public inbox for [email protected]
help / color / mirror / Atom feedFrom: [email protected] <[email protected]>
To: Amit Kapila <[email protected]>
To: [email protected] <[email protected]>
To: [email protected] <[email protected]>
Cc: Tomas Vondra <[email protected]>
Cc: Alvaro Herrera <[email protected]>
Cc: Peter Eisentraut <[email protected]>
Cc: Pg Hackers <[email protected]>
Subject: RE: bogus: logical replication rows/cols combinations
Date: Tue, 17 May 2022 09:10:00 +0000
Message-ID: <OS0PR01MB571647F2739F710E3D1B045D94CE9@OS0PR01MB5716.jpnprd01.prod.outlook.com> (raw)
In-Reply-To: <CAA4eK1JhCpGdvBbyvs9RwJZeKe5TX1Whd6fSQ2VGUh6AUdWL2Q@mail.gmail.com>
References: <CAA4eK1KQimd=0gLS1DHTK2HrZA3kPdaY3sDY9Jko8o-nXRQw9A@mail.gmail.com>
<[email protected]>
<CAA4eK1Le0Z6W_+oY+M4vV=fjwFMJRg3WguPr3LQAfEqATrSLbg@mail.gmail.com>
<[email protected]>
<CAA4eK1JJ_cJZFanVFDC-1ipKC9aounno5JcWx1ceH035FZZ1MA@mail.gmail.com>
<[email protected]>
<CAA4eK1JC8v8ysNS9L1et0_yyZJOEE1wB+GrvOcar_f5dVadvBg@mail.gmail.com>
<OS0PR01MB5716A594C58DE4FFD1F8100B94C89@OS0PR01MB5716.jpnprd01.prod.outlook.com>
<CAA4eK1KYp1408g9P54iVX2uPwCvHR8R_8_H1oRZ9bv2iJvOOSw@mail.gmail.com>
<OS0PR01MB571698DEFCE9ACC6DC40510794CA9@OS0PR01MB5716.jpnprd01.prod.outlook.com>
<CAA4eK1+UCfKUaf41H6zHS2qR4CZoVtZe64yx_T-JVPSb1pdw_w@mail.gmail.com>
<OS0PR01MB5716D2CF3F2E39A6246A76CE94CF9@OS0PR01MB5716.jpnprd01.prod.outlook.com>
<CAA4eK1JhCpGdvBbyvs9RwJZeKe5TX1Whd6fSQ2VGUh6AUdWL2Q@mail.gmail.com>
On Tuesday, May 17, 2022 2:53 PM Amit Kapila <[email protected]> wrote:
>
> Few minor comments:
> ==================
> 1.
> + <para>
> + Names of table columns included in the publication. This contains all
> + the columns of table when user didn't specify column list for the
> + table.
> + </para></entry>
>
> Can we slightly change it to: "Names of table columns included in the
> publication. This contains all the columns of the table when the user
> didn't specify the column list for the table."
>
> 2. Below comments needs to be removed from tablesync.c as we don't
> combine column lists after this patch.
>
> * For initial synchronization, column lists can be ignored in following
> * cases:
> *
> * 1) one of the subscribed publications for the table hasn't specified
> * any column list
> *
> * 2) one of the subscribed publications has puballtables set to true
> *
> * 3) one of the subscribed publications is declared as ALL TABLES IN
> * SCHEMA that includes this relation
>
> 3.
> Note that we don't support the case where column list is different for
> + * the same table when combining publications. But we still need to check
> + * all the given publication-table mappings and report an error if any
> + * publications have different column list.
>
> Can we change this comment to: "Note that we don't support the case
> where the column list is different for the same table when combining
> publications. But one can later change the publication so we still
> need to check all the given publication-table mappings and report an
> error if any publications have a different column list."?
>
> 4. Can we add a test for different column lists if it is not already there?
Thanks for the comments.
Attach the new version patch which addressed all the above comments and
comments from Shi yu[1] and Osumi-san[2].
[1] https://www.postgresql.org/message-id/OSZPR01MB6310F32344884F9C12F45071FDCE9%40OSZPR01MB6310.jpnprd0...
[2] https://www.postgresql.org/message-id/TYCPR01MB83736AEC2493FCBB75CC7556EDCE9%40TYCPR01MB8373.jpnprd0...
Best regards,
Hou zj
Attachments:
[application/octet-stream] v3-0001-Extend-pg_publication_tables-to-display-column-list-.patch (15.0K, ../OS0PR01MB571647F2739F710E3D1B045D94CE9@OS0PR01MB5716.jpnprd01.prod.outlook.com/2-v3-0001-Extend-pg_publication_tables-to-display-column-list-.patch)
download | inline diff:
From 87104fadcbdba46874a52e72e844ed4c3a73de1d Mon Sep 17 00:00:00 2001
From: "houzj.fnst" <[email protected]>
Date: Fri, 13 May 2022 11:35:14 +0800
Subject: [PATCH] Extend pg_publication_tables to display column list and row
filter.
Commit 923def9a53 and 52e4f0cd47 allowed to specify column lists and row
filters for publication tables. This commit extends the
pg_publication_tables view and pg_get_publication_tables function to
display that information.
This information will be useful to users and we also need this for the
later commit that prohibits combining multiple publications with different
column lists for the same table.
