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
50 'datacenter.cfg' => 1,
51 'replication.cfg' => 1,
53 'corosync.conf.new' => 1,
56 'priv/shadow.cfg' => 1,
58 'priv/token.cfg' => 1,
59 'priv/acme/plugins.cfg' => 1,
64 'ha/crm_commands' => 1,
65 'ha/manager_status' => 1,
66 'ha/resources.cfg' => 1,
73 'sdn/controllers.cfg' => 1,
74 'sdn/subnets.cfg' => 1,
77 'sdn/.running-config' => 1,
78 'virtual-guest/cpu-models.conf' => 1,
81 sub prepare_observed_file_basedirs
{
83 if (!check_cfs_is_mounted
(1)) {
84 warn "pmxcfs isn't mounted (/etc/pve), chickening out..\n";
88 for my $f (sort keys %$observed) {
89 next if $f !~ m!^(.*)/[^/]+$!;
90 my $dir = "$basedir/$1";
91 next if -e
$dir; # can also be a link, so just use -e xist check
92 print "creating directory '$dir' for observerd files\n";
105 sub check_cfs_quorum
{
108 # note: -w filename always return 1 for root, so wee need
109 # to use File::lstat here
110 my $st = File
::stat::lstat("$basedir/local");
111 my $quorate = ($st && (($st->mode & 0200) != 0));
113 die "cluster not ready - no quorum?\n" if !$quorate && !$noerr;
118 sub check_cfs_is_mounted
{
121 my $res = -l
"$basedir/local";
123 die "pve configuration filesystem (pmxcfs) not mounted\n" if !$res && !$noerr;
132 my $ipcc_send_rec = sub {
133 my ($msgid, $data) = @_;
135 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
137 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
142 my $ipcc_send_rec_json = sub {
143 my ($msgid, $data) = @_;
145 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
147 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
149 return decode_json
($res);
152 my $ipcc_get_config = sub {
155 my $bindata = pack "Z*", $path;
156 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_CONFIG
, $bindata);
157 if (!defined($res)) {
159 return undef if $! == ENOENT
;
168 my $ipcc_get_status = sub {
169 my ($name, $nodename) = @_;
171 my $bindata = pack "Z[256]Z[256]", $name, ($nodename || "");
172 return PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_STATUS
, $bindata);
175 my $ipcc_remove_status = sub {
177 # we just omit the data payload, pmxcfs takes this as hint and removes this
178 # key from the status hashtable
179 my $bindata = pack "Z[256]", $name;
180 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
183 my $ipcc_update_status = sub {
184 my ($name, $data) = @_;
186 my $raw = ref($data) ? encode_json
($data) : $data;
188 my $bindata = pack "Z[256]Z*", $name, $raw;
190 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
194 my ($priority, $ident, $tag, $msg) = @_;
196 my $bindata = pack "CCCZ*Z*Z*", $priority, bytes
::length($ident) + 1,
197 bytes
::length($tag) + 1, $ident, $tag, $msg;
199 return &$ipcc_send_rec(CFS_IPC_LOG_CLUSTER_MSG
, $bindata);
202 my $ipcc_get_cluster_log = sub {
203 my ($user, $max) = @_;
205 $max = 0 if !defined($max);
207 my $bindata = pack "VVVVZ*", $max, 0, 0, 0, ($user || "");
208 return &$ipcc_send_rec(CFS_IPC_GET_CLUSTER_LOG
, $bindata);
211 my $ipcc_verify_token = sub {
212 my ($full_token) = @_;
214 my $bindata = pack "Z*", $full_token;
215 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_VERIFY_TOKEN
, $bindata);
218 return 0 if $! == ENOENT
;
228 my $res = &$ipcc_send_rec_json(CFS_IPC_GET_FS_VERSION
);
229 die "no starttime\n" if !$res->{starttime
};
231 if (!$res->{starttime
} || !$versions->{starttime
} ||
232 $res->{starttime
} != $versions->{starttime
}) {
233 #print "detected changed starttime\n";
252 if (!$clinfo->{version
} || $clinfo->{version
} != $versions->{clinfo
}) {
253 #warn "detected new clinfo\n";
254 $clinfo = &$ipcc_send_rec_json(CFS_IPC_GET_CLUSTER_INFO
);
265 if (!$vmlist->{version
} || $vmlist->{version
} != $versions->{vmlist
}) {
266 #warn "detected new vmlist1\n";
267 $vmlist = &$ipcc_send_rec_json(CFS_IPC_GET_GUEST_LIST
);
287 return $clinfo->{nodelist
};
291 my $nodelist = $clinfo->{nodelist
};
293 my $nodename = PVE
::INotify
::nodename
();
295 if (!$nodelist || !$nodelist->{$nodename}) {
296 return [ $nodename ];
299 return [ keys %$nodelist ];
302 # only stored in a in-memory hashtable inside pmxcfs, local data is gone after
303 # a restart (of pmxcfs or the node), peer data is still available then
304 # best used for status data, like running (ceph) services, package versions, ...
