ('', '');
+ PVE::Cluster::assert_joinable($param->{ring0_addr}, $param->{ring1_addr}, $param->{force});
- my $error = sub {
- my ($msg, $suppress) = @_;
+ my $worker = sub {
- if ($suppress) {
- $warnings .= "* $msg\n";
- } else {
- $errors .= "* $msg\n";
- }
- };
-
- if (!$param->{force}) {
-
- if (-f $authfile) {
- &$error("authentication key '$authfile' already exists", $param->{force});
- }
+ if (!$param->{use_ssh}) {
+ print "Please enter superuser (root) password for '$host':\n";
+ my $password = PVE::PTY::read_password("Password for root\@$host: ");
- if (-f $clusterconf) {
- &$error("cluster config '$clusterconf' already exists", $param->{force});
- }
+ delete $param->{use_ssh};
+ $param->{password} = $password;
- my $vmlist = PVE::Cluster::get_vmlist();
- if ($vmlist && $vmlist->{ids} && scalar(keys %{$vmlist->{ids}})) {
- &$error("this host already contains virtual guests", $param->{force});
- }
+ my $local_cluster_lock = "/var/lock/pvecm.lock";
+ PVE::Tools::lock_file($local_cluster_lock, 10, \&PVE::Cluster::join, $param);
- if (system("corosync-quorumtool -l >/dev/null 2>&1") == 0) {
- &$error("corosync is already running, is this node already in a cluster?!", $param->{force});
- }
- }
-
- # check if corosync ring IPs are configured on the current nodes interfaces
- my $check_ip = sub {
- my $ip = shift;
- if (defined($ip)) {
- if (!PVE::JSONSchema::pve_verify_ip($ip, 1)) {
- my $host = $ip;
- eval { $ip = PVE::Network::get_ip_from_hostname($host); };
- if ($@) {
- &$error("cannot use '$host': $@\n") ;
- return;
+ if (my $err = $@) {
+ if (ref($err) eq 'PVE::APIClient::Exception' && defined($err->{code}) && $err->{code} == 501) {
+ $err = "Remote side is not able to use API for Cluster join!\n" .
+ "Pass the 'use_ssh' switch or update the remote side.\n";
}
+ die $err;
}
-
- my $cidr = (Net::IP::ip_is_ipv6($ip)) ? "$ip/128" : "$ip/32";
- my $configured_ips = PVE::Network::get_local_ip_from_cidr($cidr);
-
- &$error("cannot use IP '$ip', it must be configured exactly once on local node!\n")
- if (scalar(@$configured_ips) != 1);
+ return; # all OK, the API join endpoint successfully set us up
}
- };
- &$check_ip($param->{ring0_addr});
- &$check_ip($param->{ring1_addr});
+ # allow fallback to old ssh only join if wished or needed
- warn "warning, ignore the following errors:\n$warnings" if $warnings;
- die "detected the following error(s):\n$errors" if $errors;
+ PVE::Cluster::setup_sshd_config();
+ PVE::Cluster::setup_rootsshconfig();
+ PVE::Cluster::setup_ssh_keys();
- # make sure known_hosts is on local filesystem
- PVE::Cluster::ssh_unmerge_known_hosts();
+ # make sure known_hosts is on local filesystem
+ PVE::Cluster::ssh_unmerge_known_hosts();
- my $cmd = ['ssh-copy-id', '-i', '/root/.ssh/id_rsa', "root\@$host"];
- run_command($cmd, 'outfunc' => sub {}, 'errfunc' => sub {},
- 'errmsg' => "unable to copy ssh ID");
+ my $cmd = ['ssh-copy-id', '-i', '/root/.ssh/id_rsa', "root\@$host"];
+ run_command($cmd, 'outfunc' => sub {}, 'errfunc' => sub {},
+ 'errmsg' => "unable to copy ssh ID");
- $cmd = ['ssh', $host, '-o', 'BatchMode=yes',
- 'pvecm', 'addnode', $nodename, '--force', 1];
-
- push @$cmd, '--nodeid', $param->{nodeid} if $param->{nodeid};
-
- push @$cmd, '--votes', $param->{votes} if defined($param->{votes});
-
- push @$cmd, '--ring0_addr', $param->{ring0_addr} if defined($param->{ring0_addr});
-
- push @$cmd, '--ring1_addr', $param->{ring1_addr} if defined($param->{ring1_addr});
-
- if (system (@$cmd) != 0) {
- my $cmdtxt = join (' ', @$cmd);
- die "unable to add node: command failed ($cmdtxt)\n";
- }
+ $cmd = ['ssh', $host, '-o', 'BatchMode=yes',
+ 'pvecm', 'addnode', $nodename, '--force', 1];
- my $tmpdir = "$libdir/.pvecm_add.tmp.$$";
- mkdir $tmpdir;
+ push @$cmd, '--nodeid', $param->{nodeid} if $param->{nodeid};
+ push @$cmd, '--votes', $param->{votes} if defined($param->{votes});
+ push @$cmd, '--ring0_addr', $param->{ring0_addr} // $local_ip_address;
+ push @$cmd, '--ring1_addr', $param->{ring1_addr} if defined($param->{ring1_addr});
- eval {
- print "copy corosync auth key\n";
- $cmd = ['rsync', '--rsh=ssh -l root -o BatchMode=yes', '-lpgoq',
- "[$host]:$authfile $clusterconf", $tmpdir];
-
- system(@$cmd) == 0 || die "can't rsync data from host '$host'\n";
-
- mkdir "/etc/corosync";
- my $confbase = basename($clusterconf);
-
