1 package DBIx::Class::Storage::DBI::Replicated;
4 use Carp::Clan qw/^DBIx::Class/;
6 croak('The following modules are required for Replication ' . DBIx::Class::Optional::Dependencies->req_missing_for ('replicated') )
7 unless DBIx::Class::Optional::Dependencies->req_ok_for ('replicated');
11 use DBIx::Class::Storage::DBI;
12 use DBIx::Class::Storage::DBI::Replicated::Pool;
13 use DBIx::Class::Storage::DBI::Replicated::Balancer;
14 use DBIx::Class::Storage::DBI::Replicated::Types qw/BalancerClassNamePart DBICSchema DBICStorageDBI/;
15 use MooseX::Types::Moose qw/ClassName HashRef Object/;
16 use Scalar::Util 'reftype';
18 use List::Util qw/min max reduce/;
20 use namespace::clean -except => 'meta';
24 DBIx::Class::Storage::DBI::Replicated - BETA Replicated database support
28 The Following example shows how to change an existing $schema to a replicated
29 storage type, add some replicated (read-only) databases, and perform reporting
32 You should set the 'storage_type attribute to a replicated type. You should
33 also define your arguments, such as which balancer you want and any arguments
34 that the Pool object should get.
36 my $schema = Schema::Class->clone;
37 $schema->storage_type( ['::DBI::Replicated', {balancer=>'::Random'}] );
38 $schema->connection(...);
40 Next, you need to add in the Replicants. Basically this is an array of
41 arrayrefs, where each arrayref is database connect information. Think of these
42 arguments as what you'd pass to the 'normal' $schema->connect method.
44 $schema->storage->connect_replicants(
45 [$dsn1, $user, $pass, \%opts],
46 [$dsn2, $user, $pass, \%opts],
47 [$dsn3, $user, $pass, \%opts],
50 Now, just use the $schema as you normally would. Automatically all reads will
51 be delegated to the replicants, while writes to the master.
53 $schema->resultset('Source')->search({name=>'etc'});
55 You can force a given query to use a particular storage using the search
56 attribute 'force_pool'. For example:
58 my $RS = $schema->resultset('Source')->search(undef, {force_pool=>'master'});
60 Now $RS will force everything (both reads and writes) to use whatever was setup
61 as the master storage. 'master' is hardcoded to always point to the Master,
62 but you can also use any Replicant name. Please see:
63 L<DBIx::Class::Storage::DBI::Replicated::Pool> and the replicants attribute for more.
65 Also see transactions and L</execute_reliably> for alternative ways to
66 force read traffic to the master. In general, you should wrap your statements
67 in a transaction when you are reading and writing to the same tables at the
68 same time, since your replicants will often lag a bit behind the master.
70 See L<DBIx::Class::Storage::DBI::Replicated::Instructions> for more help and
75 Warning: This class is marked BETA. This has been running a production
76 website using MySQL native replication as its backend and we have some decent
77 test coverage but the code hasn't yet been stressed by a variety of databases.
78 Individual DBs may have quirks we are not aware of. Please use this in first
79 development and pass along your experiences/bug fixes.
81 This class implements replicated data store for DBI. Currently you can define
82 one master and numerous slave database connections. All write-type queries
83 (INSERT, UPDATE, DELETE and even LAST_INSERT_ID) are routed to master
84 database, all read-type queries (SELECTs) go to the slave database.
86 Basically, any method request that L<DBIx::Class::Storage::DBI> would normally
87 handle gets delegated to one of the two attributes: L</read_handler> or to
88 L</write_handler>. Additionally, some methods need to be distributed
89 to all existing storages. This way our storage class is a drop in replacement
90 for L<DBIx::Class::Storage::DBI>.
92 Read traffic is spread across the replicants (slaves) occurring to a user
93 selected algorithm. The default algorithm is random weighted.
97 The consistency between master and replicants is database specific. The Pool
98 gives you a method to validate its replicants, removing and replacing them
99 when they fail/pass predefined criteria. Please make careful use of the ways
100 to force a query to run against Master when needed.