305 sub broadcast_node_kv
{
306 my ($key, $data) = @_;
308 if (!defined($data)) {
309 eval { $ipcc_remove_status->("kv/$key") };
311 die "cannot send a reference\n" if ref($data);
312 my $size = length($data);
313 die "data for '$key' too big\n" if $size >= (32 * 1024); # limit from pmxfs
315 eval { $ipcc_update_status->("kv/$key", $data) };
320 # nodename is optional
322 my ($key, $nodename) = @_;
325 my $get_node_data = sub {
327 my $raw = $ipcc_get_status->("kv/$key", $node);
328 $res->{$node} = unpack("Z*", $raw) if $raw;
332 $get_node_data->($nodename);
334 for my $node (get_nodelist
()->@*) {
335 $get_node_data->($node);
342 # property: a config property you want to get, e.g., this is perfect to get
343 # the 'lock' entry of a guest _fast_ (>100 faster than manual parsing here)
344 # vmid: optional, if a valid is passed we only check that one, else return all
345 # NOTE: does *not* searches snapshot and PENDING entries sections!
346 sub get_guest_config_property
{
347 my ($property, $vmid) = @_;
349 die "property is required" if !defined($property);
351 my $bindata = pack "VZ*", $vmid // 0, $property;
352 my $res = $ipcc_send_rec_json->(CFS_IPC_GET_GUEST_CONFIG_PROPERTY
, $bindata);
357 # $data must be a chronological descending ordered array of tasks
358 sub broadcast_tasklist
{
361 # the serialized list may not get bigger than 128 KiB (CFS_MAX_STATUS_SIZE from pmxcfs)
362 # drop older items until we satisfy this constraint
363 my $size = length(encode_json
($data));
364 while ($size >= (32 * 1024)) { # TODO: update to 128 KiB in PVE 8.x
366 $size = length(encode_json
($data));
369 eval { $ipcc_update_status->("tasklist", $data) };
373 my $tasklistcache = {};
378 my $kvstore = $versions->{kvstore
} || {};
380 my $nodelist = get_nodelist
();
383 foreach my $node (@$nodelist) {
384 next if $nodename && ($nodename ne $node);
386 my $ver = $kvstore->{$node}->{tasklist
} if $kvstore->{$node};
387 my $cache = $tasklistcache->{$node};
388 if (!$cache || !$ver || !$cache->{version
} || ($cache->{version
} != $ver)) {
390 if (my $raw = $ipcc_get_status->("tasklist", $node)) {
391 my $json_str = unpack("Z*", $raw);
392 $tasks = decode_json
($json_str);
395 $tasklistcache->{$node} = {
399 } elsif ($cache && $cache->{data
}) {
400 push @$res, $cache->{data
}->@*;
404 syslog
('err', $err) if $err;
411 my ($rrdid, $data) = @_;
414 &$ipcc_update_status("rrd/$rrdid", $data);
421 my $last_rrd_dump = 0;
422 my $last_rrd_data = "";
428 my $diff = $ctime - $last_rrd_dump;
430 return $last_rrd_data;
435 $raw = &$ipcc_send_rec(CFS_IPC_GET_RRD_DUMP
);
447 while ($raw =~ s/^(.*)\n//) {
448 my ($key, @ela) = split(/:/, $1);
450 next if !(scalar(@ela) > 1);
451 $res->{$key} = [ map { $_ eq 'U' ?