Author: Hou Zhijie
Reviewed By: Amit Kapila, Alvaro Herrera, Shi Yu, Takamichi Osumi
Discussion: https://postgr.es/m/[email protected]
---
doc/src/sgml/catalogs.sgml | 27 +++++++++++++--
src/backend/catalog/pg_publication.c | 54 +++++++++++++++++++++++++++--
src/backend/catalog/system_views.sql | 10 +++++-
src/backend/replication/logical/tablesync.c | 14 +++-----
src/include/catalog/pg_proc.dat | 8 ++---
src/test/regress/expected/publication.out | 42 +++++++++++-----------
src/test/regress/expected/rules.out | 13 +++++--
7 files changed, 125 insertions(+), 43 deletions(-)
diff --git a/doc/src/sgml/catalogs.sgml b/doc/src/sgml/catalogs.sgml
index a533a21..d96c72e 100644
--- a/doc/src/sgml/catalogs.sgml
+++ b/doc/src/sgml/catalogs.sgml
@@ -9691,7 +9691,7 @@ SCRAM-SHA-256$<replaceable><iteration count></replaceable>:<replaceable>&l
<row>
<entry><link linkend="view-pg-publication-tables"><structname>pg_publication_tables</structname></link></entry>
- <entry>publications and their associated tables</entry>
+ <entry>publications and information of their associated tables</entry>
</row>
<row>
@@ -11635,8 +11635,8 @@ SELECT * FROM pg_locks pl LEFT JOIN pg_prepared_xacts ppx
<para>
The view <structname>pg_publication_tables</structname> provides
- information about the mapping between publications and the tables they
- contain. Unlike the underlying catalog
+ information about the mapping between publications and information of
+ tables they contain. Unlike the underlying catalog
<link linkend="catalog-pg-publication-rel"><structname>pg_publication_rel</structname></link>,
this view expands publications defined as <literal>FOR ALL TABLES</literal>
and <literal>FOR ALL TABLES IN SCHEMA</literal>, so for such publications
@@ -11687,6 +11687,27 @@ SELECT * FROM pg_locks pl LEFT JOIN pg_prepared_xacts ppx
Name of table
</para></entry>
</row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>attnames</structfield> <type>name[]</type>
+ (references <link linkend="catalog-pg-attribute"><structname>pg_attribute</structname></link>.<structfield>attname</structfield>)
+ </para>
+ <para>
+ Names of table columns included in the publication. This contains all
+ the columns of the table when the user didn't specify the column list
+ for the table.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>rowfilter</structfield> <type>text</type>
+ </para>
+ <para>
+ Expression for the table's publication qualifying condition
+ </para></entry>
+ </row>
</tbody>
</tgroup>
</table>
diff --git a/src/backend/catalog/pg_publication.c b/src/backend/catalog/pg_publication.c
index e2c8bcb..152b4ba 100644
--- a/src/backend/catalog/pg_publication.c
+++ b/src/backend/catalog/pg_publication.c
@@ -1077,11 +1077,12 @@ get_publication_name(Oid pubid, bool missing_ok)
}
/*
- * Returns Oids of tables in a publication.
+ * Returns information of tables in a publication.
*/
Datum
pg_get_publication_tables(PG_FUNCTION_ARGS)
{
+#define NUM_PUBLICATOIN_TABLES_ELEM 3
FuncCallContext *funcctx;
char *pubname = text_to_cstring(PG_GETARG_TEXT_PP(0));
Publication *publication;
@@ -1090,6 +1091,7 @@ pg_get_publication_tables(PG_FUNCTION_ARGS)
/* stuff done only on the first call of the function */
if (SRF_IS_FIRSTCALL())
{
+ TupleDesc tupdesc;
MemoryContext oldcontext;
/* create a function context for cross-call persistence */
@@ -1136,6 +1138,16 @@ pg_get_publication_tables(PG_FUNCTION_ARGS)
tables = filter_partitions(tables);
}
+ /* Construct a tuple descriptor for the result rows. */
+ tupdesc = CreateTemplateTupleDesc(NUM_PUBLICATOIN_TABLES_ELEM);
+ TupleDescInitEntry(tupdesc, (AttrNumber) 1, "relid",
+ OIDOID, -1, 0);
+ TupleDescInitEntry(tupdesc, (AttrNumber) 2, "attrs",
+ INT2VECTOROID, -1, 0);
+ TupleDescInitEntry(tupdesc, (AttrNumber) 3, "qual",
+ PG_NODE_TREEOID, -1, 0);
+
+ funcctx->tuple_desc = BlessTupleDesc(tupdesc);
funcctx->user_fctx = (void *) tables;
MemoryContextSwitchTo(oldcontext);
@@ -1147,9 +1159,47 @@ pg_get_publication_tables(PG_FUNCTION_ARGS)
if (funcctx->call_cntr < list_length(tables))
{
+ HeapTuple pubtuple = NULL;
+ HeapTuple rettuple;
Oid relid = list_nth_oid(tables, funcctx->call_cntr);
+ Datum values[NUM_PUBLICATOIN_TABLES_ELEM];
+ bool nulls[NUM_PUBLICATOIN_TABLES_ELEM];
+
+ /*
+ * Form tuple with appropriate data.
