From 07a79e31c07f85764a5d8b651ce420b208d9c164 Mon Sep 17 00:00:00 2001 From: wangw Date: Tue, 14 Jun 2022 13:44:09 +0800 Subject: [PATCH v16 1/2] Fix data replicated twice when specifying publish_via_partition_root option. If there are two publications - one of them publishing a parent table (using publish_via_partition_root = true) and the other is publishing one of the parent's child tables - then subscribing to both publications from one subscription results in the same initial child data being copied twice. It should only be copied once. To fix this, the API function pg_get_publication_tables has been extended to take a publication list. Now, when getting the table information, if the publish_via_partition_root is true, the function can exclude a partition table whose ancestor is also published by the same publication list. --- src/backend/catalog/pg_publication.c | 206 ++++++++++++++------- src/backend/commands/subscriptioncmds.c | 81 ++++++-- src/include/catalog/pg_proc.dat | 12 +- src/test/regress/expected/rules.out | 2 +- src/test/subscription/t/013_partition.pl | 18 +- src/test/subscription/t/028_row_filter.pl | 13 +- src/test/subscription/t/031_column_list.pl | 5 +- src/tools/pgindent/typedefs.list | 1 + 8 files changed, 231 insertions(+), 107 deletions(-) diff --git a/src/backend/catalog/pg_publication.c b/src/backend/catalog/pg_publication.c index a98fcad421..a1d0398c69 100644 --- a/src/backend/catalog/pg_publication.c +++ b/src/backend/catalog/pg_publication.c @@ -45,6 +45,14 @@ #include "utils/rel.h" #include "utils/syscache.h" +/* Records association between publication and published table */ +typedef struct +{ + Oid relid; /* OID of published table */ + Oid pubid; /* OID of publication that publishes this + * table. */ +} published_rel; + static void publication_translate_columns(Relation targetrel, List *columns, int *natts, AttrNumber **attrs); @@ -175,39 +183,45 @@ pg_relation_is_publishable(PG_FUNCTION_ARGS) * Filter out the partitions whose parent tables were also specified in * the publication. */ -static List * -filter_partitions(List *relids) +static void +filter_partitions(List *table_infos) { - List *result = NIL; ListCell *lc; - ListCell *lc2; - foreach(lc, relids) + foreach(lc, table_infos) { - bool skip = false; - List *ancestors = NIL; - Oid relid = lfirst_oid(lc); + bool skip = false; + List *ancestors = NIL; + ListCell *lc2; + published_rel *table_info = (published_rel *) lfirst(lc); - if (get_rel_relispartition(relid)) - ancestors = get_partition_ancestors(relid); + if (get_rel_relispartition(table_info->relid)) + ancestors = get_partition_ancestors(table_info->relid); foreach(lc2, ancestors) { Oid ancestor = lfirst_oid(lc2); + ListCell *lc3; /* Check if the parent table exists in the published table list. */ - if (list_member_oid(relids, ancestor)) + foreach(lc3, table_infos) { - skip = true; - break; + Oid relid = ((published_rel *) lfirst(lc3))->relid; + + if (relid == ancestor) + { + skip = true; + break; + } } + + if (skip) + break; } - if (!skip) - result = lappend_oid(result, relid); + if (skip) + table_infos = foreach_delete_current(table_infos, lc); } - - return result; } /* @@ -1026,22 +1040,28 @@ GetPublicationByName(const char *pubname, bool missing_ok) } /* - * Returns information of tables in a publication. + * Get information of the tables in the given publication array. + * + * Returns pubid, relid, column list, row filter for each table. */ Datum pg_get_publication_tables(PG_FUNCTION_ARGS) { -#define NUM_PUBLICATION_TABLES_ELEM 3 +#define NUM_PUBLICATION_TABLES_ELEM 4 