BF failure: could not open relation with OID XXXX while querying pg_views

Started by Thomas Munroover 6 years ago12 messages
#1Thomas Munro
thomas.munro@gmail.com

Hi,

Here are three strange recent failures in the "rules" test:

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=blenny&dt=2019-08-13%2022:19:27
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=alewife&dt=2019-07-27%2009:39:05
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-07-18%2003:00:27

They all raised "ERROR: could not open relation with OID <varies>"
while running:

SELECT viewname, definition FROM pg_views
WHERE schemaname IN ('pg_catalog', 'public')
ORDER BY viewname;

--
Thomas Munro
https://enterprisedb.com

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Thomas Munro (#1)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

Thomas Munro <thomas.munro@gmail.com> writes:

Here are three strange recent failures in the "rules" test:
...
They all raised "ERROR: could not open relation with OID <varies>"
while running:
SELECT viewname, definition FROM pg_views
WHERE schemaname IN ('pg_catalog', 'public')
ORDER BY viewname;

I think the problem is probably that Peter ignored this bit of advice
in parallel_schedule:

# rules cannot run concurrently with any test that creates
# a view or rule in the public schema

when he inserted "collate.linux.utf8" concurrently with "rules".

(I suspect BTW that the point is not so much that you better not
*create* such an object, as that you better not *drop* it concurrently
with that query.)

regards, tom lane

#3Peter Eisentraut
peter.eisentraut@2ndquadrant.com
In reply to: Tom Lane (#2)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On 2019-08-14 05:52, Tom Lane wrote:

Thomas Munro <thomas.munro@gmail.com> writes:

Here are three strange recent failures in the "rules" test:
...
They all raised "ERROR: could not open relation with OID <varies>"
while running:
SELECT viewname, definition FROM pg_views
WHERE schemaname IN ('pg_catalog', 'public')
ORDER BY viewname;

I think the problem is probably that Peter ignored this bit of advice
in parallel_schedule:

# rules cannot run concurrently with any test that creates
# a view or rule in the public schema

when he inserted "collate.linux.utf8" concurrently with "rules".

This test file is set up to create everything in the "collate_tests"
schema. Is that not working correctly?

--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Peter Eisentraut (#3)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

Peter Eisentraut <peter.eisentraut@2ndquadrant.com> writes:

On 2019-08-14 05:52, Tom Lane wrote:

Thomas Munro <thomas.munro@gmail.com> writes:

They all raised "ERROR: could not open relation with OID <varies>"
while running:
SELECT viewname, definition FROM pg_views
WHERE schemaname IN ('pg_catalog', 'public')
ORDER BY viewname;

I think the problem is probably that Peter ignored this bit of advice
in parallel_schedule:
# rules cannot run concurrently with any test that creates
# a view or rule in the public schema
when he inserted "collate.linux.utf8" concurrently with "rules".

This test file is set up to create everything in the "collate_tests"
schema. Is that not working correctly?

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

regards, tom lane

#5Thomas Munro
thomas.munro@gmail.com
In reply to: Tom Lane (#4)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Wed, Aug 14, 2019 at 5:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

Looks like stats_ext.sql could be the problem. It creates and drops
priv_test_view, not in a schema. Adding Dean, author of commit
d7f8d26d.

--
Thomas Munro
https://enterprisedb.com

#6Tomas Vondra
tomas.vondra@2ndquadrant.com
In reply to: Thomas Munro (#5)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Wed, Aug 14, 2019 at 05:24:26PM +1200, Thomas Munro wrote:

On Wed, Aug 14, 2019 at 5:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

Looks like stats_ext.sql could be the problem. It creates and drops
priv_test_view, not in a schema. Adding Dean, author of commit
d7f8d26d.

Yeah, that seems like it might be the cause. I'll take a look at fixing
this, probably by creating the view in a different schema.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

#7Tom Lane
tgl@sss.pgh.pa.us
In reply to: Tomas Vondra (#6)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:

On Wed, Aug 14, 2019 at 05:24:26PM +1200, Thomas Munro wrote:

On Wed, Aug 14, 2019 at 5:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

Looks like stats_ext.sql could be the problem. It creates and drops
priv_test_view, not in a schema. Adding Dean, author of commit
d7f8d26d.

Yeah, that seems like it might be the cause. I'll take a look at fixing
this, probably by creating the view in a different schema.