104 Replicated Storage has additional requirements not currently part of
105 L<DBIx::Class>. See L<DBIx::Class::Optional::Dependencies> for more details.
109 This class defines the following attributes.
113 The underlying L<DBIx::Class::Schema> object this storage is attaching
126 Contains the classname which will instantiate the L</pool> object. Defaults
127 to: L<DBIx::Class::Storage::DBI::Replicated::Pool>.
134 default=>'DBIx::Class::Storage::DBI::Replicated::Pool',
136 'create_pool' => 'new',
142 Contains a hashref of initialized information to pass to the Balancer object.
143 See L<DBIx::Class::Storage::DBI::Replicated::Pool> for available arguments.
157 The replication pool requires a balance class to provider the methods for
158 choose how to spread the query load across each replicant in the pool.
162 has 'balancer_type' => (
164 isa=>BalancerClassNamePart,
167 default=> 'DBIx::Class::Storage::DBI::Replicated::Balancer::First',
169 'create_balancer' => 'new',
175 Contains a hashref of initialized information to pass to the Balancer object.
176 See L<DBIx::Class::Storage::DBI::Replicated::Balancer> for available arguments.
180 has 'balancer_args' => (
190 Is a <DBIx::Class::Storage::DBI::Replicated::Pool> or derived class. This is a
191 container class for one or more replicated databases.
197 isa=>'DBIx::Class::Storage::DBI::Replicated::Pool',
208 Is a <DBIx::Class::Storage::DBI::Replicated::Balancer> or derived class. This
209 is a class that takes a pool (<DBIx::Class::Storage::DBI::Replicated::Pool>)
215 isa=>'DBIx::Class::Storage::DBI::Replicated::Balancer',
217 handles=>[qw/auto_validate_every/],
222 The master defines the canonical state for a pool of connected databases. All
223 the replicants are expected to match this databases state. Thus, in a classic
224 Master / Slaves distributed system, all the slaves are expected to replicate
225 the Master's state as quick as possible. This is the only database in the
226 pool of databases that is allowed to handle write traffic.
236 =head1 ATTRIBUTES IMPLEMENTING THE DBIx::Storage::DBI INTERFACE
238 The following methods are delegated all the methods required for the
239 L<DBIx::Class::Storage::DBI> interface.
243 Defines an object that implements the read side of L<BIx::Class::Storage::DBI>.
247 has 'read_handler' => (
255 _dbh_columns_info_for
262 Defines an object that implements the write side of L<BIx::Class::Storage::DBI>,
263 as well as methods that don't write or read that can be called on only one
264 storage, methods that return a C<$dbh>, and any methods that don't make sense to
269 has 'write_handler' => (
284 deployment_statements
287 build_datetime_parser
301 with_deferred_fk_checks
304 with_deferred_fk_checks
309 _supports_insert_returning
315 relname_to_table_alias
316 _straight_join_to_node
319 _default_dbi_connect_attributes
324 bind_attribute_by_data_type
333 _per_row_update_delete
335 _dbh_execute_inserts_with_no_binds
336 _select_args_to_query
338 _multipk_update_delete
339 source_bind_attributes
340 _normalize_connect_info
344 _placeholders_supported
346 _sqlt_minimum_version
349 _typeless_placeholders_supported
356 _adjust_select_args_for_complex_prefetch
357 _resolve_ident_sources
360 _strip_cond_qualifiers
362 _resolve_aliastypes_from_select_args
367 _prefetch_insert_auto_nextvals
372 my @unimplemented = qw(
373 _arm_global_destructor
374 _preserve_foreign_dbh
379 for my $method (@unimplemented) {
380 __PACKAGE__->meta->add_method($method, sub {
381 croak "$method must not be called on ".(blessed shift).' objects';
385 has _master_connect_info_opts =>
386 (is => 'rw', isa => HashRef, default => sub { {} });