FuncCallContext *funcctx; - char *pubname = text_to_cstring(PG_GETARG_TEXT_PP(0)); - Publication *publication; - List *tables; + List *table_infos = NIL; /* stuff done only on the first call of the function */ if (SRF_IS_FIRSTCALL()) { TupleDesc tupdesc; MemoryContext oldcontext; + ArrayType *arr; + Datum *elems; + int nelems, + i; + bool viaroot = false; + ListCell *lc; /* create a function context for cross-call persistence */ funcctx = SRF_FIRSTCALL_INIT(); @@ -1049,68 +1069,109 @@ pg_get_publication_tables(PG_FUNCTION_ARGS) /* switch to memory context appropriate for multiple function calls */ oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx); - publication = GetPublicationByName(pubname, false); - /* - * Publications support partitioned tables, although all changes are - * replicated using leaf partition identity and schema, so we only - * need those. + * Deconstruct the parameter into elements where each element is a + * publication name. */ - if (publication->alltables) - { - tables = GetAllTablesPublicationRelations(publication->pubviaroot); - } - else + arr = PG_GETARG_ARRAYTYPE_P(0); + deconstruct_array(arr, TEXTOID, -1, false, TYPALIGN_INT, + &elems, NULL, &nelems); + + /* Get Oids of tables from each publication. */ + for (i = 0; i < nelems; i++) { - List *relids, - *schemarelids; - - relids = GetPublicationRelations(publication->oid, - publication->pubviaroot ? - PUBLICATION_PART_ROOT : - PUBLICATION_PART_LEAF); - schemarelids = GetAllSchemaPublicationRelations(publication->oid, - publication->pubviaroot ? - PUBLICATION_PART_ROOT : - PUBLICATION_PART_LEAF); - tables = list_concat_unique_oid(relids, schemarelids); + Publication *pub_elem; + List *pub_elem_tables = NIL; + + pub_elem = GetPublicationByName(TextDatumGetCString(elems[i]), false); /* - * If the publication publishes partition changes via their - * respective root partitioned tables, we must exclude partitions - * in favor of including the root partitioned tables. Otherwise, - * the function could return both the child and parent tables - * which could cause data of the child table to be - * double-published on the subscriber side. + * Publications support partitioned tables. If + * publish_via_partition_root is false, all changes are replicated + * using leaf partition identity and schema, so we only need those. + * Otherwise, get the partitioned table itself. */ - if (publication->pubviaroot) - tables = filter_partitions(tables); + if (pub_elem->alltables) + pub_elem_tables = GetAllTablesPublicationRelations(pub_elem->pubviaroot); + else + { + List *relids, + *schemarelids; + + relids = GetPublicationRelations(pub_elem->oid, + pub_elem->pubviaroot ? + PUBLICATION_PART_ROOT : + PUBLICATION_PART_LEAF); + schemarelids = GetAllSchemaPublicationRelations(pub_elem->oid, + pub_elem->pubviaroot ? + PUBLICATION_PART_ROOT : + PUBLICATION_PART_LEAF); + pub_elem_tables = list_concat_unique_oid(relids, schemarelids); + } + + /* + * Record the published table and the corresponding publication so + * that we can get row filters and column list later. + * + * When a table is published by multiple publications, to obtain + * all row filters and column list, the structure related to this + * table will be recorded multiple times. + */ + foreach(lc, pub_elem_tables) + { + published_rel *table_info = (published_rel *) malloc(sizeof(published_rel)); + + table_info->relid = lfirst_oid(lc); + table_info->pubid = pub_elem->oid; + table_infos = lappend(table_infos, table_info); + } + + /* At least one publication is using publish_via_partition_root. */ + if (pub_elem->pubviaroot) + viaroot = true; } + pfree(elems); + + /* + * If the publication publishes partition changes via their respective + * root partitioned tables, we must exclude partitions in favor of + * including the root partitioned tables. Otherwise, the function + * could return both the child and parent tables which could cause + * data of the child table to be double-published on the subscriber + * side. + */ + if (viaroot) + filter_partitions(table_infos); + /* Construct a tuple descriptor for the result rows. */ tupdesc = CreateTemplateTupleDesc(NUM_PUBLICATION_TABLES_ELEM); - TupleDescInitEntry(tupdesc, (AttrNumber) 1, "relid", + TupleDescInitEntry(tupdesc, (AttrNumber) 1, "pubid", + OIDOID, -1, 0); + TupleDescInitEntry(tupdesc, (AttrNumber) 2, "relid", OIDOID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 2, "attrs", + TupleDescInitEntry(tupdesc, (AttrNumber) 3, "attrs", INT2VECTOROID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 3, "qual", + TupleDescInitEntry(tupdesc, (AttrNumber) 4, "qual", PG_NODE_TREEOID, -1, 0); funcctx->tuple_desc = BlessTupleDesc(tupdesc); - funcctx->user_fctx = (void *) tables; + funcctx->user_fctx = (void *) table_infos; MemoryContextSwitchTo(oldcontext); } /* stuff done on every call of the function */ funcctx = SRF_PERCALL_SETUP(); - tables = (List *) funcctx->user_fctx; + table_infos = (List *) funcctx->user_fctx; - if (funcctx->call_cntr < list_length(tables)) + if (funcctx->call_cntr < list_length(table_infos)) { HeapTuple pubtuple = NULL; HeapTuple rettuple; - Oid relid = list_nth_oid(tables, funcctx->call_cntr); + Publication *pub; + published_rel *table_info = (published_rel *) list_nth(table_infos, funcctx->call_cntr); + Oid relid = table_info->relid; Oid schemaid = get_rel_namespace(relid); Datum values[NUM_PUBLICATION_TABLES_ELEM] = {0}; bool nulls[NUM_PUBLICATION_TABLES_ELEM] = {0}; @@ -1119,42 +1180,43 @@ pg_get_publication_tables(PG_FUNCTION_ARGS) * Form tuple with appropriate data. */ - publication = GetPublicationByName(pubname, false); + pub = GetPublication(table_info->pubid); - values[0] = ObjectIdGetDatum(relid); + values[0] = ObjectIdGetDatum(pub->oid); + values[1] = ObjectIdGetDatum(relid); /* * We don't consider row filters or column lists for FOR ALL TABLES or * FOR TABLES IN SCHEMA publications. */ - if (!publication->alltables && + if (!pub->alltables && !SearchSysCacheExists2(PUBLICATIONNAMESPACEMAP, ObjectIdGetDatum(schemaid), - ObjectIdGetDatum(publication->oid))) + ObjectIdGetDatum(pub->oid))) pubtuple = SearchSysCacheCopy2(PUBLICATIONRELMAP, ObjectIdGetDatum(relid), - ObjectIdGetDatum(publication->oid)); + ObjectIdGetDatum(pub->oid)); if (HeapTupleIsValid(pubtuple)) { /* Lookup the column list attribute. */ - values[1] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple, + values[2] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple, Anum_pg_publication_rel_prattrs, - &(nulls[1])); + &(nulls[2])); /* Null indicates no filter. */ - values[2] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple, + values[3] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple, Anum_pg_publication_rel_prqual, - &(nulls[2])); + &(nulls[3])); } else { - nulls[1] = true; nulls[2] = true; + nulls[3] = true; } /* Show all columns when the column list is not specified. */ - if (nulls[1] == true) + if (nulls[2] == true) { Relation rel = table_open(relid, AccessShareLock); int nattnums = 0; @@ -1176,8 +1238,8 @@ pg_get_publication_tables(PG_FUNCTION_ARGS) if (nattnums > 