Ping? We're still getting intermittent failures of this ilk, eg

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dragonet&amp;dt=2019-09-14%2003%3A37%3A03

With v12 release approaching, I'd like to not have failures
like this in a released branch.

regards, tom lane

#8Dean Rasheed
dean.a.rasheed@gmail.com
In reply to: Tom Lane (#7)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Sat, 14 Sep 2019 at 05:25, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:

On Wed, Aug 14, 2019 at 05:24:26PM +1200, Thomas Munro wrote:

On Wed, Aug 14, 2019 at 5:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

Looks like stats_ext.sql could be the problem. It creates and drops
priv_test_view, not in a schema. Adding Dean, author of commit
d7f8d26d.

Yeah, that seems like it might be the cause. I'll take a look at fixing
this, probably by creating the view in a different schema.

Ping? We're still getting intermittent failures of this ilk, eg

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dragonet&amp;dt=2019-09-14%2003%3A37%3A03

With v12 release approaching, I'd like to not have failures
like this in a released branch.

Ah sorry, I missed this thread before. As author of that commit, it's
really on me to fix it, and the cause seems pretty clear-cut, so I'll
aim to get that done today.

Regards,
Dean

#9Tomas Vondra
tomas.vondra@2ndquadrant.com
In reply to: Dean Rasheed (#8)
1 attachment(s)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Sun, Sep 15, 2019 at 10:16:30AM +0100, Dean Rasheed wrote:

On Sat, 14 Sep 2019 at 05:25, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:

On Wed, Aug 14, 2019 at 05:24:26PM +1200, Thomas Munro wrote:

On Wed, Aug 14, 2019 at 5:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Oh, hmm --- yeah, that should mean it's safe. Maybe somebody incautiously
changed one of the other tests that run concurrently with "rules"?

Looks like stats_ext.sql could be the problem. It creates and drops
priv_test_view, not in a schema. Adding Dean, author of commit
d7f8d26d.

Yeah, that seems like it might be the cause. I'll take a look at fixing
this, probably by creating the view in a different schema.

Ping? We're still getting intermittent failures of this ilk, eg

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dragonet&amp;dt=2019-09-14%2003%3A37%3A03

With v12 release approaching, I'd like to not have failures
like this in a released branch.

Ah sorry, I missed this thread before. As author of that commit, it's
really on me to fix it, and the cause seems pretty clear-cut, so I'll
aim to get that done today.

FWIW here is a draft patch that I was going to propose - it simply moves
the table+view into a "tststats" schema. I suppose that's rougly what we
discussed earlier in this thread.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Attachments:

stats_ext_view.patchtext/plain; charset=us-asciiDownload
diff --git a/src/test/regress/expected/stats_ext.out b/src/test/regress/expected/stats_ext.out
index e56c75cd33..b2fb260172 100644
--- a/src/test/regress/expected/stats_ext.out
+++ b/src/test/regress/expected/stats_ext.out
@@ -752,19 +752,21 @@ SELECT * FROM check_estimated_rows('SELECT * FROM mcv_lists_bool WHERE NOT a AND
 -- the underlying table.
 --
 -- Currently this is only relevant for MCV stats.
-CREATE TABLE priv_test_tbl (
+CREATE schema tststats;
+CREATE TABLE tststats.priv_test_tbl (
     a int,
     b int
 );
-INSERT INTO priv_test_tbl
+INSERT INTO tststats.priv_test_tbl
      SELECT mod(i,5), mod(i,10) FROM generate_series(1,100) s(i);
-CREATE STATISTICS priv_test_stats (mcv) ON a, b
-  FROM priv_test_tbl;
-ANALYZE priv_test_tbl;
+CREATE STATISTICS tststats.priv_test_stats (mcv) ON a, b
+  FROM tststats.priv_test_tbl;
+ANALYZE tststats.priv_test_tbl;
 -- User with no access
 CREATE USER regress_stats_user1;
+GRANT USAGE ON SCHEMA tststats TO regress_stats_user1;
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_tbl; -- Permission denied
+SELECT * FROM tststats.priv_test_tbl; -- Permission denied
 ERROR:  permission denied for table priv_test_tbl
 -- Attempt to gain access using a leaky operator
 CREATE FUNCTION op_leak(int, int) RETURNS bool
@@ -772,39 +774,41 @@ CREATE FUNCTION op_leak(int, int) RETURNS bool
     LANGUAGE plpgsql;
 CREATE OPERATOR <<< (procedure = op_leak, leftarg = int, rightarg = int,
                      restrict = scalarltsel);
-SELECT * FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
+SELECT * FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
 ERROR:  permission denied for table priv_test_tbl
-DELETE FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
+DELETE FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
 ERROR:  permission denied for table priv_test_tbl
 -- Grant access via a security barrier view, but hide all data
 RESET SESSION AUTHORIZATION;
-CREATE VIEW priv_test_view WITH (security_barrier=true)
-    AS SELECT * FROM priv_test_tbl WHERE false;
-GRANT SELECT, DELETE ON priv_test_view TO regress_stats_user1;
+CREATE VIEW tststats.priv_test_view WITH (security_barrier=true)
+    AS SELECT * FROM tststats.priv_test_tbl WHERE false;
+GRANT SELECT, DELETE ON tststats.priv_test_view TO regress_stats_user1;
 -- Should now have access via the view, but see nothing and leak nothing
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
+SELECT * FROM tststats.priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
  a | b 
 ---+---
 (0 rows)
 