+ */
+ MemSet(nulls, 0, sizeof(nulls));
+ MemSet(values, 0, sizeof(values));
+
+ publication = GetPublicationByName(pubname, false);
+
+ values[0] = ObjectIdGetDatum(relid);
+
+ pubtuple = SearchSysCacheCopy2(PUBLICATIONRELMAP,
+ ObjectIdGetDatum(relid),
+ ObjectIdGetDatum(publication->oid));
+
+ if (HeapTupleIsValid(pubtuple))
+ {
+ /* Lookup the column list attribute. */
+ values[1] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple,
+ Anum_pg_publication_rel_prattrs,
+ &(nulls[1]));
+
+ /* Null indicates no filter. */
+ values[2] = SysCacheGetAttr(PUBLICATIONRELMAP, pubtuple,
+ Anum_pg_publication_rel_prqual,
+ &(nulls[2]));
+ }
+ else
+ {
+ nulls[1] = true;
+ nulls[2] = true;
+ }
+
+ rettuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
- SRF_RETURN_NEXT(funcctx, ObjectIdGetDatum(relid));
+ SRF_RETURN_NEXT(funcctx, HeapTupleGetDatum(rettuple));
}
SRF_RETURN_DONE(funcctx);
diff --git a/src/backend/catalog/system_views.sql b/src/backend/catalog/system_views.sql
index 0fc614e..fedaed5 100644
--- a/src/backend/catalog/system_views.sql
+++ b/src/backend/catalog/system_views.sql
@@ -368,7 +368,15 @@ CREATE VIEW pg_publication_tables AS
SELECT
P.pubname AS pubname,
N.nspname AS schemaname,
- C.relname AS tablename
+ C.relname AS tablename,
+ ( SELECT array_agg(a.attname ORDER BY a.attnum)
+ FROM unnest(CASE WHEN GPT.attrs IS NOT NULL THEN GPT.attrs
+ ELSE (SELECT array_agg(g) FROM generate_series(1, C.relnatts) g)
+ END) k
+ JOIN pg_attribute a
+ ON (a.attrelid = GPT.relid AND a.attnum = k)
+ ) AS attnames,
+ pg_get_expr(GPT.qual, GPT.relid) AS rowfilter
FROM pg_publication P,
LATERAL pg_get_publication_tables(P.pubname) GPT,
pg_class C JOIN pg_namespace N ON (N.oid = C.relnamespace)
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index b03e0f5..994c7a0 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -795,15 +795,12 @@ fetch_remote_table_info(char *nspname, char *relname,
resetStringInfo(&cmd);
appendStringInfo(&cmd,
"SELECT DISTINCT unnest"
- " FROM pg_publication p"
- " LEFT OUTER JOIN pg_publication_rel pr"
- " ON (p.oid = pr.prpubid AND pr.prrelid = %u)"
- " LEFT OUTER JOIN unnest(pr.prattrs) ON TRUE,"
+ " FROM pg_publication p,"
" LATERAL pg_get_publication_tables(p.pubname) gpt"
+ " LEFT OUTER JOIN unnest(gpt.attrs) ON TRUE"
" WHERE gpt.relid = %u"
" AND p.pubname IN ( %s )",
lrel->remoteid,
- lrel->remoteid,
pub_names.data);
pubres = walrcv_exec(LogRepWorkerWalRcvConn, cmd.data,
@@ -965,15 +962,12 @@ fetch_remote_table_info(char *nspname, char *relname,
/* Check for row filters. */
resetStringInfo(&cmd);
appendStringInfo(&cmd,
- "SELECT DISTINCT pg_get_expr(pr.prqual, pr.prrelid)"
- " FROM pg_publication p"
- " LEFT OUTER JOIN pg_publication_rel pr"
- " ON (p.oid = pr.prpubid AND pr.prrelid = %u),"
+ "SELECT DISTINCT pg_get_expr(gpt.qual, gpt.relid)"
+ " FROM pg_publication p,"
" LATERAL pg_get_publication_tables(p.pubname) gpt"
" WHERE gpt.relid = %u"
" AND p.pubname IN ( %s )",
lrel->remoteid,
- lrel->remoteid,
pub_names.data);
res = walrcv_exec(LogRepWorkerWalRcvConn, cmd.data, 1, qualRow);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index babe16f..87aa571 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11673,11 +11673,11 @@
prosrc => 'pg_show_replication_origin_status' },
# publications
-{ oid => '6119', descr => 'get OIDs of tables in a publication',
+{ oid => '6119', descr => 'get information of tables in a publication',
proname => 'pg_get_publication_tables', prorows => '1000', proretset => 't',
- provolatile => 's', prorettype => 'oid', proargtypes => 'text',
- proallargtypes => '{text,oid}', proargmodes => '{i,o}',
- proargnames => '{pubname,relid}', prosrc => 'pg_get_publication_tables' },
+ provolatile => 's', prorettype => 'record', proargtypes => 'text',
+ proallargtypes => '{text,oid,int2vector,pg_node_tree}', proargmodes => '{i,o,o,o}',
+ proargnames => '{pubname,relid,attrs,qual}', prosrc => 'pg_get_publication_tables' },
{ oid => '6121',
descr => 'returns whether a relation can be part of a publication',
proname => 'pg_relation_is_publishable', provolatile => 's',
diff --git a/src/test/regress/expected/publication.out b/src/test/regress/expected/publication.out
index 398c0f3..274b37d 100644
--- a/src/test/regress/expected/publication.out
+++ b/src/test/regress/expected/publication.out
@@ -1585,52 +1585,52 @@ CREATE TABLE sch2.tbl1_part1 PARTITION OF sch1.tbl1 FOR VALUES FROM (1) to (10);
-- Schema publication that does not include the schema that has the parent table
CREATE PUBLICATION pub FOR ALL TABLES IN SCHEMA sch2 WITH (PUBLISH_VIA_PARTITION_ROOT=1);
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+------------
- pub | sch2 | tbl1_part1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+------------+----------+-----------
+ pub | sch2 | tbl1_part1 | {a} |
(1 row)
DROP PUBLICATION pub;
-- Table publication that does not include the parent table
CREATE PUBLICATION pub FOR TABLE sch2.tbl1_part1 WITH (PUBLISH_VIA_PARTITION_ROOT=1);
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+------------
