public inbox for [email protected]
help / color / mirror / Atom feedFrom: [email protected] <[email protected]>
To: Peter Smith <[email protected]>
Cc: Takamichi Osumi (Fujitsu) <[email protected]>
Cc: Amit Kapila <[email protected]>
Cc: [email protected] <[email protected]>
Cc: [email protected] <[email protected]>
Cc: Amit Langote <[email protected]>
Cc: Peter Eisentraut <[email protected]>
Cc: [email protected] <[email protected]>
Cc: Dilip Kumar <[email protected]>
Cc: Greg Nancarrow <[email protected]>
Cc: vignesh C <[email protected]>
Subject: RE: Data is copied twice when specifying both child and parent table in publication
Date: Fri, 11 Nov 2022 05:43:10 +0000
Message-ID: <OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com> (raw)
In-Reply-To: <CAHut+Pt+1PNx6VsZ-xKzAU-18HmNXhjCC1TGakKX46Wg7YNT1Q@mail.gmail.com>
References: <OS3PR01MB62751414E4DABF144F5E481C9E529@OS3PR01MB6275.jpnprd01.prod.outlook.com>
<CAHut+PtTN1Udug3x1eZnDdh8Z6PL5VN=xz1b44VbB9T1z_PtcQ@mail.gmail.com>
<OS3PR01MB6275A9B8C65C381C6828DF9D9E549@OS3PR01MB6275.jpnprd01.prod.outlook.com>
<CAHut+PvaSDuZ7PhzZOm4c2u-sbN4DUsZr22qyYQ8dncUpqMpXg@mail.gmail.com>
<OS3PR01MB6275843B2BBE92870F7881C19E299@OS3PR01MB6275.jpnprd01.prod.outlook.com>
<CAHut+Pt+1PNx6VsZ-xKzAU-18HmNXhjCC1TGakKX46Wg7YNT1Q@mail.gmail.com>
On Fri, Oct 21, 2022 at 17:02 PM Peter Smith <[email protected]> wrote:
>
Thanks for your comments. Sorry for not replying in time.
> On Mon, Oct 17, 2022 at 4:49 PM [email protected]
> <[email protected]> wrote:
> >
> > On Wed, Oct 5, 2022 at 11:08 AM Peter Smith <[email protected]>
> wrote:
> > > Hi Wang-san. Here are my review comments for HEAD_v12-0001 patch.
> >
> ...
> > >
> > > 3. QUESTION - pg_get_publication_tables / fetch_table_list
> > >
> > > When the same table is published by different publications (but there
> > > are other differences like row-filters/column-lists in each
> > > publication) the result tuple of this function does not include the
> > > pubid. Maybe the SQL of pg_publication_tables/fetch_table_list() is OK
> > > as-is but how does it manage to associate each table with the correct
> > > tuple?
> > >
> > > I know it apparently all seems to work but I’m not how does that
> > > happen? Can you explain why a puboid is not needed for the result
> > > tuple of this function?
> >
> > Sorry, I am not sure I understand your question.
> > I try to answer your question by explaining the two functions you mentioned:
> >
> > First, the function pg_get_publication_tables gets the list (see table_infos)
> > that included published table and the corresponding publication. Then based
> > on this list, the function pg_get_publication_tables returns information
> > (scheme, relname, row filter and column list) about the published tables in the
> > publications list. It just doesn't return pubid.
> >
> > Then, the SQL in the function fetch_table_list will get the columns in the
> > column list from pg_attribute. (This is to return all columns when the column
> > list is not specified)
> >
>
> I meant, for example, if the different publications specified
> different col-lists for the same table then IIUC the
> fetch_table_lists() is going to return 2 list elements
> (schema,rel_name,row_filter,col_list). But when the schema/rel_name
> are the same for 2 elements then (without also a pubid) how are you
> going to know where the list element came from, and how come that is
> not important?
>
> > > ~~
> > >
> > > test_pub=# create table t1(a int, b int, c int);
> > > CREATE TABLE
> > > test_pub=# create publication pub1 for table t1(a) where (a > 99);
> > > CREATE PUBLICATION
> > > test_pub=# create publication pub2 for table t1(a,b) where (b < 33);
> > > CREATE PUBLICATION
> > >
> > > Following seems OK when I swap orders of publication names...