-DELETE FROM priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
+DELETE FROM tststats.priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
 -- Grant table access, but hide all data with RLS
 RESET SESSION AUTHORIZATION;
-ALTER TABLE priv_test_tbl ENABLE ROW LEVEL SECURITY;
-GRANT SELECT, DELETE ON priv_test_tbl TO regress_stats_user1;
+ALTER TABLE tststats.priv_test_tbl ENABLE ROW LEVEL SECURITY;
+GRANT SELECT, DELETE ON tststats.priv_test_tbl TO regress_stats_user1;
 -- Should now have direct table access, but see nothing and leak nothing
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
+SELECT * FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
  a | b 
 ---+---
 (0 rows)
 
-DELETE FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
+DELETE FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
 -- Tidy up
 DROP OPERATOR <<< (int, int);
 DROP FUNCTION op_leak(int, int);
 RESET SESSION AUTHORIZATION;
-DROP VIEW priv_test_view;
-DROP TABLE priv_test_tbl;
+DROP SCHEMA tststats CASCADE;
+NOTICE:  drop cascades to 2 other objects
+DETAIL:  drop cascades to table tststats.priv_test_tbl
+drop cascades to view tststats.priv_test_view
 DROP USER regress_stats_user1;
diff --git a/src/test/regress/sql/stats_ext.sql b/src/test/regress/sql/stats_ext.sql
index 65d1e39f0c..62956d25f8 100644
--- a/src/test/regress/sql/stats_ext.sql
+++ b/src/test/regress/sql/stats_ext.sql
@@ -493,23 +493,27 @@ SELECT * FROM check_estimated_rows('SELECT * FROM mcv_lists_bool WHERE NOT a AND
 -- the underlying table.
 --
 -- Currently this is only relevant for MCV stats.
-CREATE TABLE priv_test_tbl (
+CREATE schema tststats;
+
+CREATE TABLE tststats.priv_test_tbl (
     a int,
     b int
 );
 
-INSERT INTO priv_test_tbl
+INSERT INTO tststats.priv_test_tbl
      SELECT mod(i,5), mod(i,10) FROM generate_series(1,100) s(i);
 
-CREATE STATISTICS priv_test_stats (mcv) ON a, b
-  FROM priv_test_tbl;
+CREATE STATISTICS tststats.priv_test_stats (mcv) ON a, b
+  FROM tststats.priv_test_tbl;
 
-ANALYZE priv_test_tbl;
+ANALYZE tststats.priv_test_tbl;
 
 -- User with no access
 CREATE USER regress_stats_user1;
+GRANT USAGE ON SCHEMA tststats TO regress_stats_user1;
+
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_tbl; -- Permission denied
+SELECT * FROM tststats.priv_test_tbl; -- Permission denied
 
 -- Attempt to gain access using a leaky operator
 CREATE FUNCTION op_leak(int, int) RETURNS bool
@@ -517,34 +521,33 @@ CREATE FUNCTION op_leak(int, int) RETURNS bool
     LANGUAGE plpgsql;
 CREATE OPERATOR <<< (procedure = op_leak, leftarg = int, rightarg = int,
                      restrict = scalarltsel);
-SELECT * FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
-DELETE FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
+SELECT * FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
+DELETE FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Permission denied
 
 -- Grant access via a security barrier view, but hide all data
 RESET SESSION AUTHORIZATION;
-CREATE VIEW priv_test_view WITH (security_barrier=true)
-    AS SELECT * FROM priv_test_tbl WHERE false;
-GRANT SELECT, DELETE ON priv_test_view TO regress_stats_user1;
+CREATE VIEW tststats.priv_test_view WITH (security_barrier=true)
+    AS SELECT * FROM tststats.priv_test_tbl WHERE false;
+GRANT SELECT, DELETE ON tststats.priv_test_view TO regress_stats_user1;
 
