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
51 'datacenter.cfg' => 1,
52 'replication.cfg' => 1,
54 'corosync.conf.new' => 1,
55 'firewall/cluster.fw' => 1,
58 'notifications.cfg' => 1,
59 'priv/notifications.cfg' => 1,
60 'priv/shadow.cfg' => 1,
62 'priv/token.cfg' => 1,
63 'priv/acme/plugins.cfg' => 1,
68 'ha/crm_commands' => 1,
69 'ha/manager_status' => 1,
70 'ha/resources.cfg' => 1,
77 'sdn/controllers.cfg' => 1,
78 'sdn/subnets.cfg' => 1,
81 'sdn/.running-config' => 1,
82 'virtual-guest/cpu-models.conf' => 1,
83 'mapping/pci.cfg' => 1,
84 'mapping/usb.cfg' => 1,
87 sub prepare_observed_file_basedirs
{
89 if (!check_cfs_is_mounted
(1)) {
90 warn "pmxcfs isn't mounted (/etc/pve), chickening out..\n";
94 for my $f (sort keys %$observed) {
95 next if $f !~ m!^(.*)/[^/]+$!;
96 my $dir = "$basedir/$1";
97 next if -e
$dir; # can also be a link, so just use -e xist check
98 print "creating directory '$dir' for observed files\n";
111 sub check_cfs_quorum
{
114 # note: -w filename always return 1 for root, so wee need
115 # to use File::lstat here
116 my $st = File
::stat::lstat("$basedir/local");
117 my $quorate = ($st && (($st->mode & 0200) != 0));
119 die "cluster not ready - no quorum?\n" if !$quorate && !$noerr;
124 sub check_cfs_is_mounted
{
127 my $res = -l
"$basedir/local";
129 die "pve configuration filesystem (pmxcfs) not mounted\n" if !$res && !$noerr;
138 my $ipcc_send_rec = sub {
139 my ($msgid, $data) = @_;
141 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
143 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
148 my $ipcc_send_rec_json = sub {
149 my ($msgid, $data) = @_;
151 my $res = PVE
::IPCC
::ipcc_send_rec
($msgid, $data);
153 die "ipcc_send_rec[$msgid] failed: $!\n" if !defined($res) && ($! != 0);
155 return decode_json
($res);
158 my $ipcc_get_config = sub {
161 my $bindata = pack "Z*", $path;
162 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_CONFIG
, $bindata);
163 if (!defined($res)) {
165 return undef if $! == ENOENT
;
174 my $ipcc_get_status = sub {
175 my ($name, $nodename) = @_;
177 my $bindata = pack "Z[256]Z[256]", $name, ($nodename || "");
178 return PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_GET_STATUS
, $bindata);
181 my $ipcc_remove_status = sub {
183 # we just omit the data payload, pmxcfs takes this as hint and removes this
184 # key from the status hashtable
185 my $bindata = pack "Z[256]", $name;
186 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
189 my $ipcc_update_status = sub {
190 my ($name, $data) = @_;
192 my $raw = ref($data) ? encode_json
($data) : $data;
194 my $bindata = pack "Z[256]Z*", $name, $raw;
196 return &$ipcc_send_rec(CFS_IPC_SET_STATUS
, $bindata);
200 my ($priority, $ident, $tag, $msg) = @_;
202 my $bindata = pack "CCCZ*Z*Z*", $priority, bytes
::length($ident) + 1,
203 bytes
::length($tag) + 1, $ident, $tag, $msg;
205 return &$ipcc_send_rec(CFS_IPC_LOG_CLUSTER_MSG
, $bindata);
208 my $ipcc_get_cluster_log = sub {
209 my ($user, $max) = @_;
211 $max = 0 if !defined($max);
213 my $bindata = pack "VVVVZ*", $max, 0, 0, 0, ($user || "");
214 return &$ipcc_send_rec(CFS_IPC_GET_CLUSTER_LOG
, $bindata);
217 my $ipcc_verify_token = sub {
218 my ($full_token) = @_;