388 =head2 around: connect_info
390 Preserves master's C<connect_info> options (for merging with replicants.)
391 Also sets any Replicated-related options from connect_info, such as
392 C<pool_type>, C<pool_args>, C<balancer_type> and C<balancer_args>.
396 around connect_info => sub {
397 my ($next, $self, $info, @extra) = @_;
399 my $wantarray = wantarray;
401 my $merge = Hash::Merge->new('LEFT_PRECEDENT');
404 for my $arg (@$info) {
405 next unless (reftype($arg)||'') eq 'HASH';
406 %opts = %{ $merge->merge($arg, \%opts) };
410 if (@opts{qw/pool_type pool_args/}) {
411 $self->pool_type(delete $opts{pool_type})
415 $merge->merge((delete $opts{pool_args} || {}), $self->pool_args)
418 $self->pool($self->_build_pool)
422 if (@opts{qw/balancer_type balancer_args/}) {
423 $self->balancer_type(delete $opts{balancer_type})
424 if $opts{balancer_type};
426 $self->balancer_args(
427 $merge->merge((delete $opts{balancer_args} || {}), $self->balancer_args)
430 $self->balancer($self->_build_balancer)
434 $self->_master_connect_info_opts(\%opts);
438 @res = $self->$next($info, @extra);
440 $res = $self->$next($info, @extra);
443 # Make sure master is blessed into the correct class and apply role to it.
444 my $master = $self->master;
445 $master->_determine_driver;
446 Moose::Meta::Class->initialize(ref $master);
448 DBIx::Class::Storage::DBI::Replicated::WithDSN->meta->apply($master);
450 # link pool back to master
451 $self->pool->master($master);
453 $wantarray ? @res : $res;
458 This class defines the following methods.
462 L<DBIx::Class::Schema> when instantiating its storage passed itself as the
463 first argument. So we need to massage the arguments a bit so that all the
464 bits get put into the correct places.
469 my ($class, $schema, $storage_type_args, @args) = @_;
480 Lazy builder for the L</master> attribute.
486 my $master = DBIx::Class::Storage::DBI->new($self->schema);
492 Lazy builder for the L</pool> attribute.
498 $self->create_pool(%{$self->pool_args});
501 =head2 _build_balancer
503 Lazy builder for the L</balancer> attribute. This takes a Pool object so that
504 the balancer knows which pool it's balancing.
508 sub _build_balancer {
510 $self->create_balancer(
512 master=>$self->master,
513 %{$self->balancer_args},
517 =head2 _build_write_handler
519 Lazy builder for the L</write_handler> attribute. The default is to set this to
524 sub _build_write_handler {
525 return shift->master;
528 =head2 _build_read_handler
530 Lazy builder for the L</read_handler> attribute. The default is to set this to
535 sub _build_read_handler {
536 return shift->balancer;
539 =head2 around: connect_replicants
541 All calls to connect_replicants needs to have an existing $schema tacked onto
542 top of the args, since L<DBIx::Storage::DBI> needs it, and any C<connect_info>
543 options merged with the master, with replicant opts having higher priority.
547 around connect_replicants => sub {
548 my ($next, $self, @args) = @_;
551 $r = [ $r ] unless reftype $r eq 'ARRAY';
553 $self->throw_exception('coderef replicant connect_info not supported')
554 if ref $r->[0] && reftype $r->[0] eq 'CODE';
556 # any connect_info options?
558 $i++ while $i < @$r && (reftype($r->[$i])||'') ne 'HASH';
561 $r->[$i] = {} unless $r->[$i];
563 # merge if two hashes
564 my @hashes = @$r[$i .. $#{$r}];
566 $self->throw_exception('invalid connect_info options')
567 if (grep { reftype($_) eq 'HASH' } @hashes) != @hashes;
569 $self->throw_exception('too many hashrefs in connect_info')
572 my $merge = Hash::Merge->new('LEFT_PRECEDENT');
573 my %opts = %{ $merge->merge(reverse @hashes) };
576 splice @$r, $i+1, ($#{$r} - $i), ();
578 # make sure master/replicants opts don't clash
579 my %master_opts = %{ $self->_master_connect_info_opts };
580 if (exists $opts{dbh_maker}) {
581 delete @master_opts{qw/dsn user password/};
583 delete $master_opts{dbh_maker};
586 %opts = %{ $merge->merge(\%opts, \%master_opts) };
592 $self->$next($self->schema, @args);
597 Returns an array of of all the connected storage backends. The first element
598 in the returned array is the master, and the remainings are each of the
605 return grep {defined $_ && blessed $_} (
607 values %{ $self->replicants },
611 =head2 execute_reliably ($coderef, ?@args)
613 Given a coderef, saves the current state of the L</read_handler>, forces it to
614 use reliable storage (e.g. sets it to the master), executes a coderef and then
615 restores the original state.
