8 use File
::Path
qw(make_path);
13 use Storable
qw(dclone);
21 use PVE
::Tools
qw(run_command);
23 use PVE
::Cluster
::IPCConst
;
33 # x509 certificate utils
35 my $basedir = "/etc/pve";
36 my $authdir = "$basedir/priv";
37 my $lockdir = "/etc/pve/priv/lock";
39 # cfs and corosync files
40 my $dbfile = "/var/lib/pve-cluster/config.db";
41 my $dbbackupdir = "/var/lib/pve-cluster/backup";
43 # this is just a readonly copy, the relevant one is in status.c from pmxcfs
44 # observed files are the one we can get directly through IPCC, they are cached
45 # using a computed version and only those can be used by the cfs_*_file methods
49 'datacenter.cfg' => 1,
50 'replication.cfg' => 1,
52 'corosync.conf.new' => 1,
55 'priv/shadow.cfg' => 1,
57 'priv/token.cfg' => 1,
58 'priv/acme/plugins.cfg' => 1,
63 'ha/crm_commands' => 1,
64 'ha/manager_status' => 1,
65 'ha/resources.cfg' => 1,
72 'sdn/controllers.cfg' => 1,
73 'sdn/subnets.cfg' => 1,
76 'virtual-guest/cpu-models.conf' => 1,
79 sub prepare_observed_file_basedirs
{
81 if (!check_cfs_is_mounted
(1)) {
82 warn "pmxcfs isn't mounted (/etc/pve), chickening out..\n";
86 for my $f (sort keys %$observed) {
87 next if $f !~ m!^(.*)/[^/]+$!;
88 my $dir = "$basedir/$1";
89 next if -e
$dir; # can also be a link, so just use -e xist check
90 print "creating directory '$dir' for observerd files\n";
103 sub check_cfs_quorum
{
106 # note: -w filename always return 1 for root, so wee need
107 # to use File::lstat here
108 my $st = File
::stat::lstat("$basedir/local");
109 my $quorate = ($st && (($st->mode & 0200) != 0));
111 die "cluster not ready - no quorum?\n" if !$quorate && !$noerr;
116 sub check_cfs_is_mounted
{
119 my $res = -l
"$basedir/local";
121 die "pve configuration filesystem not mounted\n"
131 my $ipcc_send_rec = sub {
132 my ($msgid, $data) = @_;
134 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
136 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
141 my $ipcc_send_rec_json = sub {
142 my ($msgid, $data) = @_;
144 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
146 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
148 return decode_json
($res);
151 my $ipcc_get_config = sub {
154 my $bindata = pack "Z*", $path;
155 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_CONFIG
, $bindata);
156 if (!defined($res)) {
158 return undef if $! == ENOENT
;
167 my $ipcc_get_status = sub {
168 my ($name, $nodename) = @_;
170 my $bindata = pack "Z[256]Z[256]", $name, ($nodename || "");
171 return PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_STATUS
, $bindata);
174 my $ipcc_remove_status = sub {
176 # we just omit the data payload, pmxcfs takes this as hint and removes this
177 # key from the status hashtable
178 my $bindata = pack "Z[256]", $name;
179 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
182 my $ipcc_update_status = sub {
183 my ($name, $data) = @_;
185 my $raw = ref($data) ? encode_json
($data) : $data;
187 my $bindata = pack "Z[256]Z*", $name, $raw;
189 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
193 my ($priority, $ident, $tag, $msg) = @_;
195 my $bindata = pack "CCCZ*Z*Z*", $priority, bytes
::length($ident) + 1,
196 bytes
::length($tag) + 1, $ident, $tag, $msg;
198 return &$ipcc_send_rec(CFS_IPC_LOG_CLUSTER_MSG
, $bindata);
201 my $ipcc_get_cluster_log = sub {
202 my ($user, $max) = @_;
204 $max = 0 if !defined($max);
206 my $bindata = pack "VVVVZ*", $max, 0, 0, 0, ($user || "");
207 return &$ipcc_send_rec(CFS_IPC_GET_CLUSTER_LOG
, $bindata);
210 my $ipcc_verify_token = sub {
211 my ($full_token) = @_;
213 my $bindata = pack "Z*", $full_token;
214 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_VERIFY_TOKEN
, $bindata);
217 return 0 if $! == ENOENT
;
227 my $res = &$ipcc_send_rec_json(CFS_IPC_GET_FS_VERSION
);
228 die "no starttime\n" if !$res->{starttime
};
230 if (!$res->{starttime
} || !$versions->{starttime
} ||
231 $res->{starttime
} != $versions->{starttime
}) {
232 #print "detected changed starttime\n";
251 if (!$clinfo->{version
} || $clinfo->{version
} != $versions->{clinfo
}) {
252 #warn "detected new clinfo\n";
253 $clinfo = &$ipcc_send_rec_json(CFS_IPC_GET_CLUSTER_INFO
);
264 if (!$vmlist->{version
} || $vmlist->{version
} != $versions->{vmlist
}) {
265 #warn "detected new vmlist1\n";
266 $vmlist = &$ipcc_send_rec_json(CFS_IPC_GET_GUEST_LIST
);
286 return $clinfo->{nodelist
};
290 my $nodelist = $clinfo->{nodelist
};
292 my $nodename = PVE
::INotify
::nodename
();
294 if (!$nodelist || !$nodelist->{$nodename}) {
295 return [ $nodename ];
298 return [ keys %$nodelist ];
301 # only stored in a in-memory hashtable inside pmxcfs, local data is gone after
302 # a restart (of pmxcfs or the node), peer data is still available then
303 # best used for status data, like running (ceph) services, package versions, ...