0) { - values[1] = PointerGetDatum(buildint2vector(attnums, nattnums)); - nulls[1] = false; + values[2] = PointerGetDatum(buildint2vector(attnums, nattnums)); + nulls[2] = false; } table_close(rel, AccessShareLock); diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c index 464db6d247..c2a9dacb0e 100644 --- a/src/backend/commands/subscriptioncmds.c +++ b/src/backend/commands/subscriptioncmds.c @@ -1936,25 +1936,84 @@ static List * fetch_table_list(WalReceiverConn *wrconn, List *publications) { WalRcvExecResult *res; - StringInfoData cmd; + StringInfoData cmd, + pub_names; TupleTableSlot *slot; Oid tableRow[3] = {TEXTOID, TEXTOID, NAMEARRAYOID}; List *tablelist = NIL; - bool check_columnlist = (walrcv_server_version(wrconn) >= 150000); + int server_version = walrcv_server_version(wrconn); + bool check_columnlist = (server_version >= 150000); - initStringInfo(&cmd); - appendStringInfoString(&cmd, "SELECT DISTINCT t.schemaname, t.tablename \n"); + initStringInfo(&pub_names); + get_publications_str(publications, &pub_names, true); - /* Get column lists for each relation if the publisher supports it */ - if (check_columnlist) - appendStringInfoString(&cmd, ", t.attnames\n"); + initStringInfo(&cmd); - appendStringInfoString(&cmd, "FROM pg_catalog.pg_publication_tables t\n" - " WHERE t.pubname IN ("); - get_publications_str(publications, &cmd, true); - appendStringInfoChar(&cmd, ')'); + /* + * Get namespace, relname and column list (if supported) of the tables + * belonging to the specified publications. + * + * Get the list of tables from the publisher. The partition table whose + * ancestor is also in this list will be ignored, otherwise the initial + * data in the partition table would be replicated twice. + * + * From version 16, the parameter of the function + * pg_get_publication_tables can be an array of publications. The + * partition table whose ancestor is also published in this publication + * array will be filtered out in this function. + */ + if (server_version >= 160000) + { + appendStringInfo(&cmd, "SELECT DISTINCT N.nspname, C.relname,\n" + " ( SELECT array_agg(a.attname ORDER BY a.attnum)\n" + " FROM pg_attribute a\n" + " WHERE a.attrelid = GPT.relid AND a.attnum > 0 AND\n" + " NOT a.attisdropped AND\n" + " (a.attnum = ANY(GPT.attrs) OR GPT.attrs IS NULL)\n" + " ) AS attnames\n" + " FROM pg_class C\n" + " JOIN pg_namespace N ON N.oid = C.relnamespace\n" + " JOIN ( SELECT (pg_get_publication_tables(VARIADIC array_agg(pubname::text))).*\n" + " FROM pg_publication\n" + " WHERE pubname IN ( %s )) as GPT\n" + " ON GPT.relid = C.oid\n", + pub_names.data); + } + else + { + appendStringInfoString(&cmd, "WITH pub_tabs AS(\n" + " SELECT DISTINCT N.nspname, C.oid, C.relname, C.relispartition\n"); + + /* Get column lists for each relation if the publisher supports it */ + if (check_columnlist) + appendStringInfoString(&cmd, ", ( SELECT array_agg(a.attname ORDER BY a.attnum)\n" + " FROM pg_attribute a\n" + " WHERE a.attrelid = GPT.relid AND a.attnum > 0 AND\n" + " NOT a.attisdropped AND\n" + " (a.attnum = ANY(GPT.attrs) OR GPT.attrs IS NULL)\n" + " ) AS attnames\n"); + + appendStringInfo(&cmd, " FROM pg_publication P,\n" + " LATERAL pg_get_publication_tables(P.pubname) GPT,\n" + " pg_class C JOIN pg_namespace N ON (N.oid = C.relnamespace)\n" + " WHERE C.oid = GPT.relid AND P.pubname IN ( %s )\n" + ")\n" + "SELECT DISTINCT pub_tabs.nspname, pub_tabs.relname\n", + pub_names.data); + + /* Get column lists for each relation if the publisher supports it */ + if (check_columnlist) + appendStringInfoString(&cmd, ", pub_tabs.attnames\n"); + + appendStringInfoString(&cmd, "FROM pub_tabs\n" + " WHERE (pub_tabs.relispartition IS FALSE\n" + " OR NOT EXISTS (SELECT 1 FROM pg_partition_ancestors(pub_tabs.oid) as PA\n" + " WHERE PA.relid IN (SELECT pub_tabs.oid FROM pub_tabs)\n" + " AND PA.relid != pub_tabs.oid))\n"); + } res = walrcv_exec(wrconn, cmd.data, check_columnlist ? 3 : 2, tableRow); + pfree(pub_names.data); pfree(cmd.data); if (res->status != WALRCV_OK_TUPLES) diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index c0f2a8a77c..f275c74233 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -11705,11 +11705,13 @@ prosrc => 'pg_show_replication_origin_status' }, # publications -{ oid => '6119', descr => 'get information of tables in a publication', - proname => 'pg_get_publication_tables', prorows => '1000', proretset => 't', - provolatile => 's', prorettype => 'record', proargtypes => 'text', - proallargtypes => '{text,oid,int2vector,pg_node_tree}', - proargmodes => '{i,o,o,o}', proargnames => '{pubname,relid,attrs,qual}', +{ oid => '6119', + descr => 'get information of the tables in the given publication array', + proname => 'pg_get_publication_tables', prorows => '1000', + provariadic => 'text', proretset => 't', provolatile => 's', + prorettype => 'record', proargtypes => '_text', + proallargtypes => '{_text,oid,oid,int2vector,pg_node_tree}', + proargmodes => '{v,o,o,o,o}', proargnames => '{pubname,pubid,relid,attrs,qual}', prosrc => 'pg_get_publication_tables' }, { oid => '6121', descr => 'returns whether a relation can be part of a publication', diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out index e7a2f5856a..1a933408b4 100644 --- a/src/test/regress/expected/rules.out +++ b/src/test/regress/expected/rules.out @@ -1449,7 +1449,7 @@ pg_publication_tables| SELECT p.pubname, WHERE ((a.attrelid = gpt.relid) AND (a.attnum = ANY ((gpt.attrs)::smallint[])))) AS attnames, pg_get_expr(gpt.qual, gpt.relid) AS rowfilter FROM pg_publication p, - LATERAL pg_get_publication_tables((p.pubname)::text) gpt(relid, attrs, qual), + LATERAL pg_get_publication_tables(VARIADIC ARRAY[(p.pubname)::text]) gpt(pubid, relid, attrs, qual), (pg_class c JOIN pg_namespace n ON ((n.oid = c.relnamespace))) WHERE (c.oid = gpt.relid); diff --git a/src/test/subscription/t/013_partition.pl b/src/test/subscription/t/013_partition.pl index 11a5c3c03e..58e28d51d0 100644 --- a/src/test/subscription/t/013_partition.pl +++ b/src/test/subscription/t/013_partition.pl @@ -473,12 +473,12 @@ $node_subscriber2->safe_psql('postgres', "CREATE TABLE tab3_1 (a int PRIMARY KEY, c text DEFAULT 'sub2_tab3_1', b text)" ); -# Note: We create two separate tables, not a partitioned one, so that we can -# easily identity through which relation were the changes replicated. +# Note: We only create one table (tab4) here. We specified +# publish_via_partition_root = true (see pub_all and pub_lower_level above), so +# all data will be replicated to that table. $node_subscriber2->safe_psql('postgres', "CREATE TABLE tab4 (a int PRIMARY KEY)"); -$node_subscriber2->safe_psql('postgres', - "CREATE TABLE tab4_1 (a int PRIMARY KEY)"); + # Publication that sub2 points to now publishes via root, so must update # subscription target relations. We set the list of publications so that # the FOR ALL TABLES publication is second (the list order matters). @@ -550,11 +550,6 @@ $result = $node_subscriber2->safe_psql('postgres', "SELECT a FROM tab4 ORDER BY 1"); is($result, qq(0), 'inserts into tab4 replicated'); -$result = - $node_subscriber2->safe_psql('postgres', "SELECT a FROM tab4_1 ORDER BY 1"); -is($result, qq(), 'inserts into tab4_1 replicated'); - - # now switch the order of publications in the list, try again, the result # should be the same (no dependence on order of pulications) $node_subscriber2->safe_psql('postgres', @@ -577,11 +572,6 @@ $result = is( $result, qq(0 1), 'inserts into tab4 replicated'); -$result = - $node_subscriber2->safe_psql('postgres', "SELECT a FROM tab4_1 ORDER BY 1"); -is($result, qq(), 'inserts into tab4_1 replicated'); - - # update (replicated as update) $node_publisher->safe_psql('postgres', "UPDATE tab1 SET a = 6 WHERE a = 5"); $node_publisher->safe_psql('postgres', "UPDATE tab2 SET a = 6 WHERE a = 5"); diff --git a/src/test/subscription/t/028_row_filter.pl b/src/test/subscription/t/028_row_filter.pl index def3fbc54b..d8972aeb93 100644 --- a/src/test/subscription/t/028_row_filter.pl +++ b/src/test/subscription/t/028_row_filter.pl @@ -386,6 +386,10 @@ $node_publisher->safe_psql('postgres', "INSERT INTO tab_rowfilter_child(a, b) VALUES(0,'0'),(30,'30'),(40,'40')" ); +# insert data into partitioned table. +$node_publisher->safe_psql('postgres', + "INSERT INTO tab_rowfilter_viaroot_part(a) VALUES(13), (17)"); + $node_subscriber->safe_psql('postgres', "CREATE SUBSCRIPTION tap_sub CONNECTION '$publisher_connstr application_name=$appname' PUBLICATION tap_pub_1, tap_pub_2, tap_pub_3, tap_pub_4a, tap_pub_4b, tap_pub_5a, tap_pub_5b, tap_pub_toast, tap_pub_inherits, tap_pub_viaroot_2, tap_pub_viaroot_1" ); @@ -707,13 +711,18 @@ is($result, qq(t|1), 'check replicated rows to tab_rowfilter_toast'); # the row filter for the top-level ancestor: # # tab_rowfilter_viaroot_part filter is: (a > 15) +# - INSERT (13) NO, 13 < 15 # - INSERT (14) NO, 14 < 15 # - INSERT (15) NO, 15 = 15 # - INSERT (16) YES, 16 > 15 +# - INSERT (17) YES, 17 > 15 $result = $node_subscriber->safe_psql('postgres', - "SELECT a FROM tab_rowfilter_viaroot_part"); -is($result, qq(16), 'check replicated rows to tab_rowfilter_viaroot_part'); + "SELECT a FROM tab_rowfilter_viaroot_part ORDER BY 1"); +is($result, qq(16 +17), + 'check replicated rows to tab_rowfilter_viaroot_part' +); # Check there is no data in tab_rowfilter_viaroot_part_1 because rows are # replicated via the top most parent table tab_rowfilter_viaroot_part diff --git a/src/test/subscription/t/031_column_list.pl b/src/test/subscription/t/031_column_list.pl index 7c313e26ee..7c8295b13b 100644 --- a/src/test/subscription/t/031_column_list.pl +++ b/src/test/subscription/t/031_column_list.pl @@ -959,7 +959,8 @@ $node_publisher->safe_psql( CREATE TABLE test_root_1 PARTITION OF test_root FOR VALUES FROM (1) TO (10); CREATE TABLE test_root_2 PARTITION OF test_root FOR VALUES FROM (10) TO (20); - CREATE PUBLICATION pub_root_true FOR TABLE test_root (a) WITH (publish_via_partition_root = true); + CREATE PUBLICATION pub_root_true_1 FOR TABLE test_root (a) WITH (publish_via_partition_root = true); + CREATE PUBLICATION pub_root_true_2 FOR TABLE test_root_1 (a, b) WITH (publish_via_partition_root = true); -- initial data INSERT INTO test_root VALUES (1, 2, 3); @@ -968,7 +969,7 @@ $node_publisher->safe_psql( $node_subscriber->safe_psql( 'postgres', qq( - CREATE SUBSCRIPTION sub1 CONNECTION '$publisher_connstr' PUBLICATION pub_root_true; + CREATE SUBSCRIPTION sub1 CONNECTION '$publisher_connstr' PUBLICATION pub_root_true_1, pub_root_true_2; )); $node_subscriber->wait_for_subscription_sync; diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 07fbb7ccf6..8f3b920b69 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3589,6 +3589,7 @@ pthread_mutex_t pthread_once_t pthread_t ptrdiff_t +published_rel pull_var_clause_context pull_varattnos_context pull_varnos_context -- 2.39.1.windows.1