220 my $bindata = pack "Z*", $full_token;
221 my $res = PVE
::IPCC
::ipcc_send_rec
(CFS_IPC_VERIFY_TOKEN
, $bindata);
224 return 0 if $! == ENOENT
;
234 my $res = &$ipcc_send_rec_json(CFS_IPC_GET_FS_VERSION
);
235 die "no starttime\n" if !$res->{starttime
};
237 if (!$res->{starttime
} || !$versions->{starttime
} ||
238 $res->{starttime
} != $versions->{starttime
}) {
239 #print "detected changed starttime\n";
258 if (!$clinfo->{version
} || $clinfo->{version
} != $versions->{clinfo
}) {
259 #warn "detected new clinfo\n";
260 $clinfo = &$ipcc_send_rec_json(CFS_IPC_GET_CLUSTER_INFO
);
271 if (!$vmlist->{version
} || $vmlist->{version
} != $versions->{vmlist
}) {
272 #warn "detected new vmlist1\n";
273 $vmlist = &$ipcc_send_rec_json(CFS_IPC_GET_GUEST_LIST
);
293 return $clinfo->{nodelist
};
297 my $nodelist = $clinfo->{nodelist
};
299 my $nodename = PVE
::INotify
::nodename
();
301 if (!$nodelist || !$nodelist->{$nodename}) {
302 return [ $nodename ];
305 return [ keys %$nodelist ];
308 # only stored in a in-memory hashtable inside pmxcfs, local data is gone after
309 # a restart (of pmxcfs or the node), peer data is still available then
310 # best used for status data, like running (ceph) services, package versions, ...
311 sub broadcast_node_kv
{
312 my ($key, $data) = @_;
314 if (!defined($data)) {
315 eval { $ipcc_remove_status->("kv/$key") };
317 die "cannot send a reference\n" if ref($data);
318 my $size = length($data);
319 die "data for '$key' too big\n" if $size >= (32 * 1024); # limit from pmxfs
321 eval { $ipcc_update_status->("kv/$key", $data) };
326 # nodename is optional
328 my ($key, $nodename) = @_;
331 my $get_node_data = sub {
333 my $raw = $ipcc_get_status->("kv/$key", $node);
334 $res->{$node} = unpack("Z*", $raw) if $raw;
338 $get_node_data->($nodename);
340 for my $node (get_nodelist
()->@*) {
341 $get_node_data->($node);
348 # properties: an array-ref of config properties you want to get, e.g., this
349 # is perfect to get multiple properties of a guest _fast_
350 # (>100 faster than manual parsing here)
351 # vmid: optional, if a valid is passed we only check that one, else return all
352 # NOTE: does *not* searches snapshot and PENDING entries sections!
353 # NOTE: returns the guest config lines (excluding trailing whitespace) as is,
354 # so for non-trivial properties, checking the validity must be done
355 # NOTE: no permission check is done, that is the responsibilty of the caller
356 sub get_guest_config_properties
{
357 my ($properties, $vmid) = @_;
359 die "properties required" if !defined($properties);
361 my $num_props = scalar(@$properties);
362 die "only up to 255 properties supported" if $num_props > 255;
363 my $bindata = pack "VC", $vmid // 0, $num_props;
364 for my $property (@$properties) {
365 $bindata .= pack "Z*", $property;
367 my $res = $ipcc_send_rec_json->(CFS_IPC_GET_GUEST_CONFIG_PROPERTIES
, $bindata);
372 # property: a config property you want to get, e.g., this is perfect to get
373 # the 'lock' entry of a guest _fast_ (>100 faster than manual parsing here)
374 # vmid: optional, if a valid is passed we only check that one, else return all
375 # NOTE: does *not* searches snapshot and PENDING entries sections!
376 # NOTE: returns the guest config lines (excluding trailing whitespace) as is,
377 # so for non-trivial properties, checking the validity must be done
378 # NOTE: no permission check is done, that is the responsibilty of the caller
379 sub get_guest_config_property
{
380 my ($property, $vmid) = @_;
382 die "property is required" if !defined($property);
384 my $bindata = pack "VZ*", $vmid // 0, $property;
385 my $res = $ipcc_send_rec_json->(CFS_IPC_GET_GUEST_CONFIG_PROPERTY
, $bindata);
390 # $data must be a chronological descending ordered array of tasks
391 sub broadcast_tasklist
{
394 # the serialized list may not get bigger than 128 KiB (CFS_MAX_STATUS_SIZE from pmxcfs)
395 # drop older items until we satisfy this constraint
396 my $size = length(encode_json
($data));
397 while ($size >= (32 * 1024)) { # TODO: update to 128 KiB in PVE 8.x
399 $size = length(encode_json
($data));
402 eval { $ipcc_update_status->("tasklist", $data) };
406 my $tasklistcache = {};
411 my $kvstore = $versions->{kvstore
} || {};
413 my $nodelist = get_nodelist
();
416 foreach my $node (@$nodelist) {
417 next if $nodename && ($nodename ne $node);
419 my $ver = exists $kvstore->{$node} ?
$kvstore->{$node}->{tasklist
} : undef;
420 my $cache = $tasklistcache->{$node};
421 if (!$cache || !$ver || !$cache->{version
} || ($cache->{version
} != $ver)) {
423 if (my $raw = $ipcc_get_status->("tasklist", $node)) {
424 my $json_str = unpack("Z*", $raw);
425 $tasks = decode_json
($json_str);
428 $tasklistcache->{$node} = {
432 } elsif ($cache && $cache->{data
}) {
433 push @$res, $cache->{data
}->@*;
437 syslog
('err', $err) if $err;
444 my ($rrdid, $data) = @_;
447 &$ipcc_update_status("rrd/$rrdid", $data);
454 my $last_rrd_dump = 0;
455 my $last_rrd_data = "";
461 my $diff = $ctime - $last_rrd_dump;
463 return $last_rrd_data;
468 $raw = &$ipcc_send_rec(CFS_IPC_GET_RRD_DUMP
);
480 while ($raw =~ s/^(.*)\n//) {
481 my ($key, @ela) = split(/:/, $1);
483 next if !(scalar(@ela) > 1);
484 $res->{$key} = [ map { $_ eq 'U' ?
undef : $_ } @ela ];
488 $last_rrd_dump = $ctime;
489 $last_rrd_data = $res;
495 # a fast way to read files (avoid fuse overhead)
499 return &$ipcc_get_config($path);
502 sub get_cluster_log
{
503 my ($user, $max) = @_;
505 return &$ipcc_get_cluster_log($user, $max);
509 my ($userid, $token) = @_;
511 return &$ipcc_verify_token("$userid $token");
516 sub cfs_register_file
{
517 my ($filename, $parser, $writer) = @_;
519 $observed->{$filename} || die "unknown file '$filename'";
521 die "file '$filename' already registered" if $file_info->{$filename};
523 $file_info->{$filename} = {
529 my $ccache_read = sub {
530 my ($filename, $parser, $version) = @_;
532 $ccache->{$filename} = {} if !$ccache->{$filename};
534 my $ci = $ccache->{$filename};
536 if (!$ci->{version
} || !$version || $ci->{version
} != $version) {
537 # we always call the parser, even when the file does not exist
538 # (in that case $data is undef)
539 my $data = get_config
($filename);
540 $ci->{data
} = &$parser("/etc/pve/$filename", $data);
541 $ci->{version
} = $version;
544 my $res = ref($ci->{data
}) ? dclone
($ci->{data
}) : $ci->{data
};
549 sub cfs_file_version
{
554 if ($filename =~ m!^nodes/[^/]+/(openvz|lxc|qemu-server)/(\d+)\.conf$!) {
555 my ($type, $vmid) = ($1, $2);
556 if ($vmlist && $vmlist->{ids
} && $vmlist->{ids
}->{$vmid}) {
557 $version = $vmlist->{ids
}->{$vmid}->{version
};
559 $infotag = "/$type/";
561 $infotag = $filename;
562 $version = $versions->{$filename};
565 my $info = $file_info->{$infotag} ||
566 die "unknown file type '$filename'\n";
568 return wantarray ?