304 sub broadcast_node_kv
{
305 my ($key, $data) = @_;
307 if (!defined($data)) {
309 $ipcc_remove_status->("kv/$key");
312 die "cannot send a reference\n" if ref($data);
313 my $size = length($data);
314 die "data for '$key' too big\n" if $size >= (32 * 1024); # limit from pmxfs
317 $ipcc_update_status->("kv/$key", $data);
324 # nodename is optional
326 my ($key, $nodename) = @_;
329 my $get_node_data = sub {
331 my $raw = $ipcc_get_status->("kv/$key", $node);
332 $res->{$node} = unpack("Z*", $raw) if $raw;
336 $get_node_data->($nodename);
338 my $nodelist = get_nodelist
();
340 foreach my $node (@$nodelist) {
341 $get_node_data->($node);
348 # property: a config property you want to get, e.g., this is perfect to get
349 # the 'lock' entry of a guest _fast_ (>100 faster than manual parsing here)
350 # vmid: optipnal, if a valid is passed we only check that one, else return all
351 # NOTE: does *not* searches snapshot and PENDING entries sections!
352 sub get_guest_config_property
{
353 my ($property, $vmid) = @_;
355 die "property is required" if !defined($property);
357 my $bindata = pack "VZ*", $vmid // 0, $property;
358 my $res = $ipcc_send_rec_json->(CFS_IPC_GET_GUEST_CONFIG_PROPERTY
, $bindata);
363 # $data must be a chronological descending ordered array of tasks
364 sub broadcast_tasklist
{
367 # the serialized list may not get bigger than 32kb (CFS_MAX_STATUS_SIZE
368 # from pmxcfs) - drop older items until we satisfy this constraint
369 my $size = length(encode_json
($data));
370 while ($size >= (32 * 1024)) {
372 $size = length(encode_json
($data));
376 &$ipcc_update_status("tasklist", $data);
382 my $tasklistcache = {};
387 my $kvstore = $versions->{kvstore
} || {};
389 my $nodelist = get_nodelist
();
392 foreach my $node (@$nodelist) {
393 next if $nodename && ($nodename ne $node);
395 my $ver = $kvstore->{$node}->{tasklist
} if $kvstore->{$node};
396 my $cd = $tasklistcache->{$node};
397 if (!$cd || !$ver || !$cd->{version
} ||
398 ($cd->{version
} != $ver)) {
399 my $raw = &$ipcc_get_status("tasklist", $node) || '[]';
400 my $data = decode_json
($raw);
402 $cd = $tasklistcache->{$node} = {
406 } elsif ($cd && $cd->{data
}) {
407 push @$res, @{$cd->{data
}};
411 syslog
('err', $err) if $err;
418 my ($rrdid, $data) = @_;
421 &$ipcc_update_status("rrd/$rrdid", $data);
428 my $last_rrd_dump = 0;
429 my $last_rrd_data = "";
435 my $diff = $ctime - $last_rrd_dump;
437 return $last_rrd_data;
442 $raw = &$ipcc_send_rec(CFS_IPC_GET_RRD_DUMP
);
454 while ($raw =~ s/^(.*)\n//) {
455 my ($key, @ela) = split(/:/, $1);
457 next if !(scalar(@ela) > 1);
458 $res->{$key} = [ map { $_ eq 'U' ?