- pub | sch2 | tbl1_part1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+------------+----------+-----------
+ pub | sch2 | tbl1_part1 | {a} |
(1 row)
-- Table publication that includes both the parent table and the child table
ALTER PUBLICATION pub ADD TABLE sch1.tbl1;
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+-----------
- pub | sch1 | tbl1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+-----------+----------+-----------
+ pub | sch1 | tbl1 | {a} |
(1 row)
DROP PUBLICATION pub;
-- Schema publication that does not include the schema that has the parent table
CREATE PUBLICATION pub FOR ALL TABLES IN SCHEMA sch2 WITH (PUBLISH_VIA_PARTITION_ROOT=0);
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+------------
- pub | sch2 | tbl1_part1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+------------+----------+-----------
+ pub | sch2 | tbl1_part1 | {a} |
(1 row)
DROP PUBLICATION pub;
-- Table publication that does not include the parent table
CREATE PUBLICATION pub FOR TABLE sch2.tbl1_part1 WITH (PUBLISH_VIA_PARTITION_ROOT=0);
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+------------
- pub | sch2 | tbl1_part1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+------------+----------+-----------
+ pub | sch2 | tbl1_part1 | {a} |
(1 row)
-- Table publication that includes both the parent table and the child table
ALTER PUBLICATION pub ADD TABLE sch1.tbl1;
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+------------
- pub | sch2 | tbl1_part1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+------------+----------+-----------
+ pub | sch2 | tbl1_part1 | {a} |
(1 row)
DROP PUBLICATION pub;
@@ -1643,9 +1643,9 @@ CREATE TABLE sch1.tbl1_part3 (a int) PARTITION BY RANGE(a);
ALTER TABLE sch1.tbl1 ATTACH PARTITION sch1.tbl1_part3 FOR VALUES FROM (20) to (30);
CREATE PUBLICATION pub FOR ALL TABLES IN SCHEMA sch1 WITH (PUBLISH_VIA_PARTITION_ROOT=1);
SELECT * FROM pg_publication_tables;
- pubname | schemaname | tablename
----------+------------+-----------
- pub | sch1 | tbl1
+ pubname | schemaname | tablename | attnames | rowfilter
+---------+------------+-----------+----------+-----------
+ pub | sch1 | tbl1 | {a} |
(1 row)
RESET client_min_messages;
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
index 21effe8..fc3cde3 100644
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -1437,9 +1437,18 @@ pg_prepared_xacts| SELECT p.transaction,
LEFT JOIN pg_database d ON ((p.dbid = d.oid)));
pg_publication_tables| SELECT p.pubname,
n.nspname AS schemaname,
- c.relname AS tablename
+ c.relname AS tablename,
+ ( SELECT array_agg(a.attname ORDER BY a.attnum) AS array_agg
+ FROM (unnest(
+ CASE
+ WHEN (gpt.attrs IS NOT NULL) THEN (gpt.attrs)::integer[]
+ ELSE ( SELECT array_agg(g.g) AS array_agg
+ FROM generate_series(1, (c.relnatts)::integer) g(g))
+ END) k(k)
+ JOIN pg_attribute a ON (((a.attrelid = gpt.relid) AND (a.attnum = k.k))))) 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),
+ LATERAL pg_get_publication_tables((p.pubname)::text) gpt(relid, attrs, qual),
(pg_class c
JOIN pg_namespace n ON ((n.oid = c.relnamespace)))
WHERE (c.oid = gpt.relid);
--
2.7.2.windows.1
[application/octet-stream] v3-0002-Prohibit-combining-publications-with-different-colum.patch (22.0K, ../OS0PR01MB571647F2739F710E3D1B045D94CE9@OS0PR01MB5716.jpnprd01.prod.outlook.com/3-v3-0002-Prohibit-combining-publications-with-different-colum.patch)
download | inline diff:
From 91a4774b9f351cb09f461ee920be9031cc4e073e Mon Sep 17 00:00:00 2001
From: "houzj.fnst" <[email protected]>
Date: Fri, 13 May 2022 13:19:49 +0800
Subject: [PATCH] Prohibit combining publications with different column lists.
The main purpose of introducing a column list is to have statically
different shapes on publisher and subscriber or hide sensitive column
data. In both cases, it doesn't seem to make sense to combine column
lists.
So, we disallow the cases where the column list is different for the same
table when combining publications. It can be later extended to combine the
column lists for selective cases where required.
Reported-by: Alvaro Herrera
Author: Hou Zhijie
Reviewed-by: Amit Kapila, Alvaro Herrera, Shi Yu, Takamichi Osumi
Discussion: https://postgr.es/m/[email protected]
---
doc/src/sgml/ref/alter_publication.sgml | 12 ++-
doc/src/sgml/ref/create_subscription.sgml | 5 +
src/backend/commands/subscriptioncmds.c | 23 ++++-
src/backend/replication/logical/tablesync.c | 72 ++++++-------
src/backend/replication/pgoutput/pgoutput.c | 80 +++++++--------
src/test/subscription/t/031_column_list.pl | 150 ++++++++++++----------------
6 files changed, 176 insertions(+), 166 deletions(-)
diff --git a/doc/src/sgml/ref/alter_publication.sgml b/doc/src/sgml/ref/alter_publication.sgml
index e2cce49..f03933a 100644
--- a/doc/src/sgml/ref/alter_publication.sgml
+++ b/doc/src/sgml/ref/alter_publication.sgml
@@ -116,7 +116,17 @@ ALTER PUBLICATION <replaceable class="parameter">name</replaceable> RENAME TO <r
<para>
Optionally, a column list can be specified. See <xref
- linkend="sql-createpublication"/> for details.
+ linkend="sql-createpublication"/> for details. Note that a subscription
+ having several publications in which the same table has been published
+ with different column lists is not supported. So, changing the column
+ lists of the tables being subscribed could cause inconsistency of column
+ lists among publications in which case <command>ALTER PUBLICATION</command>
+ command will be successful but later the WalSender in publisher or the
+ subscriber may throw an error. In this scenario, the user needs to
+ recreate the subscription after adjusting the column list or drop the
+ problematic publication using
+ <literal>ALTER SUBSCRIPTION ... DROP PUBLICATION</literal> and then add
+ it back after adjusting the column list.
</para>
<para>
diff --git a/doc/src/sgml/ref/create_subscription.sgml b/doc/src/sgml/ref/create_subscription.sgml
index 203bb41..f6f82a0 100644
--- a/doc/src/sgml/ref/create_subscription.sgml
+++ b/doc/src/sgml/ref/create_subscription.sgml
@@ -356,6 +356,11 @@ CREATE SUBSCRIPTION <replaceable class="parameter">subscription_name</replaceabl
</para>
<para>
+ Subscription having several publications in which the same table has been
+ published with different column lists is not supported.
+ </para>
+
+ <para>
We allow non-existent publications to be specified so that users can add
those later. This means
<link linkend="catalog-pg-subscription"><structname>pg_subscription</structname></link>
diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c
index 690cdaa..991b2c1 100644
--- a/src/backend/commands/subscriptioncmds.c
+++ b/src/backend/commands/subscriptioncmds.c
@@ -1753,7 +1753,8 @@ AlterSubscriptionOwner_oid(Oid subid, Oid newOwnerId)
/*
* Get the list of tables which belong to specified publications on the
- * publisher connection.
+ * publisher connection. Also get the column list for each table and check if
+ * column lists are the same in different publications.
*/
static List *
fetch_table_list(WalReceiverConn *wrconn, List *publications)
@@ -1761,17 +1762,18 @@ fetch_table_list(WalReceiverConn *wrconn, List *publications)
WalRcvExecResult *res;
StringInfoData cmd;
TupleTableSlot *slot;
- Oid tableRow[2] = {TEXTOID, TEXTOID};
+ Oid tableRow[3] = {TEXTOID, TEXTOID, NAMEARRAYOID};
List *tablelist = NIL;
initStringInfo(&cmd);
- appendStringInfoString(&cmd, "SELECT DISTINCT t.schemaname, t.tablename\n"
+ appendStringInfoString(&cmd, "SELECT DISTINCT t.schemaname, t.tablename, \n"
+ " t.attnames\n"
" FROM pg_catalog.pg_publication_tables t\n"
" WHERE t.pubname IN (");
get_publications_str(publications, &cmd, true);
appendStringInfoChar(&cmd, ')');
- res = walrcv_exec(wrconn, cmd.data, 2, tableRow);
+ res = walrcv_exec(wrconn, cmd.data, 3, tableRow);
pfree(cmd.data);
if (res->status != WALRCV_OK_TUPLES)
@@ -1795,7 +1797,18 @@ fetch_table_list(WalReceiverConn *wrconn, List *publications)
Assert(!isnull);
rv = makeRangeVar(nspname, relname, -1);
- tablelist = lappend(tablelist, rv);
+
+ /*
+ * We don't support the case where column list is different for the
+ * same table in different publications.
+ */
+ if (list_member(tablelist, rv))
+ ereport(ERROR,
+ errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot use different column lists for table \"%s.%s\" in different publications",
+ nspname, relname));
+ else
+ tablelist = lappend(tablelist, rv);
ExecClearTuple(slot);
}
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 994c7a0..afc5bcd 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -753,17 +753,6 @@ fetch_remote_table_info(char *nspname, char *relname,
/*
* Get column lists for each relation.
*
- * For initial synchronization, column lists can be ignored in following
- * cases:
- *
- * 1) one of the subscribed publications for the table hasn't specified
- * any column list
- *
- * 2) one of the subscribed publications has puballtables set to true
- *
- * 3) one of the subscribed publications is declared as ALL TABLES IN
- * SCHEMA that includes this relation
- *
* We need to do this before fetching info about column names and types,
* so that we can skip columns that should not be replicated.
*/
@@ -771,7 +760,7 @@ fetch_remote_table_info(char *nspname, char *relname,
{
WalRcvExecResult *pubres;
TupleTableSlot *slot;
- Oid attrsRow[] = {INT2OID};
+ Oid attrsRow[] = {INT2VECTOROID};
StringInfoData pub_names;
bool first = true;
@@ -786,19 +775,17 @@ fetch_remote_table_info(char *nspname, char *relname,
/*
* Fetch info about column lists for the relation (from all the
- * publications). We unnest the int2vector values, because that makes
- * it easier to combine lists by simply adding the attnums to a new
- * bitmap (without having to parse the int2vector data). This
- * preserves NULL values, so that if one of the publications has no
- * column list, we'll know that.
+ * publications).
*/
resetStringInfo(&cmd);
appendStringInfo(&cmd,
- "SELECT DISTINCT unnest"
+ "SELECT DISTINCT"
+ " (CASE WHEN (array_length(gpt.attrs, 1) = c.relnatts)"
+ " THEN NULL ELSE gpt.attrs END)"
" FROM pg_publication p,"
- " LATERAL pg_get_publication_tables(p.pubname) gpt"
- " LEFT OUTER JOIN unnest(gpt.attrs) ON TRUE"
- " WHERE gpt.relid = %u"
+ " LATERAL pg_get_publication_tables(p.pubname) gpt,"
+ " pg_class c"
+ " WHERE gpt.relid = %u AND c.oid = gpt.relid"
" AND p.pubname IN ( %s )",
lrel->remoteid,
pub_names.data);
@@ -813,26 +800,43 @@ fetch_remote_table_info(char *nspname, char *relname,
nspname, relname, pubres->err)));
/*
- * Merge the column lists (from different publications) by creating a
- * single bitmap with all the attnums. If we find a NULL value, that
- * means one of the publications has no column list for the table
- * we're syncing.
