Re: pg_restore failed on foreign key constraint

On Thu, 2026-02-12 at 16:35 -0500, Ron Johnson wrote:
> When running pg_restore 17.7 against a PG 14.20 database directory dump, I got this in the log:
>
> pg_restore: while PROCESSING TOC:
> pg_restore: from TOC entry 171927; 2606 3149123877 FK CONSTRAINT rel_user_email fk_rel_user_email_2 TAP
> pg_restore: error: could not execute query: ERROR:  insert or update on table "rel_user_email" violates foreign key
constraint"fk_rel_user_email_2" 
> DETAIL:  Key (access_email_id)=(2073) is not present in table "access_email".
> Command was: ALTER TABLE ONLY public.rel_user_email
>     ADD CONSTRAINT fk_rel_user_email_2 FOREIGN KEY (access_email_id) REFERENCES public.access_email(access_email_id);
>
> So, I went to the source database:
>
> TAPd=# \d rel_user_email
>                          Table "public.rel_user_email"
>      Column      |            Type             | Collation | Nullable | Default
> -----------------+-----------------------------+-----------+----------+---------
>  user_id         | integer                     |           | not null |
>  access_email_id | integer                     |           | not null |
>  modified_by     | integer                     |           |          |
>  modified_on     | timestamp without time zone |           | not null |
> Indexes:
>     "idx_rel_user_email" UNIQUE, btree (user_id, access_email_id)
> Foreign-key constraints:
>     "fk_rel_user_email_1" FOREIGN KEY (user_id) REFERENCES access_user(user_id)
>     "fk_rel_user_email_2" FOREIGN KEY (access_email_id) REFERENCES access_email(access_email_id)
>
> TAPd=# select * from rel_user_email where access_email_id=2073;
>  user_id | access_email_id | modified_by |       modified_on      
> ---------+-----------------+-------------+-------------------------
>     2452 |            2073 |          41 | 2013-03-11 10:52:20.331
> (1 row)
>
> TAPd=# \d access_email
>                                                  Table "public.access_email"
>      Column      |            Type             | Collation | Nullable |                        Default              
         
>
-----------------+-----------------------------+-----------+----------+-------------------------------------------------------
>  access_email_id | integer                     |           | not null |
nextval('access_email_access_email_id_seq'::regclass)
>  type            | numeric(10,0)               |           |          |
>  email_address   | character varying(255)      |           |          |
>  created_on      | timestamp without time zone |           | not null |
>  modified_on     | timestamp without time zone |           |          |
>  created_by      | integer                     |           |          |
>  modified_by     | integer                     |           |          |
> Indexes:
>     "pk_access_email" PRIMARY KEY, btree (access_email_id)
> Referenced by:
>     TABLE "rel_user_email" CONSTRAINT "fk_rel_user_email_2" FOREIGN KEY (access_email_id) REFERENCES
access_email(access_email_id)
>
> TAPd=# select * from access_email where access_email_id=2073;
>  access_email_id | type | email_address | created_on | modified_on | created_by | modified_by
> -----------------+------+---------------+------------+-------------+------------+-------------
> (0 rows)
>
> Looks like index corruption.  
>
> $ pg_amcheck -t public.rel_user_email -t public.access_email TAPd
> $ echo $?
> 0
> $ pg_amcheck -i idx_rel_user_email -i pk_access_email --parent-check --heapallindexed TAPd
> $ echo $?
> 0
>
> But amcheck shows no problems.
>
> Before I get worried that there' s corrupt data: am I missing something obvious?

Try

  SET enable_indexscan = off;

  SELECT * FROM access_email WHERE access_email_id = 2073;

Only if that returns a row, I would assume index corruption, and that one should have been
caught with "heapallindexed".

It is the foreign key that is violated.  The normal ways to end up with broken foreign
keys are

  SET session_replication_role = replica;

and

  ALTER TABLE rel_user_email DISABLE TRIGGER ALL;

both of which require superuser privileges.

Yours,
Laurenz Albe