undef : $_ } @ela ];
462 $last_rrd_dump = $ctime;
463 $last_rrd_data = $res;
469 # a fast way to read files (avoid fuse overhead)
473 return &$ipcc_get_config($path);
476 sub get_cluster_log
{
477 my ($user, $max) = @_;
479 return &$ipcc_get_cluster_log($user, $max);
483 my ($userid, $token) = @_;
485 return &$ipcc_verify_token("$userid $token");
490 sub cfs_register_file
{
491 my ($filename, $parser, $writer) = @_;
493 $observed->{$filename} || die "unknown file '$filename'";
495 die "file '$filename' already registered" if $file_info->{$filename};
497 $file_info->{$filename} = {
503 my $ccache_read = sub {
504 my ($filename, $parser, $version) = @_;
506 $ccache->{$filename} = {} if !$ccache->{$filename};
508 my $ci = $ccache->{$filename};
510 if (!$ci->{version
} || !$version || $ci->{version
} != $version) {
511 # we always call the parser, even when the file does not exist
512 # (in that case $data is undef)
513 my $data = get_config
($filename);
514 $ci->{data
} = &$parser("/etc/pve/$filename", $data);
515 $ci->{version
} = $version;
518 my $res = ref($ci->{data
}) ? dclone
($ci->{data
}) : $ci->{data
};
523 sub cfs_file_version
{
528 if ($filename =~ m!^nodes/[^/]+/(openvz|lxc|qemu-server)/(\d+)\.conf$!) {
529 my ($type, $vmid) = ($1, $2);
530 if ($vmlist && $vmlist->{ids
} && $vmlist->{ids
}->{$vmid}) {
531 $version = $vmlist->{ids
}->{$vmid}->{version
};
533 $infotag = "/$type/";
535 $infotag = $filename;
536 $version = $versions->{$filename};
539 my $info = $file_info->{$infotag} ||
540 die "unknown file type '$filename'\n";
542 return wantarray ?
($version, $info) : $version;
548 my ($version, $info) = cfs_file_version
($filename);
549 my $parser = $info->{parser
};
551 return &$ccache_read($filename, $parser, $version);
555 my ($filename, $data) = @_;
557 my ($version, $info) = cfs_file_version
($filename);
559 my $writer = $info->{writer
} || die "no writer defined";
561 my $fsname = "/etc/pve/$filename";
563 my $raw = &$writer($fsname, $data);
565 if (my $ci = $ccache->{$filename}) {
566 $ci->{version
} = undef;
569 PVE
::Tools
::file_set_contents
($fsname, $raw);
573 my ($lockid, $timeout, $code, @param) = @_;
575 my $prev_alarm = alarm(0); # suspend outer alarm early
580 # this timeout is for acquire the lock
581 $timeout = 10 if !$timeout;
583 my $filename = "$lockdir/$lockid";
590 die "pve cluster filesystem not online.\n";
593 my $timeout_err = sub { die "got lock request timeout\n"; };
594 local $SIG{ALRM
} = $timeout_err;
598 $got_lock = mkdir($filename);
599 $timeout = alarm(0) - 1; # we'll sleep for 1s, see down below
603 $timeout_err->() if $timeout <= 0;
605 print STDERR
"trying to acquire cfs lock '$lockid' ...\n";
606 utime (0, 0, $filename); # cfs unlock request
610 # fixed command timeout: cfs locks have a timeout of 120
611 # using 60 gives us another 60 seconds to abort the task
612 local $SIG{ALRM
} = sub { die "got lock timeout - aborting command\n"; };
615 cfs_update
(); # make sure we read latest versions inside code()
617 $res = &$code(@param);
624 $err = "no quorum!\n" if !$got_lock && !check_cfs_quorum
(1);
626 rmdir $filename if $got_lock; # if we held the lock always unlock again
631 if (ref($err) eq 'PVE::Exception') {
632 # re-raise defined exceptions
635 # add lock info for plain errors
636 $@ = "error during cfs-locked '$lockid' operation: $err";
647 my ($filename, $timeout, $code, @param) = @_;
649 my $info = $observed->{$filename} || die "unknown file '$filename'";
651 my $lockid = "file-$filename";
652 $lockid =~ s/[.\/]/_
/g
;
654 &$cfs_lock($lockid, $timeout, $code, @param);
657 sub cfs_lock_storage
{
658 my ($storeid, $timeout, $code, @param) = @_;
660 my $lockid = "storage-$storeid";
662 &$cfs_lock($lockid, $timeout, $code, @param);
665 sub cfs_lock_domain
{
666 my ($domainname, $timeout, $code, @param) = @_;
668 my $lockid = "domain-$domainname";
670 &$cfs_lock($lockid, $timeout, $code, @param);
674 my ($account, $timeout, $code, @param) = @_;
676 my $lockid = "acme-$account";
678 &$cfs_lock($lockid, $timeout, $code, @param);
681 sub cfs_lock_authkey
{
682 my ($timeout, $code, @param) = @_;
684 $cfs_lock->('authkey', $timeout, $code, @param);
687 sub cfs_lock_firewall
{
688 my ($scope, $timeout, $code, @param) = @_;
690 my $lockid = "firewall-$scope";
692 $cfs_lock->($lockid, $timeout, $code, @param);
710 my ($priority, $ident, $msg) = @_;
712 if (my $tmp = $log_levels->{$priority}) {
716 die "need numeric log priority" if $priority !~ /^\d+$/;
718 my $tag = PVE
::SafeSyslog
::tag
();
720 $msg = "empty message" if !$msg;
722 $ident = "" if !$ident;
723 $ident = encode
("ascii", $ident,
724 sub { sprintf "\\u%04x", shift });
726 my $ascii = encode
("ascii", $msg, sub { sprintf "\\u%04x", shift });
729 syslog
($priority, "<%s> %s", $ident, $ascii);
731 syslog
($priority, "%s", $ascii);
734 eval { &$ipcc_log($priority, $ident, $tag, $ascii); };
736 syslog
("err", "writing cluster log failed: $@") if $@;
739 sub check_vmid_unused
{
740 my ($vmid, $noerr) = @_;
742 my $vmlist = get_vmlist
();
744 my $d = $vmlist->{ids
}->{$vmid};
745 return 1 if !defined($d);
747 return undef if $noerr;
749 my $vmtypestr = $d->{type
} eq 'qemu' ?
