Re: Logical replication woes - Mailing list pgsql-hackers

From Heikki Linnakangas
Subject Re: Logical replication woes
Date
Msg-id 5374F44B.1060003@vmware.com
Whole thread Raw
In response to Logical replication woes  (Heikki Linnakangas <hlinnakangas@vmware.com>)
Responses Re: Logical replication woes  (Andres Freund <andres@2ndquadrant.com>)
Re: Logical replication woes  (Andres Freund <andres@2ndquadrant.com>)
List pgsql-hackers
On 05/15/2014 07:57 PM, Heikki Linnakangas wrote:
> Spotted while testing pg_recvlogical:
>
> 1. Set up pg_recvlogical to receive:
>
> ./pg_recvlogical -S fooslot -d postgres --create
> ./pg_recvlogical -S fooslot -d postgres --start -f -
>
> 2. In another terminal, with psql:
>
> create table foo (id int4);
> begin;
>    insert into foo values (4);
>    alter table foo alter column id type text;
> prepare transaction 'foo';
> commit prepared 'foo';
> insert into foo values (1);
>
> 3.  With current HEAD, after commit
> bb38fb0d43c8d7ff54072bfd8bd63154e536b384, this produces an assertion
> failure:
>
> TRAP: FailedAssertion("!(((xid) != ((TransactionId) 0)))", File:
> "reorderbuffer.c", Line: 508)
>
> I believe that's we no longer assign another XID to the transaction that
> does the COMMIT PREPARED. Previously, an extra XID, in addition to the
> XID of the prepared transaction, was assigned for use in locking the
> global transaction entry in shared memory, but that's no longer required.
>
> However, even with that patch reverted, it doesn't work correctly:
>
> ERROR:  could not map filenode "base/12142/16390" to relation OID
> LOG:  starting logical decoding for slot fooslot
> DETAIL:  streaming transactions committing after 0/16D1670, reading WAL
> from 0/16BC470

Ok, so the immediate cause was quick to find: when decoding a
commit-prepared WAL record, we have to use the XID from the record
content (patch attached). The XID in the record header is the XID of the
transaction doing the COMMIT PREPARED, which is always 0 after patch
bb38fb0d43c8d7ff54072bfd8bd63154e536b384 which causes the assertion. But
it was always wrong. After fixing, it no longer asserts or gives the
above "could not map filenode" error.

However, it still doesn't seem right. When I do the above as a regular
transaction, ie:

begin; insert into foo values (6);  alter table foo alter column id type
text; commit;

pg_recvlogical prints this:

BEGIN 708
table public.foo: INSERT: id[text]:'6'
COMMIT 708

But if I do it as a prepared transaction:

begin; insert into foo values (7);  alter table foo alter column id type
text; prepare transaction 'foo'; commit prepared 'foo';

pg_recvlogical prints nothing.

- Heikki

Attachment

pgsql-hackers by date:

Previous
From: Andres Freund
Date:
Subject: Re: pg_recvlogical, stdout and SIGHUP
Next
From: Heikki Linnakangas
Date:
Subject: Re: pg_recvlogical, stdout and SIGHUP