+ * We don't support the case where the column list is different for the
+ * same table when combining publications. So there should be only one
+ * row returned. Although we already checked this when creating
+ * subscription, we still need to check here in case the column list
+ * was changed after creating the subscription and before the sync
+ * worker is started.
+ */
+ if (tuplestore_tuple_count(pubres->tuplestore) > 1)
+ ereport(ERROR,
+ errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot use different column lists for table \"%s.%s\" in different publications",
+ nspname, relname));
+
+ /*
+ * Get the column list and build a single bitmap with the attnums.
+ *
+ * If we find a NULL value, it means all the columns should be
+ * replicated.
*/
slot = MakeSingleTupleTableSlot(pubres->tupledesc, &TTSOpsMinimalTuple);
- while (tuplestore_gettupleslot(pubres->tuplestore, true, false, slot))
+ if (tuplestore_gettupleslot(pubres->tuplestore, true, false, slot))
{
Datum cfval = slot_getattr(slot, 1, &isnull);
- /* NULL means empty column list, so we're done. */
- if (isnull)
+ if (!isnull)
{
- bms_free(included_cols);
- included_cols = NULL;
- break;
- }
+ ArrayType *arr;
+ int nelems;
+ int16 *elems;
- included_cols = bms_add_member(included_cols,
- DatumGetInt16(cfval));
+ arr = DatumGetArrayTypeP(cfval);
+ nelems = ARR_DIMS(arr)[0];
+ elems = (int16 *) ARR_DATA_PTR(arr);
+
+ for (natt = 0; natt < nelems; natt++)
+ included_cols = bms_add_member(included_cols, elems[natt]);
+ }
ExecClearTuple(slot);
}
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index 42c06af..e9079bd 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -979,30 +979,31 @@ pgoutput_column_list_init(PGOutputData *data, List *publications,
RelationSyncEntry *entry)
{
ListCell *lc;
+ bool first = true;
+ Relation relation = RelationIdGetRelation(entry->publish_as_relid);
/*
* Find if there are any column lists for this relation. If there are,
- * build a bitmap merging all the column lists.
- *
- * All the given publication-table mappings must be checked.
+ * build a bitmap using the column lists.
*
* Multiple publications might have multiple column lists for this
* relation.
*
+ * Note that we don't support the case where the column list is different
+ * for the same table when combining publications. But one can later change
+ * the publication so we still need to check all the given
+ * publication-table mappings and report an error if any publications have
+ * a different column list.
+ *
* FOR ALL TABLES and FOR ALL TABLES IN SCHEMA implies "don't use column
- * list" so it takes precedence.
+ * list".
*/
foreach(lc, publications)
{
Publication *pub = lfirst(lc);
HeapTuple cftuple = NULL;
Datum cfdatum = 0;
-
- /*
- * Assume there's no column list. Only if we find pg_publication_rel
- * entry with a column list we'll switch it to false.
- */
- bool pub_no_list = true;
+ Bitmapset *cols = NULL;
/*
* If the publication is FOR ALL TABLES then it is treated the same as
@@ -1011,6 +1012,8 @@ pgoutput_column_list_init(PGOutputData *data, List *publications,
*/
if (!pub->alltables)
{
+ bool pub_no_list = true;
+
/*
* Check for the presence of a column list in this publication.
*
@@ -1024,51 +1027,48 @@ pgoutput_column_list_init(PGOutputData *data, List *publications,
if (HeapTupleIsValid(cftuple))
{
- /*
- * Lookup the column list attribute.
- *
- * Note: We update the pub_no_list value directly, because if
- * the value is NULL, we have no list (and vice versa).
- */
+ /* Lookup the column list attribute. */
cfdatum = SysCacheGetAttr(PUBLICATIONRELMAP, cftuple,
Anum_pg_publication_rel_prattrs,
&pub_no_list);
- /*
- * Build the column list bitmap in the per-entry context.
- *
- * We need to merge column lists from all publications, so we
- * update the same bitmapset. If the column list is null, we
- * interpret it as replicating all columns.
- */
+ /* Build the column list bitmap in the per-entry context. */
if (!pub_no_list) /* when not null */
{
pgoutput_ensure_entry_cxt(data, entry);
- entry->columns = pub_collist_to_bitmapset(entry->columns,
- cfdatum,
- entry->entry_cxt);
+ cols = pub_collist_to_bitmapset(cols, cfdatum,
+ entry->entry_cxt);
+
+ /*
+ * If column list includes all the columns of the table,
+ * set it to NULL.
+ */
+ if (bms_num_members(cols) == RelationGetNumberOfAttributes(relation))
+ {
+ bms_free(cols);
+ cols = NULL;
+ }
}
+
+ ReleaseSysCache(cftuple);
}
}
- /*
- * Found a publication with no column list, so we're done. But first
- * discard column list we might have from preceding publications.