> > >
> > > test_pub=# SELECT gpt.relid, gpt.attrs, pg_get_expr(gpt.qual,
> > > gpt.relid) AS rowfilter from pg_get_publication_tables(VARIADIC
> > > ARRAY['pub2','pub1']) gpt(relid, attrs, qual);
> > > relid | attrs | rowfilter
> > > -------+-------+-----------
> > > 16385 | 1 2 | (b < 33)
> > > 16385 | 1 | (a > 99)
> > > (2 rows)
> > >
> > > test_pub=# SELECT gpt.relid, gpt.attrs, pg_get_expr(gpt.qual,
> > > gpt.relid) AS rowfilter from pg_get_publication_tables(VARIADIC
> > > ARRAY['pub1','pub2']) gpt(relid, attrs, qual);
> > > relid | attrs | rowfilter
> > > -------+-------+-----------
> > > 16385 | 1 | (a > 99)
> > > 16385 | 1 2 | (b < 33)
> > > (2 rows)
> > >
> > > But what about this (this is similar to the SQL fragment from
> > > fetch_table_list); I swapped the pub names but the results are the
> > > same...
> > >
> > > test_pub=# SELECT pg_get_publication_tables(VARIADIC
> > > array_agg(p.pubname)) from pg_publication p where pubname
> > > IN('pub2','pub1');
> > >
> > > pg_get_publication_tables
> > >
> > > ---------------------------------------------------------------------------------------------
> ----
> > > ---------------------------------------------------------------------
> > > ---------------------------------------------------------------------------------------------
> ----
> > > ---------------------------------------------------------------------
> > > -------------------------------------------------------------------
> > > (16385,1,"{OPEXPR :opno 521 :opfuncid 147 :opresulttype 16 :opretset
> > > false :opcollid 0 :inputcollid 0 :args ({VAR :varno 1 :varattno 1
> > > :vartype 23 :vartypmod -1 :var
> > > collid 0 :varlevelsup 0 :varnosyn 1 :varattnosyn 1 :location 47}
> > > {CONST :consttype 23 :consttypmod -1 :constcollid 0 :constlen 4
> > > :constbyval true :constisnull false :
> > > location 51 :constvalue 4 [ 99 0 0 0 0 0 0 0 ]}) :location 49}")
> > > (16385,"1 2","{OPEXPR :opno 97 :opfuncid 66 :opresulttype 16
> > > :opretset false :opcollid 0 :inputcollid 0 :args ({VAR :varno 1
> > > :varattno 2 :vartype 23 :vartypmod -1 :v
> > > arcollid 0 :varlevelsup 0 :varnosyn 1 :varattnosyn 2 :location 49}
> > > {CONST :consttype 23 :consttypmod -1 :constcollid 0 :constlen 4
> > > :constbyval true :constisnull false
> > > :location 53 :constvalue 4 [ 33 0 0 0 0 0 0 0 ]}) :location 51}")
> > > (2 rows)
> > >
> > > test_pub=# SELECT pg_get_publication_tables(VARIADIC
> > > array_agg(p.pubname)) from pg_publication p where pubname
> > > IN('pub1','pub2');
> > >
> > > pg_get_publication_tables
> > >
> > > ---------------------------------------------------------------------------------------------
> ----
> > > ---------------------------------------------------------------------
> > > ---------------------------------------------------------------------------------------------
> ----
> > > ---------------------------------------------------------------------
> > > -------------------------------------------------------------------
> > > (16385,1,"{OPEXPR :opno 521 :opfuncid 147 :opresulttype 16 :opretset
> > > false :opcollid 0 :inputcollid 0 :args ({VAR :varno 1 :varattno 1
> > > :vartype 23 :vartypmod -1 :var
> > > collid 0 :varlevelsup 0 :varnosyn 1 :varattnosyn 1 :location 47}
> > > {CONST :consttype 23 :consttypmod -1 :constcollid 0 :constlen 4
> > > :constbyval true :constisnull false :
> > > location 51 :constvalue 4 [ 99 0 0 0 0 0 0 0 ]}) :location 49}")
> > > (16385,"1 2","{OPEXPR :opno 97 :opfuncid 66 :opresulttype 16
> > > :opretset false :opcollid 0 :inputcollid 0 :args ({VAR :varno 1