undef : $_ } @ela ];
455 $last_rrd_dump = $ctime;
456 $last_rrd_data = $res;
462 # a fast way to read files (avoid fuse overhead)
466 return &$ipcc_get_config($path);
469 sub get_cluster_log
{
470 my ($user, $max) = @_;
472 return &$ipcc_get_cluster_log($user, $max);
476 my ($userid, $token) = @_;
478 return &$ipcc_verify_token("$userid $token");
483 sub cfs_register_file
{
484 my ($filename, $parser, $writer) = @_;
486 $observed->{$filename} || die "unknown file '$filename'";
488 die "file '$filename' already registered" if $file_info->{$filename};
490 $file_info->{$filename} = {
496 my $ccache_read = sub {
497 my ($filename, $parser, $version) = @_;
499 $ccache->{$filename} = {} if !$ccache->{$filename};
501 my $ci = $ccache->{$filename};
503 if (!$ci->{version
} || !$version || $ci->{version
} != $version) {
504 # we always call the parser, even when the file does not exist
505 # (in that case $data is undef)
506 my $data = get_config
($filename);
507 $ci->{data
} = &$parser("/etc/pve/$filename", $data);
508 $ci->{version
} = $version;
511 my $res = ref($ci->{data
}) ? dclone
($ci->{data
}) : $ci->{data
};
516 sub cfs_file_version
{
521 if ($filename =~ m!^nodes/[^/]+/(openvz|lxc|qemu-server)/(\d+)\.conf$!) {
522 my ($type, $vmid) = ($1, $2);
523 if ($vmlist && $vmlist->{ids
} && $vmlist->{ids
}->{$vmid}) {
524 $version = $vmlist->{ids
}->{$vmid}->{version
};
526 $infotag = "/$type/";
528 $infotag = $filename;
529 $version = $versions->{$filename};
532 my $info = $file_info->{$infotag} ||
533 die "unknown file type '$filename'\n";
535 return wantarray ?
($version, $info) : $version;
541 my ($version, $info) = cfs_file_version
($filename);
542 my $parser = $info->{parser
};
544 return &$ccache_read($filename, $parser, $version);
548 my ($filename, $data) = @_;
550 my ($version, $info) = cfs_file_version
($filename);
552 my $writer = $info->{writer
} || die "no writer defined";
554 my $fsname = "/etc/pve/$filename";
556 my $raw = &$writer($fsname, $data);
558 if (my $ci = $ccache->{$filename}) {
559 $ci->{version
} = undef;
562 PVE
::Tools
::file_set_contents
($fsname, $raw);
566 my ($lockid, $timeout, $code, @param) = @_;
568 my $prev_alarm = alarm(0); # suspend outer alarm early
573 # this timeout is for acquire the lock
574 $timeout = 10 if !$timeout;
576 my $filename = "$lockdir/$lockid";
584 die "pve cluster filesystem not online.\n";
587 my $timeout_err = sub { die "got lock request timeout\n"; };
588 local $SIG{ALRM
} = $timeout_err;
592 $got_lock = mkdir($filename);
593 $timeout = alarm(0) - 1; # we'll sleep for 1s, see down below
597 $timeout_err->() if $timeout <= 0;
599 print STDERR
"trying to acquire cfs lock '$lockid' ...\n";
600 utime (0, 0, $filename); # cfs unlock request
604 # fixed command timeout: cfs locks have a timeout of 120
605 # using 60 gives us another 60 seconds to abort the task
606 local $SIG{ALRM
} = sub { die "'$lockid'-locked command timed out - aborting\n"; };
609 cfs_update
(); # make sure we read latest versions inside code()
611 $is_code_err = 1; # allows to differ between locking and actual-work errors
613 $res = &$code(@param);
620 $err = "no quorum!\n" if !$got_lock && !check_cfs_quorum
(1);
622 rmdir $filename if $got_lock; # if we held the lock always unlock again
627 if (ref($err) eq 'PVE::Exception' || $is_code_err) {
628 # re-raise defined exceptions
631 # add lock info for plain errors comming from the locking itself
632 $@ = "cfs-lock '$lockid' error: $err";
643 my ($filename, $timeout, $code, @param) = @_;
645 my $info = $observed->{$filename} || die "unknown file '$filename'";
647 my $lockid = "file-$filename";
648 $lockid =~ s/[.\/]/_
/g
;
650 &$cfs_lock($lockid, $timeout, $code, @param);
653 sub cfs_lock_storage
{
654 my ($storeid, $timeout, $code, @param) = @_;
656 my $lockid = "storage-$storeid";
658 &$cfs_lock($lockid, $timeout, $code, @param);
661 sub cfs_lock_domain
{
662 my ($domainname, $timeout, $code, @param) = @_;
664 my $lockid = "domain-$domainname";
666 &$cfs_lock($lockid, $timeout, $code, @param);
670 my ($account, $timeout, $code, @param) = @_;
672 my $lockid = "acme-$account";
674 &$cfs_lock($lockid, $timeout, $code, @param);
677 sub cfs_lock_authkey
{
678 my ($timeout, $code, @param) = @_;
680 $cfs_lock->('authkey', $timeout, $code, @param);
683 sub cfs_lock_firewall
{
684 my ($scope, $timeout, $code, @param) = @_;
686 my $lockid = "firewall-$scope";
688 $cfs_lock->($lockid, $timeout, $code, @param);
706 my ($priority, $ident, $msg) = @_;
708 if (my $tmp = $log_levels->{$priority}) {
712 die "need numeric log priority" if $priority !~ /^\d+$/;
714 my $tag = PVE
::SafeSyslog
::tag
();
716 $msg = "empty message" if !$msg;
718 $ident = "" if !$ident;
719 $ident = encode
("ascii", $ident,
720 sub { sprintf "\\u%04x", shift });
722 my $ascii = encode
("ascii", $msg, sub { sprintf "\\u%04x", shift });
725 syslog
($priority, "<%s> %s", $ident, $ascii);
727 syslog
($priority, "%s", $ascii);
730 eval { &$ipcc_log($priority, $ident, $tag, $ascii); };
732 syslog
("err", "writing cluster log failed: $@") if $@;
735 sub check_vmid_unused
{
736 my ($vmid, $noerr) = @_;
738 my $vmlist = get_vmlist
();
740 my $d = $vmlist->{ids
}->{$vmid};
741 return 1 if !defined($d);
743 return undef if $noerr;
745 my $vmtypestr = $d->{type
} eq 'qemu' ?