- */
- if (pub_no_list)
+ if (first)
{
- if (cftuple)
- ReleaseSysCache(cftuple);
-
- bms_free(entry->columns);
- entry->columns = NULL;
-
- break;
+ entry->columns = cols;
+ first = false;
}
-
- ReleaseSysCache(cftuple);
+ else if (!bms_equal(entry->columns, cols))
+ ereport(ERROR,
+ errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot use different column lists for table \"%s.%s\" in different publications",
+ get_namespace_name(RelationGetNamespace(relation)),
+ RelationGetRelationName(relation)));
} /* loop all subscribed publications */
+
+ RelationClose(relation);
}
/*
diff --git a/src/test/subscription/t/031_column_list.pl b/src/test/subscription/t/031_column_list.pl
index 19812e1..7f031bc 100644
--- a/src/test/subscription/t/031_column_list.pl
+++ b/src/test/subscription/t/031_column_list.pl
@@ -20,6 +20,7 @@ $node_subscriber->append_conf('postgresql.conf',
$node_subscriber->start;
my $publisher_connstr = $node_publisher->connstr . ' dbname=postgres';
+my $offset = 0;
sub wait_for_subscription_sync
{
@@ -361,13 +362,13 @@ is( $result, qq(1|abc
2|xyz), 'update on column tab2.c is not replicated');
-# TEST: add a table to two publications with different column lists, and
+# TEST: add a table to two publications with same column lists, and
# create a single subscription replicating both publications
$node_publisher->safe_psql(
'postgres', qq(
CREATE TABLE tab5 (a int PRIMARY KEY, b int, c int, d int);
CREATE PUBLICATION pub2 FOR TABLE tab5 (a, b);
- CREATE PUBLICATION pub3 FOR TABLE tab5 (a, d);
+ CREATE PUBLICATION pub3 FOR TABLE tab5 (a, b);
-- insert a couple initial records
INSERT INTO tab5 VALUES (1, 11, 111, 1111);
@@ -388,8 +389,7 @@ wait_for_subscription_sync($node_subscriber);
$node_publisher->wait_for_catchup('sub1');
-# insert data and make sure all the columns (union of the columns lists)
-# get fully replicated
+# insert data and make sure the columns in column list get fully replicated
$node_publisher->safe_psql(
'postgres', qq(
INSERT INTO tab5 VALUES (3, 33, 333, 3333);
@@ -399,42 +399,12 @@ $node_publisher->safe_psql(
$node_publisher->wait_for_catchup('sub1');
is( $node_subscriber->safe_psql('postgres', "SELECT * FROM tab5 ORDER BY a"),
- qq(1|11|1111
-2|22|2222
-3|33|3333
-4|44|4444),
+ qq(1|11|
+2|22|
+3|33|
+4|44|),
'overlapping publications with overlapping column lists');
-# and finally, remove the column list for one of the publications, which
-# means replicating all columns (removing the column list), but first add
-# the missing column to the table on subscriber
-$node_publisher->safe_psql(
- 'postgres', qq(
- ALTER PUBLICATION pub3 SET TABLE tab5;
-));
-
-$node_subscriber->safe_psql(
- 'postgres', qq(
- ALTER SUBSCRIPTION sub1 REFRESH PUBLICATION;
- ALTER TABLE tab5 ADD COLUMN c INT;
-));
-
-wait_for_subscription_sync($node_subscriber);
-
-$node_publisher->safe_psql(
- 'postgres', qq(
- INSERT INTO tab5 VALUES (5, 55, 555, 5555);
-));
-
-$node_publisher->wait_for_catchup('sub1');
-
-is( $node_subscriber->safe_psql('postgres', "SELECT * FROM tab5 ORDER BY a"),
- qq(1|11|1111|
-2|22|2222|
-3|33|3333|
-4|44|4444|
-5|55|5555|555),
- 'overlapping publications with overlapping column lists');
# TEST: create a table with a column list, then change the replica
# identity by replacing a primary key (but use a different column in
@@ -900,57 +870,21 @@ is( $node_subscriber->safe_psql(
3|),
'partitions with different replica identities not replicated correctly');
-# TEST: With a table included in multiple publications, we should use a
-# union of the column lists. So with column lists (a,b) and (a,c) we
-# should replicate (a,b,c).
-
-$node_publisher->safe_psql(
- 'postgres', qq(
- CREATE TABLE test_mix_1 (a int PRIMARY KEY, b int, c int);
- CREATE PUBLICATION pub_mix_1 FOR TABLE test_mix_1 (a, b);
- CREATE PUBLICATION pub_mix_2 FOR TABLE test_mix_1 (a, c);
-
- -- initial data
- INSERT INTO test_mix_1 VALUES (1, 2, 3);
-));
-
-$node_subscriber->safe_psql(
- 'postgres', qq(
- CREATE TABLE test_mix_1 (a int PRIMARY KEY, b int, c int);
- ALTER SUBSCRIPTION sub1 SET PUBLICATION pub_mix_1, pub_mix_2;
-));
-
-wait_for_subscription_sync($node_subscriber);
-
-$node_publisher->safe_psql(
- 'postgres', qq(
- INSERT INTO test_mix_1 VALUES (4, 5, 6);
-));
-
-$node_publisher->wait_for_catchup('sub1');
-
-is( $node_subscriber->safe_psql(
- 'postgres', "SELECT * FROM test_mix_1 ORDER BY a"),
- qq(1|2|3
-4|5|6),
- 'a mix of publications should use a union of column list');
-
-# TEST: With a table included in multiple publications, we should use a
-# union of the column lists. If any of the publications is FOR ALL
-# TABLES, we should replicate all columns.
+# TEST: With a table included in the publications which is FOR ALL TABLES, it
+# means replicate all columns.