($version, $info) : $version;
574 my ($version, $info) = cfs_file_version
($filename);
575 my $parser = $info->{parser
};
577 return &$ccache_read($filename, $parser, $version);
581 my ($filename, $data, $force_utf8) = @_;
583 my ($version, $info) = cfs_file_version
($filename);
585 my $writer = $info->{writer
} || die "no writer defined";
587 my $fsname = "/etc/pve/$filename";
589 my $raw = &$writer($fsname, $data);
591 if (my $ci = $ccache->{$filename}) {
592 $ci->{version
} = undef;
595 PVE
::Tools
::file_set_contents
($fsname, $raw, undef, 1);
599 my ($lockid, $timeout, $code, @param) = @_;
601 my $prev_alarm = alarm(0); # suspend outer alarm early
606 # this timeout is for acquire the lock
607 $timeout = 10 if !$timeout;
609 my $filename = "$lockdir/$lockid";
617 die "pve cluster filesystem not online.\n";
620 my $timeout_err = sub { die "got lock request timeout\n"; };
621 local $SIG{ALRM
} = $timeout_err;
625 $got_lock = mkdir($filename);
626 $timeout = alarm(0) - 1; # we'll sleep for 1s, see down below
630 $timeout_err->() if $timeout <= 0;
632 print STDERR
"trying to acquire cfs lock '$lockid' ...\n";
633 utime (0, 0, $filename); # cfs unlock request
637 # fixed command timeout: cfs locks have a timeout of 120
638 # using 60 gives us another 60 seconds to abort the task
639 local $SIG{ALRM
} = sub { die "'$lockid'-locked command timed out - aborting\n"; };
642 cfs_update
(); # make sure we read latest versions inside code()
644 $is_code_err = 1; # allows to differ between locking and actual-work errors
646 $res = &$code(@param);
653 $err = "no quorum!\n" if !$got_lock && !check_cfs_quorum
(1);
655 rmdir $filename if $got_lock; # if we held the lock always unlock again
660 if (ref($err) eq 'PVE::Exception' || $is_code_err) {
661 # re-raise defined exceptions
664 # add lock info for plain errors comming from the locking itself
665 $@ = "cfs-lock '$lockid' error: $err";
676 my ($filename, $timeout, $code, @param) = @_;
678 my $info = $observed->{$filename} || die "unknown file '$filename'";
680 my $lockid = "file-$filename";
681 $lockid =~ s/[.\/]/_
/g
;
683 &$cfs_lock($lockid, $timeout, $code, @param);
686 sub cfs_lock_storage
{
687 my ($storeid, $timeout, $code, @param) = @_;
689 my $lockid = "storage-$storeid";
691 &$cfs_lock($lockid, $timeout, $code, @param);
694 sub cfs_lock_domain
{
695 my ($domainname, $timeout, $code, @param) = @_;
697 my $lockid = "domain-$domainname";
699 &$cfs_lock($lockid, $timeout, $code, @param);
703 my ($account, $timeout, $code, @param) = @_;
705 my $lockid = "acme-$account";
707 &$cfs_lock($lockid, $timeout, $code, @param);
710 sub cfs_lock_authkey
{
711 my ($timeout, $code, @param) = @_;
713 $cfs_lock->('authkey', $timeout, $code, @param);
716 sub cfs_lock_firewall
{
717 my ($scope, $timeout, $code, @param) = @_;
719 my $lockid = "firewall-$scope";
721 $cfs_lock->($lockid, $timeout, $code, @param);
739 my ($priority, $ident, $msg) = @_;
741 if (my $tmp = $log_levels->{$priority}) {
745 die "need numeric log priority" if $priority !~ /^\d+$/;
747 my $tag = PVE
::SafeSyslog
::tag
();
749 $msg = "empty message" if !$msg;
751 $ident = "" if !$ident;
752 $ident = encode
("ascii", $ident,
753 sub { sprintf "\\u%04x", shift });
755 my $ascii = encode
("ascii", $msg, sub { sprintf "\\u%04x", shift });
758 syslog
($priority, "<%s> %s", $ident, $ascii);
760 syslog
($priority, "%s", $ascii);
763 eval { &$ipcc_log($priority, $ident, $tag, $ascii); };
765 syslog
("err", "writing cluster log failed: $@") if $@;
768 sub check_vmid_unused
{
769 my ($vmid, $noerr) = @_;
771 my $vmlist = get_vmlist
();
773 my $d = $vmlist->{ids
}->{$vmid};
774 return 1 if !defined($d);
776 return undef if $noerr;
778 my $vmtypestr = $d->{type
} eq 'qemu' ?