 -- Should now have access via the view, but see nothing and leak nothing
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
-DELETE FROM priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
+SELECT * FROM tststats.priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
+DELETE FROM tststats.priv_test_view WHERE a <<< 0 AND b <<< 0; -- Should not leak
 
 -- Grant table access, but hide all data with RLS
 RESET SESSION AUTHORIZATION;
-ALTER TABLE priv_test_tbl ENABLE ROW LEVEL SECURITY;
-GRANT SELECT, DELETE ON priv_test_tbl TO regress_stats_user1;
+ALTER TABLE tststats.priv_test_tbl ENABLE ROW LEVEL SECURITY;
+GRANT SELECT, DELETE ON tststats.priv_test_tbl TO regress_stats_user1;
 
 -- Should now have direct table access, but see nothing and leak nothing
 SET SESSION AUTHORIZATION regress_stats_user1;
-SELECT * FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
-DELETE FROM priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
+SELECT * FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
+DELETE FROM tststats.priv_test_tbl WHERE a <<< 0 AND b <<< 0; -- Should not leak
 
 -- Tidy up
 DROP OPERATOR <<< (int, int);
 DROP FUNCTION op_leak(int, int);
 RESET SESSION AUTHORIZATION;
-DROP VIEW priv_test_view;
-DROP TABLE priv_test_tbl;
+DROP SCHEMA tststats CASCADE;
 DROP USER regress_stats_user1;
#10Dean Rasheed
dean.a.rasheed@gmail.com
In reply to: Tomas Vondra (#9)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Sun, 15 Sep 2019 at 11:11, Tomas Vondra <tomas.vondra@2ndquadrant.com> wrote:

On Sun, Sep 15, 2019 at 10:16:30AM +0100, Dean Rasheed wrote:

Ah sorry, I missed this thread before. As author of that commit, it's
really on me to fix it, and the cause seems pretty clear-cut, so I'll
aim to get that done today.

FWIW here is a draft patch that I was going to propose - it simply moves
the table+view into a "tststats" schema. I suppose that's rougly what we
discussed earlier in this thread.

Yes, that matches what I just drafted.
Do you want to push it, or shall I?

Regards,
Dean

#11Tomas Vondra
tomas.vondra@2ndquadrant.com
In reply to: Dean Rasheed (#10)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Sun, Sep 15, 2019 at 11:27:19AM +0100, Dean Rasheed wrote:

On Sun, 15 Sep 2019 at 11:11, Tomas Vondra <tomas.vondra@2ndquadrant.com> wrote:

On Sun, Sep 15, 2019 at 10:16:30AM +0100, Dean Rasheed wrote:

Ah sorry, I missed this thread before. As author of that commit, it's
really on me to fix it, and the cause seems pretty clear-cut, so I'll
aim to get that done today.

FWIW here is a draft patch that I was going to propose - it simply moves
the table+view into a "tststats" schema. I suppose that's rougly what we
discussed earlier in this thread.

Yes, that matches what I just drafted.
Do you want to push it, or shall I?

Please go ahead and push. I'm temporarily without commit access.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

#12Dean Rasheed
dean.a.rasheed@gmail.com
In reply to: Tomas Vondra (#11)
Re: BF failure: could not open relation with OID XXXX while querying pg_views

On Sun, 15 Sep 2019 at 12:20, Tomas Vondra <tomas.vondra@2ndquadrant.com> wrote:

On Sun, Sep 15, 2019 at 11:27:19AM +0100, Dean Rasheed wrote:

On Sun, 15 Sep 2019 at 11:11, Tomas Vondra <tomas.vondra@2ndquadrant.com> wrote:

On Sun, Sep 15, 2019 at 10:16:30AM +0100, Dean Rasheed wrote:

Ah sorry, I missed this thread before. As author of that commit, it's
really on me to fix it, and the cause seems pretty clear-cut, so I'll
aim to get that done today.

FWIW here is a draft patch that I was going to propose - it simply moves
the table+view into a "tststats" schema. I suppose that's rougly what we
discussed earlier in this thread.

Yes, that matches what I just drafted.
Do you want to push it, or shall I?

Please go ahead and push. I'm temporarily without commit access.

OK, pushed.

Regards,
Dean