> > > :varattno 2 :vartype 23 :vartypmod -1 :v
> > > arcollid 0 :varlevelsup 0 :varnosyn 1 :varattnosyn 2 :location 49}
> > > {CONST :consttype 23 :consttypmod -1 :constcollid 0 :constlen 4
> > > :constbyval true :constisnull false
> > > :location 53 :constvalue 4 [ 33 0 0 0 0 0 0 0 ]}) :location 51}")
> > > (2 rows)
> >
> > I think this is because the usage of SELECT statement. The order seems
> depend
> > on pg_publication. Such as:
> >
> > postgres=# SELECT array_agg(p.pubname) FROM pg_publication p WHERE
> pubname IN ('pub1','pub2');
> > array_agg
> > -------------
> > {pub1,pub2}
> > (1 row)
> >
> > postgres=# SELECT array_agg(p.pubname) FROM pg_publication p WHERE
> pubname IN ('pub2','pub1');
> > array_agg
> > -------------
> > {pub1,pub2}
> > (1 row)
> >
>
> Right, so I felt it was a bit dubious that the result of the function
> “seems to depend on” something. That’s why I was asking why the list
> elements did not include a pubid. Then a caller could be certain what
> element belonged with what publication. It's not quite clear to me why
> that is not important for this patch - but anyway, even if it's not
> necessary for this patch's usage, this is a function that is exposed
> to users who might have different needs/expectations than this patch
> has, so shouldn't the result be less fuzzy for them?
Yes, I agree that there may be such a need in the future. Added 'pubid' to the
output of this function.
BTW, I think the usage of the function pg_get_publication_tables is not
documented in the pg-doc now, it doesn't seem to be a function provided to
users. So I didn't modify the documentation.
Attach new patches.
Regards,
Wang wei
Attachments:
[application/octet-stream] HEAD_v14-0001-Fix-data-replicated-twice-when-specifying-publis.patch (21.4K, ../OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com/2-HEAD_v14-0001-Fix-data-replicated-twice-when-specifying-publis.patch)
download | inline diff:
From c343eb4c5bf23c3c565c6a5e73fa541e837afbca Mon Sep 17 00:00:00 2001
From: wangw <[email protected]>
Date: Tue, 14 Jun 2022 13:44:09 +0800
Subject: [PATCH v14 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 | 177 +++++++++++++++------
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, 222 insertions(+), 87 deletions(-)
diff --git a/src/backend/catalog/pg_publication.c b/src/backend/catalog/pg_publication.c
index 59967098b3..1a5dedc1a8 100644
--- a/src/backend/catalog/pg_publication.c
+++ b/src/backend/catalog/pg_publication.c
@@ -1026,22 +1026,35 @@ 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
+ typedef struct
+ {
+ Oid relid; /* OID of published table */
+ Oid pubid; /* OID of publication that publishes this
+ * table. */
+ } published_rel;
FuncCallContext *funcctx;
- char *pubname = text_to_cstring(PG_GETARG_TEXT_PP(0));
- Publication *publication;
- List *tables;
+ List *tables = NIL,
+ *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 +1062,127 @@ 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(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);
+ }
+
+ tables = list_concat(tables, pub_elem_tables);
+
+ /* At least one publication is using publish_via_partition_root. */
+ if (pub_elem->pubviaroot)
+ viaroot = true;
+ }
+
+ pfree(elems);
+
+ /* Now sort and de-duplicate the tables list */
+ list_sort(tables, list_oid_cmp);
+ list_deduplicate_oid(tables);
+
+ /*
+ * 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)
+ tables = filter_partitions(tables);
+
+ /*
+ * For tables that have been filtered out, delete the corresponding
+ * table information in the table_infos list.