- $cmd = "cp '$tmpdir/$confbase' '/etc/corosync/$confbase'";
- system($cmd) == 0 || die "can't copy cluster configuration\n";
-
- my $keybase = basename($authfile);
- system ("cp '$tmpdir/$keybase' '$authfile'") == 0 ||
- die "can't copy '$tmpdir/$keybase' to '$authfile'\n";
-
- print "stopping pve-cluster service\n";
-
- system("umount $basedir -f >/dev/null 2>&1");
- system("systemctl stop pve-cluster") == 0 ||
- die "can't stop pve-cluster service\n";
-
- backup_database();
-
- unlink $dbfile;
-
- system("systemctl start pve-cluster") == 0 ||
- die "starting pve-cluster failed\n";
+ if (system (@$cmd) != 0) {
+ my $cmdtxt = join (' ', @$cmd);
+ die "unable to add node: command failed ($cmdtxt)\n";
+ }
- system("systemctl start corosync");
+ my $tmpdir = "$libdir/.pvecm_add.tmp.$$";
+ mkdir $tmpdir;
- # wait for quorum
- my $printqmsg = 1;
- while (!PVE::Cluster::check_cfs_quorum(1)) {
- if ($printqmsg) {
- print "waiting for quorum...";
- STDOUT->flush();
- $printqmsg = 0;
- }
- sleep(1);
- }
- print "OK\n" if !$printqmsg;
+ eval {
+ print "copy corosync auth key\n";
+ $cmd = ['rsync', '--rsh=ssh -l root -o BatchMode=yes', '-lpgoq',
+ "[$host]:$authfile $clusterconf", $tmpdir];
- my $local_ip_address = PVE::Cluster::remote_node_ip($nodename);
+ system(@$cmd) == 0 || die "can't rsync data from host '$host'\n";
- print "generating node certificates\n";
- PVE::Cluster::gen_pve_node_files($nodename, $local_ip_address);
+ my $corosync_conf = PVE::Tools::file_get_contents("$tmpdir/corosync.conf");
+ my $corosync_authkey = PVE::Tools::file_get_contents("$tmpdir/authkey");
- print "merge known_hosts file\n";
- PVE::Cluster::ssh_merge_known_hosts($nodename, $local_ip_address, 1);
+ PVE::Cluster::finish_join($host, $corosync_conf, $corosync_authkey);
+ };
+ my $err = $@;
- print "restart services\n";
- # restart pvedaemon (changed certs)
- system("systemctl restart pvedaemon");
- # restart pveproxy (changed certs)
- system("systemctl restart pveproxy");
+ rmtree $tmpdir;
- print "successfully added node '$nodename' to cluster.\n";
+ die $err if $err;
};
- my $err = $@;
- rmtree $tmpdir;
+ # use a synced worker so we get a nice task log when joining through CLI
+ my $rpcenv = PVE::RPCEnvironment::get();
+ my $authuser = $rpcenv->get_user();
- die $err if $err;
+ $rpcenv->fork_worker('clusterjoin', '', $authuser, $worker);
return undef;
}});
code => sub {
my ($param) = @_;
- PVE::Cluster::setup_rootsshconfig();
-
- PVE::Cluster::gen_pve_vzdump_symlink();
-
- if (!PVE::Cluster::check_cfs_quorum(1)) {
- return undef if $param->{silent};
- die "no quorum - unable to update files\n";
- }
-
- PVE::Cluster::setup_ssh_keys();
-
- my $nodename = PVE::INotify::nodename();
-
- my $local_ip_address = PVE::Cluster::remote_node_ip($nodename);
-
- PVE::Cluster::gen_pve_node_files($nodename, $local_ip_address, $param->{force});
- PVE::Cluster::ssh_merge_keys();
- PVE::Cluster::ssh_merge_known_hosts($nodename, $local_ip_address);
- PVE::Cluster::gen_pve_vzdump_files();
+ # we get called by the pve-cluster.service ExecStartPost and as we do
+ # IO (on /etc/pve) which can hang (uninterruptedly D state). That'd be
+ # no-good for ExecStartPost as it fails the whole service in this case
+ PVE::Tools::run_fork_with_timeout(30, sub {
+ PVE::Cluster::updatecerts_and_ssh($param->@{qw(force silent)});
+ });
return undef;
}});
print "tunnel online\n";
*STDOUT->flush();
- while (my $line = <>) {
+ while (my $line = <STDIN>) {
chomp $line;
last if $line =~ m/^quit$/;
}
our $cmddef = {
keygen => [ __PACKAGE__, 'keygen', ['filename']],
- create => [ __PACKAGE__, 'create', ['clustername']],
+ create => [ 'PVE::API2::ClusterConfig', 'create', ['clustername']],
add => [ __PACKAGE__, 'add', ['hostname']],
- addnode => [ __PACKAGE__, 'addnode', ['node']],
- delnode => [ __PACKAGE__, 'delnode', ['node']],
+ addnode => [ 'PVE::API2::ClusterConfig', 'addnode', ['node']],
+ delnode => [ 'PVE::API2::ClusterConfig', 'delnode', ['node']],
status => [ __PACKAGE__, 'status' ],
nodes => [ __PACKAGE__, 'nodes' ],
expected => [ __PACKAGE__, 'expected', ['expected']],
updatecerts => [ __PACKAGE__, 'updatecerts', []],
mtunnel => [ __PACKAGE__, 'mtunnel', ['extra-args']],
+ qdevice => {
+ setup => [ __PACKAGE__, 'setup_qdevice', ['address']],
+ remove => [ __PACKAGE__, 'remove_qdevice', []],
+ }
};
1;
projects / pve-cluster.git / blobdiff