'VM' : 'CT';
750 die "$vmtypestr $vmid already exists on node '$d->{node}'\n";
753 sub check_node_exists
{
754 my ($nodename, $noerr) = @_;
756 my $nodelist = $clinfo->{nodelist
};
757 return 1 if $nodelist && $nodelist->{$nodename};
759 return undef if $noerr;
761 die "no such cluster node '$nodename'\n";
764 # this is also used to get the IP of the local node
766 my ($nodename, $noerr) = @_;
768 my $nodelist = $clinfo->{nodelist
};
769 if ($nodelist && $nodelist->{$nodename}) {
770 if (my $ip = $nodelist->{$nodename}->{ip
}) {
771 return $ip if !wantarray;
772 my $family = $nodelist->{$nodename}->{address_family
};
774 $nodelist->{$nodename}->{address_family
} =
776 PVE
::Tools
::get_host_address_family
($ip);
778 return wantarray ?
($ip, $family) : $ip;
782 # fallback: try to get IP by other means
783 return PVE
::Network
::get_ip_from_hostname
($nodename, $noerr);
786 sub get_node_fingerprint
{
789 my $cert_path = "/etc/pve/nodes/$node/pve-ssl.pem";
790 my $custom_cert_path = "/etc/pve/nodes/$node/pveproxy-ssl.pem";
792 $cert_path = $custom_cert_path if -f
$custom_cert_path;
794 return PVE
::Certificate
::get_certificate_fingerprint
($cert_path);
797 # bash completion helpers
799 sub complete_next_vmid
{
801 my $vmlist = get_vmlist
() || {};
802 my $idlist = $vmlist->{ids
} || {};
804 for (my $i = 100; $i < 10000; $i++) {
805 return [$i] if !defined($idlist->{$i});
813 my $vmlist = get_vmlist
();
814 my $ids = $vmlist->{ids
} || {};
816 return [ keys %$ids ];
819 sub complete_local_vmid
{
821 my $vmlist = get_vmlist
();
822 my $ids = $vmlist->{ids
} || {};
824 my $nodename = PVE
::INotify
::nodename
();
827 foreach my $vmid (keys %$ids) {
828 my $d = $ids->{$vmid};
829 next if !$d->{node
} || $d->{node
} ne $nodename;
836 sub complete_migration_target
{
840 my $nodename = PVE
::INotify
::nodename
();
842 my $nodelist = get_nodelist
();
843 foreach my $node (@$nodelist) {
844 next if $node eq $nodename;
852 # NOTE: filesystem must be offline here, no DB changes allowed
853 sub cfs_backup_database
{
857 my $backup_fn = "$dbbackupdir/config-$ctime.sql.gz";
859 print "backup old database to '$backup_fn'\n";
861 my $cmd = [ ['sqlite3', $dbfile, '.dump'], ['gzip', '-', \
">${backup_fn}"] ];
862 run_command
($cmd, 'errmsg' => "cannot backup old database\n");
864 my $maxfiles = 10; # purge older backup
865 my $backups = [ sort { $b cmp $a } <$dbbackupdir/config-*.sql
.gz
> ];
867 if ((my $count = scalar(@$backups)) > $maxfiles) {
868 foreach my $f (@$backups[$maxfiles..$count-1]) {
869 next if $f !~ m/^(\S+)$/; # untaint
870 print "delete old backup '$1'\n";