+ */
+ foreach(lc, table_infos)
+ {
+ published_rel *table_info = (published_rel *) lfirst(lc);
+
+ if (!list_member_oid(tables, table_info->relid))
+ table_infos = foreach_delete_current(table_infos, lc);
}
/* 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};
@@ -1119,38 +1191,39 @@ 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;
}
rettuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c
index f0cec2ad5e..f3e002d380 100644
--- a/src/backend/commands/subscriptioncmds.c
+++ b/src/backend/commands/subscriptioncmds.c
@@ -1920,25 +1920,84 @@ 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 20f5aa56ea..2b2d45d838 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11615,11 +11615,13 @@
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 624d0e5aae..5d595db68b 100644
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -1446,7 +1446,7 @@ pg_publication_tables| SELECT p.pubname,
WHERE ((a.attrelid = gpt.relid) AND (a.attnum > 0) AND (NOT a.attisdropped) AND ((a.attnum = ANY ((gpt.attrs)::smallint[])) OR (gpt.attrs IS NULL)))) 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 8b33e4e7ae..100e2d2fd0 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 da52289dde..056ceb8096 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 ae022faa78..a39f3d27ff 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 245aea1dd1..d64b0f6390 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -3581,6 +3581,7 @@ pthread_mutex_t
pthread_once_t
pthread_t
ptrdiff_t
+published_rel
pull_var_clause_context
pull_varattnos_context
pull_varnos_context
--
2.23.0.windows.1
[application/octet-stream] HEAD_v14-0002-Add-clarification-for-the-behaviour-of-the-publi.patch (2.4K, ../OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com/3-HEAD_v14-0002-Add-clarification-for-the-behaviour-of-the-publi.patch)
download | inline diff:
From 6917fc349c36fe153b385bf53dfc8b2ed9c3754e Mon Sep 17 00:00:00 2001
From: wangw <[email protected]>
Date: Wed, 28 Sep 2022 14:35:41 +0800
Subject: [PATCH v14 2/2] Add clarification for the behaviour of the
publication parameter publish_via_partition_root
Assume a subscription is subscribing to multiple publications, and these
publications publish a partitioned table and its partitions respectively:
[publisher-side]
create table parent (a int primary key) partition by range (a);
create table child partition of parent default;
create publication pub1 for table parent;
create publication pub2 for table child;
[subscriber-side]
create subscription sub connection 'xxxx' publication pub1, pub2;
The manual does not clearly describe the behaviour when the user had
specified the parameter 'publish_via_partition_root' on just one of
the publications. This patch modifies documentation to clarify the
following rules:
- If the parameter publish_via_partition_root is specified only in pub1,
changes will be published using the identity and schema of the table 'parent'.
- If the parameter publish_via_partition_root is specified only in pub2,
changes will be published using the identity and schema of the table 'child'.
---
doc/src/sgml/ref/create_publication.sgml | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/doc/src/sgml/ref/create_publication.sgml b/doc/src/sgml/ref/create_publication.sgml
index e229384e6f..74d50dd9fb 100644
--- a/doc/src/sgml/ref/create_publication.sgml
+++ b/doc/src/sgml/ref/create_publication.sgml
@@ -201,6 +201,16 @@ CREATE PUBLICATION <replaceable class="parameter">name</replaceable>
consisting of a different set of partitions.
</para>
+ <para>
+ There can be a case where a subscription combines multiple
+ publications. If a root partitioned table is published by any
+ subscribed publications which set
+ <literal>publish_via_partition_root</literal> = true, changes on this
+ root partitioned table (or on its partitions) will be published using
+ the identity and schema of this root partitioned table rather than
+ that of the individual partitions.
+ </para>
+
<para>
This parameter also affects how row filters and column lists are
chosen for partitions; see below for details.
--
2.23.0.windows.1
[application/octet-stream] REL15_v14-0001-Fix-data-replicated-twice-when-specifying-publis_patch (8.6K, ../OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com/4-REL15_v14-0001-Fix-data-replicated-twice-when-specifying-publis_patch)
download
[application/octet-stream] REL14_v14-0001-Fix-data-replicated-twice-when-specifying-publis_patch (5.3K, ../OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com/5-REL14_v14-0001-Fix-data-replicated-twice-when-specifying-publis_patch)
download
view thread (91+ messages) latest in thread
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Subject: RE: Data is copied twice when specifying both child and parent table in publication
In-Reply-To: <OS3PR01MB6275FB5397C6A647F262A3A69E009@OS3PR01MB6275.jpnprd01.prod.outlook.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox