Обсуждение: ERROR: cache lookup failed for type 14237017

Yesterday, I had twelve thousand  "cache lookup failed for type N"
messages, like this:

2010-09-20 00:00:00 PDT ERROR:  cache lookup failed for type 14237017
2010-09-20 00:00:00 PDT CONTEXT:  SQL statement "INSERT INTO
mycluster.sl_log_2 (log_origin, log_xid, log_tableid, log_actionseq,
log_cmdtype, log_cmddata) VALUES (1, $1, $2,
nextval('mycluster.sl_action_seq'), $3, $4);"

The context is always Slony sl_log_2 table.  All twelve thousand
errors occurred within 40 minutes.  This did happen right after a
Slony cluster set drop and recreate.

What does "type 14237017" mean?  What cache are we talking about?

I am concerned because this is classified as ERROR rather than WARNING.

Thanks,
Aleksey

On Dienstag, 21. September 2010, Aleksey Tsalolikhin wrote:
> Yesterday, I had twelve thousand  "cache lookup failed for type N"
> messages, like this:

> What does "type 14237017" mean?

pg_type oid

> What cache are we talking about?

Did you alter a type before?
There's a bug in postgres, that prevents the plan cache to be
invalidated on type changes. We ran into this bug as well on 8.4.4.

HTH, Jens

On Tue, Sep 21, 2010 at 2:06 PM, Jens Wilke <jens@wilke.org> wrote:
> On Dienstag, 21. September 2010, Aleksey Tsalolikhin wrote:
>> Yesterday, I had twelve thousand  "cache lookup failed for type N"
>> messages, like this:
>
>> What does "type 14237017" mean?
>
> pg_type oid

Dear Jens,

  I am trying to understand your reply.

  pg_type is a catalog of data types:
http://www.postgresql.org/docs/8.4/static/catalog-pg-type.html

  An OID is a key to system table:
http://www.postgresql.org/docs/8.4/static/datatype-oid.html
I guess it's used to identify a row in a system table.

  So PostgreSQL was trying to lookup a row in a system table and did
not find it in a cache.


>> What cache are we talking about?
>
> Did you alter a type before?

No.  I don't even know how to alter a type.

> There's a bug in postgres, that prevents the plan cache to be
> invalidated on type changes. We ran into this bug as well on 8.4.4.

Is there a bug id or a URL for the bug, please?  I'd like to study the
bug description to
understand it better.

Sincerely,
Aleksey

On Dienstag, 21. September 2010, Aleksey Tsalolikhin wrote:

Hi Aleksey,

>   So PostgreSQL was trying to lookup a row in a system table and
> did not find it in a cache.

yes,
select * from pg_type where oid =14237017

>> Did you alter a type before?

> No.  I don't even know how to alter a type.

Sorry, what i mean is drop and recreate a type.

> Is there a bug id or a URL for the bug, please?  I'd like to study

No idea, sorry

Jens

On Tue, Sep 21, 2010 at 3:10 PM, Jens Wilke <jens@wilke.org> wrote:
> On Dienstag, 21. September 2010, Aleksey Tsalolikhin wrote:
>
>>   So PostgreSQL was trying to lookup a row in a system table and
>> did not find it in a cache.
>
> yes,
> select * from pg_type where oid =14237017

Thank you.

>>> Did you alter a type before?
>
>> No.  I don't even know how to alter a type.
>
> Sorry, what i mean is drop and recreate a type.

I did drop the entire slony schema and then recreated it.  It seems
to be slony-related issue; I am taking it further on the Slony mailing list.

Thanks!!
Aleksey