# drop unnecessary tables, so as not to interfere with the FOR ALL TABLES
$node_publisher->safe_psql(
'postgres', qq(
- DROP TABLE tab1, tab2, tab3, tab4, tab5, tab6, tab7, test_mix_1,
+ DROP TABLE tab1, tab2, tab3, tab4, tab5, tab6, tab7,
test_part, test_part_a, test_part_b, test_part_c, test_part_d;
));
$node_publisher->safe_psql(
'postgres', qq(
CREATE TABLE test_mix_2 (a int PRIMARY KEY, b int, c int);
- CREATE PUBLICATION pub_mix_3 FOR TABLE test_mix_2 (a, b);
+ CREATE PUBLICATION pub_mix_3 FOR TABLE test_mix_2 (a, b, c);
CREATE PUBLICATION pub_mix_4 FOR ALL TABLES;
-- initial data
@@ -976,12 +910,11 @@ $node_publisher->wait_for_catchup('sub1');
is( $node_subscriber->safe_psql('postgres', "SELECT * FROM test_mix_2"),
qq(1|2|3
4|5|6),
- 'a mix of publications should use a union of column list');
+ 'all columns should be replicated');
-# TEST: With a table included in multiple publications, we should use a
-# union of the column lists. If any of the publications is FOR ALL
-# TABLES IN SCHEMA, we should replicate all columns.
+# TEST: With a table included in the publication which is FOR ALL TABLES
+# IN SCHEMA, it means replicate all columns.
$node_subscriber->safe_psql(
'postgres', qq(
@@ -993,7 +926,7 @@ $node_publisher->safe_psql(
'postgres', qq(
DROP TABLE test_mix_2;
CREATE TABLE test_mix_3 (a int PRIMARY KEY, b int, c int);
- CREATE PUBLICATION pub_mix_5 FOR TABLE test_mix_3 (a, b);
+ CREATE PUBLICATION pub_mix_5 FOR TABLE test_mix_3 (a, b, c);
CREATE PUBLICATION pub_mix_6 FOR ALL TABLES IN SCHEMA public;
-- initial data
@@ -1017,7 +950,7 @@ $node_publisher->wait_for_catchup('sub1');
is( $node_subscriber->safe_psql('postgres', "SELECT * FROM test_mix_3"),
qq(1|2|3
4|5|6),
- 'a mix of publications should use a union of column list');
+ 'all columns should be replicated');
# TEST: Check handling of publish_via_partition_root - if a partition is
@@ -1074,7 +1007,7 @@ is( $node_subscriber->safe_psql(
# TEST: Multiple publications which publish schema of parent table and
# partition. The partition is published through two publications, once
# through a schema (so no column list) containing the parent, and then
-# also directly (with a columns list). The expected outcome is there is
+# also directly (with all columns). The expected outcome is there is
# no column list.
$node_publisher->safe_psql(
@@ -1086,7 +1019,7 @@ $node_publisher->safe_psql(
CREATE TABLE t_1 PARTITION OF s1.t FOR VALUES FROM (1) TO (10);
CREATE PUBLICATION pub1 FOR ALL TABLES IN SCHEMA s1;
- CREATE PUBLICATION pub2 FOR TABLE t_1(b);
+ CREATE PUBLICATION pub2 FOR TABLE t_1(a, b, c);
-- initial data
INSERT INTO s1.t VALUES (1, 2, 3);
@@ -1233,6 +1166,51 @@ is( $node_subscriber->safe_psql(
'publication containing both parent and child relation');
+# TEST: With a table included in multiple publications with different column
+# lists, we should catch the error when creating the subscription.
+
+$node_publisher->safe_psql('postgres', qq(
+ CREATE TABLE test_mix_1 (a int PRIMARY KEY, b int, c int);
+ CREATE PUBLICATION pub_mix_1 FOR TABLE test_mix_1 (a, b);
+ CREATE PUBLICATION pub_mix_2 FOR TABLE test_mix_1 (a, c);
+));
+
+$node_subscriber->safe_psql('postgres', qq(
+ DROP SUBSCRIPTION sub1;
+ CREATE TABLE test_mix_1 (a int PRIMARY KEY, b int, c int);
+));
+
+my ($cmdret, $stdout, $stderr) = $node_subscriber->psql(
+ 'postgres', qq(
+ CREATE SUBSCRIPTION sub1 CONNECTION '$publisher_connstr' PUBLICATION pub_mix_1, pub_mix_2;
+));
+
+ok( $stderr =~
+ qr/cannot use different column lists for table "public.test_mix_1" in different publications/,
+ 'different column lists detected');
+
+# TEST: If the column list is changed after creating the subscription, we
+# should catch the error reported by walsender.
+
+$node_publisher->safe_psql('postgres', qq(
+ ALTER PUBLICATION pub_mix_1 SET TABLE test_mix_1 (a, c);
+));
+
+$node_subscriber->safe_psql('postgres', qq(
+ CREATE SUBSCRIPTION sub1 CONNECTION '$publisher_connstr' PUBLICATION pub_mix_1, pub_mix_2;
+));
+
+$node_publisher->wait_for_catchup('sub1');
+
+$node_publisher->safe_psql('postgres', qq(
+ ALTER PUBLICATION pub_mix_1 SET TABLE test_mix_1 (a, b);
+ INSERT INTO test_mix_1 VALUES(1, 1, 1);
+));
+
+$offset = $node_publisher->wait_for_log(
+ qr/cannot use different column lists for table "public.test_mix_1" in different publications/,
+ $offset);
+
$node_subscriber->stop('fast');
$node_publisher->stop('fast');
--
2.7.2.windows.1
view thread (63+ 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]
Subject: RE: bogus: logical replication rows/cols combinations
In-Reply-To: <OS0PR01MB571647F2739F710E3D1B045D94CE9@OS0PR01MB5716.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