621 $schema->resultset('User')->create({name=>$name});
622 my $user_rs = $schema->resultset('User')->find({name=>$name});
626 my $user_rs = $schema->storage->execute_reliably($reliably, 'John');
628 Use this when you must be certain of your database state, such as when you just
629 inserted something and need to get a resultset including it, etc.
633 sub execute_reliably {
634 my ($self, $coderef, @args) = @_;
636 unless( ref $coderef eq 'CODE') {
637 $self->throw_exception('Second argument must be a coderef');
640 ##Get copy of master storage
641 my $master = $self->master;
643 ##Get whatever the current read hander is
644 my $current = $self->read_handler;
646 ##Set the read handler to master
647 $self->read_handler($master);
649 ## do whatever the caller needs
651 my $want_array = wantarray;
655 @result = $coderef->(@args);
656 } elsif(defined $want_array) {
657 ($result[0]) = ($coderef->(@args));
663 ##Reset to the original state
664 $self->read_handler($current);
666 ##Exception testing has to come last, otherwise you might leave the
667 ##read_handler set to master.
670 $self->throw_exception("coderef returned an error: $@");
672 return $want_array ? @result : $result[0];
676 =head2 set_reliable_storage
678 Sets the current $schema to be 'reliable', that is all queries, both read and
679 write are sent to the master
683 sub set_reliable_storage {
685 my $schema = $self->schema;
686 my $write_handler = $self->schema->storage->write_handler;
688 $schema->storage->read_handler($write_handler);
691 =head2 set_balanced_storage
693 Sets the current $schema to be use the </balancer> for all reads, while all
694 writes are sent to the master only
698 sub set_balanced_storage {
700 my $schema = $self->schema;
701 my $balanced_handler = $self->schema->storage->balancer;
703 $schema->storage->read_handler($balanced_handler);
708 Check that the master and at least one of the replicants is connected.
715 $self->master->connected &&
716 $self->pool->connected_replicants;
719 =head2 ensure_connected
721 Make sure all the storages are connected.
725 sub ensure_connected {
727 foreach my $source ($self->all_storages) {
728 $source->ensure_connected(@_);
734 Set the limit_dialect for all existing storages
740 foreach my $source ($self->all_storages) {
741 $source->limit_dialect(@_);
743 return $self->master->quote_char;
748 Set the quote_char for all existing storages
754 foreach my $source ($self->all_storages) {
755 $source->quote_char(@_);
757 return $self->master->quote_char;
762 Set the name_sep for all existing storages
768 foreach my $source ($self->all_storages) {
769 $source->name_sep(@_);
771 return $self->master->name_sep;
776 Set the schema object for all existing storages
782 foreach my $source ($self->all_storages) {
783 $source->set_schema(@_);
789 set a debug flag across all storages
796 foreach my $source ($self->all_storages) {
800 return $self->master->debug;
811 return $self->master->debugobj(@_);
822 return $self->master->debugfh(@_);
833 return $self->master->debugcb(@_);
838 disconnect everything
844 foreach my $source ($self->all_storages) {
845 $source->disconnect(@_);
851 set cursor class on all storages, or return master's
856 my ($self, $cursor_class) = @_;
859 $_->cursor_class($cursor_class) for $self->all_storages;
861 $self->master->cursor_class;
866 set cursor class on all storages, or return master's, alias for L</cursor_class>
872 my ($self, $cursor_class) = @_;
875 $_->cursor($cursor_class) for $self->all_storages;
877 $self->master->cursor;
882 sets the L<DBIx::Class::Storage::DBI/unsafe> option on all storages or returns
883 master's current setting
891 $_->unsafe(@_) for $self->all_storages;
894 return $self->master->unsafe;
897 =head2 disable_sth_caching
899 sets the L<DBIx::Class::Storage::DBI/disable_sth_caching> option on all storages
900 or returns master's current setting
904 sub disable_sth_caching {
908 $_->disable_sth_caching(@_) for $self->all_storages;
911 return $self->master->disable_sth_caching;
914 =head2 lag_behind_master