'VM' : 'CT';
779 die "$vmtypestr $vmid already exists on node '$d->{node}'\n";
782 sub check_node_exists
{
783 my ($nodename, $noerr) = @_;
785 my $nodelist = $clinfo->{nodelist
};
786 return 1 if $nodelist && $nodelist->{$nodename};
788 return undef if $noerr;
790 die "no such cluster node '$nodename'\n";
793 # this is also used to get the IP of the local node
795 my ($nodename, $noerr) = @_;
797 my $nodelist = $clinfo->{nodelist
};
798 if ($nodelist && $nodelist->{$nodename}) {
799 if (my $ip = $nodelist->{$nodename}->{ip
}) {
800 return $ip if !wantarray;
801 my $family = $nodelist->{$nodename}->{address_family
};
803 $nodelist->{$nodename}->{address_family
} =
805 PVE
::Tools
::get_host_address_family
($ip);
807 return wantarray ?
($ip, $family) : $ip;
811 # fallback: try to get IP by other means
812 return PVE
::Network
::get_ip_from_hostname
($nodename, $noerr);
815 sub get_node_fingerprint
{
818 my $cert_path = "/etc/pve/nodes/$node/pve-ssl.pem";
819 my $custom_cert_path = "/etc/pve/nodes/$node/pveproxy-ssl.pem";
821 $cert_path = $custom_cert_path if -f
$custom_cert_path;
823 return PVE
::Certificate
::get_certificate_fingerprint
($cert_path);
826 # bash completion helpers
828 sub complete_next_vmid
{
830 my $vmlist = get_vmlist
() || {};
831 my $idlist = $vmlist->{ids
} || {};
833 for (my $i = 100; $i < 10000; $i++) {
834 return [$i] if !defined($idlist->{$i});
842 my $vmlist = get_vmlist
();
843 my $ids = $vmlist->{ids
} || {};
845 return [ keys %$ids ];
848 sub complete_local_vmid
{
850 my $vmlist = get_vmlist
();
851 my $ids = $vmlist->{ids
} || {};
853 my $nodename = PVE
::INotify
::nodename
();
856 foreach my $vmid (keys %$ids) {
857 my $d = $ids->{$vmid};
858 next if !$d->{node
} || $d->{node
} ne $nodename;
865 sub complete_migration_target
{
869 my $nodename = PVE
::INotify
::nodename
();
871 my $nodelist = get_nodelist
();
872 foreach my $node (@$nodelist) {
873 next if $node eq $nodename;
881 # NOTE: filesystem must be offline here, no DB changes allowed
882 sub cfs_backup_database
{
886 my $backup_fn = "$dbbackupdir/config-$ctime.sql.gz";
888 print "backup old database to '$backup_fn'\n";
890 my $cmd = [ ['sqlite3', $dbfile, '.dump'], ['gzip', '-', \
">${backup_fn}"] ];
891 run_command
($cmd, 'errmsg' => "cannot backup old database\n");
893 my $maxfiles = 10; # purge older backup
894 my $backups = [ sort { $b cmp $a } <$dbbackupdir/config-*.sql
.gz
> ];
896 if ((my $count = scalar(@$backups)) > $maxfiles) {
897 foreach my $f (@$backups[$maxfiles..$count-1]) {
898 next if $f !~ m/^(\S+)$/; # untaint
899 print "delete old backup '$1'\n";