'VM' : 'CT';
746 die "$vmtypestr $vmid already exists on node '$d->{node}'\n";
749 sub check_node_exists
{
750 my ($nodename, $noerr) = @_;
752 my $nodelist = $clinfo->{nodelist
};
753 return 1 if $nodelist && $nodelist->{$nodename};
755 return undef if $noerr;
757 die "no such cluster node '$nodename'\n";
760 # this is also used to get the IP of the local node
762 my ($nodename, $noerr) = @_;
764 my $nodelist = $clinfo->{nodelist
};
765 if ($nodelist && $nodelist->{$nodename}) {
766 if (my $ip = $nodelist->{$nodename}->{ip
}) {
767 return $ip if !wantarray;
768 my $family = $nodelist->{$nodename}->{address_family
};
770 $nodelist->{$nodename}->{address_family
} =
772 PVE
::Tools
::get_host_address_family
($ip);
774 return wantarray ?
($ip, $family) : $ip;
778 # fallback: try to get IP by other means
779 return PVE
::Network
::get_ip_from_hostname
($nodename, $noerr);
782 sub get_node_fingerprint
{
785 my $cert_path = "/etc/pve/nodes/$node/pve-ssl.pem";
786 my $custom_cert_path = "/etc/pve/nodes/$node/pveproxy-ssl.pem";
788 $cert_path = $custom_cert_path if -f
$custom_cert_path;
790 return PVE
::Certificate
::get_certificate_fingerprint
($cert_path);
793 # bash completion helpers
795 sub complete_next_vmid
{
797 my $vmlist = get_vmlist
() || {};
798 my $idlist = $vmlist->{ids
} || {};
800 for (my $i = 100; $i < 10000; $i++) {
801 return [$i] if !defined($idlist->{$i});
809 my $vmlist = get_vmlist
();
810 my $ids = $vmlist->{ids
} || {};
812 return [ keys %$ids ];
815 sub complete_local_vmid
{
817 my $vmlist = get_vmlist
();
818 my $ids = $vmlist->{ids
} || {};
820 my $nodename = PVE
::INotify
::nodename
();
823 foreach my $vmid (keys %$ids) {
824 my $d = $ids->{$vmid};
825 next if !$d->{node
} || $d->{node
} ne $nodename;
832 sub complete_migration_target
{
836 my $nodename = PVE
::INotify
::nodename
();
838 my $nodelist = get_nodelist
();
839 foreach my $node (@$nodelist) {
840 next if $node eq $nodename;
848 # NOTE: filesystem must be offline here, no DB changes allowed
849 sub cfs_backup_database
{
853 my $backup_fn = "$dbbackupdir/config-$ctime.sql.gz";
855 print "backup old database to '$backup_fn'\n";
857 my $cmd = [ ['sqlite3', $dbfile, '.dump'], ['gzip', '-', \
">${backup_fn}"] ];
858 run_command
($cmd, 'errmsg' => "cannot backup old database\n");
860 my $maxfiles = 10; # purge older backup
861 my $backups = [ sort { $b cmp $a } <$dbbackupdir/config-*.sql
.gz
> ];
863 if ((my $count = scalar(@$backups)) > $maxfiles) {
864 foreach my $f (@$backups[$maxfiles..$count-1]) {
865 next if $f !~ m/^(\S+)$/; # untaint
866 print "delete old backup '$1'\n";