916 returns the highest Replicant L<DBIx::Class::Storage::DBI/lag_behind_master>
921 sub lag_behind_master {
924 return max map $_->lag_behind_master, $self->replicants;
927 =head2 is_replicating
929 returns true if all replicants return true for
930 L<DBIx::Class::Storage::DBI/is_replicating>
937 return (grep $_->is_replicating, $self->replicants) == ($self->replicants);
940 =head2 connect_call_datetime_setup
942 calls L<DBIx::Class::Storage::DBI/connect_call_datetime_setup> for all storages
946 sub connect_call_datetime_setup {
948 $_->connect_call_datetime_setup for $self->all_storages;
953 $_->_populate_dbh for $self->all_storages;
958 $_->_connect for $self->all_storages;
963 $_->_rebless for $self->all_storages;
966 sub _determine_driver {
968 $_->_determine_driver for $self->all_storages;
971 sub _driver_determined {
975 $_->_driver_determined(@_) for $self->all_storages;
978 return $self->master->_driver_determined;
984 $_->_init for $self->all_storages;
987 sub _run_connection_actions {
990 $_->_run_connection_actions for $self->all_storages;
993 sub _do_connection_actions {
997 $_->_do_connection_actions(@_) for $self->all_storages;
1001 sub connect_call_do_sql {
1003 $_->connect_call_do_sql(@_) for $self->all_storages;
1006 sub disconnect_call_do_sql {
1008 $_->disconnect_call_do_sql(@_) for $self->all_storages;
1011 sub _seems_connected {
1014 return min map $_->_seems_connected, $self->all_storages;
1020 return min map $_->_ping, $self->all_storages;
1023 my $numify_ver = sub {
1025 my @numparts = split /\D+/, $ver;
1026 my $format = '%d.' . (join '', ('%05d') x (@numparts - 1));
1028 return sprintf $format, @numparts;
1034 if (not $self->_server_info_hash) {
1035 my $min_version_info = (
1036 reduce { $a->[0] < $b->[0] ? $a : $b }
1037 map [ $numify_ver->($_->{dbms_version}), $_ ],
1038 map $_->_server_info, $self->all_storages
1041 $self->_server_info_hash($min_version_info); # on master
1044 return $self->_server_info_hash;
1047 sub _get_server_version {
1050 return $self->_server_info->{dbms_version};
1055 Due to the fact that replicants can lag behind a master, you must take care to
1056 make sure you use one of the methods to force read queries to a master should
1057 you need realtime data integrity. For example, if you insert a row, and then
1058 immediately re-read it from the database (say, by doing $row->discard_changes)
1059 or you insert a row and then immediately build a query that expects that row
1060 to be an item, you should force the master to handle reads. Otherwise, due to
1061 the lag, there is no certainty your data will be in the expected state.
1063 For data integrity, all transactions automatically use the master storage for
1064 all read and write queries. Using a transaction is the preferred and recommended
1065 method to force the master to handle all read queries.
1067 Otherwise, you can force a single query to use the master with the 'force_pool'
1070 my $row = $resultset->search(undef, {force_pool=>'master'})->find($pk);
1072 This attribute will safely be ignore by non replicated storages, so you can use
1073 the same code for both types of systems.
1075 Lastly, you can use the L</execute_reliably> method, which works very much like
1078 For debugging, you can turn replication on/off with the methods L</set_reliable_storage>
1079 and L</set_balanced_storage>, however this operates at a global level and is not
1080 suitable if you have a shared Schema object being used by multiple processes,
1081 such as on a web application server. You can get around this limitation by
1082 using the Schema clone method.
1084 my $new_schema = $schema->clone;
1085 $new_schema->set_reliable_storage;
1087 ## $new_schema will use only the Master storage for all reads/writes while
1088 ## the $schema object will use replicated storage.
1092 John Napiorkowski <john.napiorkowski@takkle.com>
1094 Based on code originated by:
1096 Norbert Csongrádi <bert@cpan.org>
1097 Peter Siklósi <einon@einon.hu>
1101 You may distribute this code under the same terms as Perl itself.
1105 __PACKAGE__->meta->make_immutable;