Обсуждение: "LockRelease: locktable lookup failed, no lock"

  Searching through the archives, I see a few other persons having this
problem, but no solutions, so I'll ask again. : )

  I'm getting "LockRelease: locktable lookup failed, no lock" trying to
insert a row into a database.   The database is very simple, it contains a
single table with 8 fields.   Over the last few weeks, the database has
been corrupted regularly, but a "vacuum analyze" would do the trick.  I
decided to see if I couldn't fix the real problem, though, so I destroyed
the database, created it again, and tried to insert data - but I keep
getting "LockRelease: locktable lookup failed, no lock" when trying to
insert records.

   If anyone has suggestions on what to do to fix this problem, please let
me know.

steve

>   I'm getting "LockRelease: locktable lookup failed, no lock" trying to
> insert a row into a database.

  Well, I think I tracked down the problem - an index on a "text" field.
Once I removed the index, everything works well.  The odd thing is that not
*all* inserts would cause the error.  I suppose I ought to dig into the
docs when I get a few spare minutes...

steve

> >   I'm getting "LockRelease: locktable lookup failed, no lock" trying to
> > insert a row into a database.
>
>   Well, I think I tracked down the problem - an index on a "text" field.
> Once I removed the index, everything works well.  The odd thing is that not
> *all* inserts would cause the error.  I suppose I ought to dig into the
> docs when I get a few spare minutes...

What is the version of PostgreSQL and what kind of platform are you
using?
--
Tatsuo Ishii

> > >   I'm getting "LockRelease: locktable lookup failed, no lock" trying
to
> > > insert a row into a database.
> >
> >   Well, I think I tracked down the problem - an index on a "text"
field.
> > Once I removed the index, everything works well.  The odd thing is that
not
> > *all* inserts would cause the error.  I suppose I ought to dig into the
> > docs when I get a few spare minutes...
>
> What is the version of PostgreSQL and what kind of platform are you
> using?

  Lemmee see...  pgsql 6.5.3 on redhat linux 6.1, kernel 2.2.12-20smp.  The
machine currently isn't running multicpu, that will happen soon.  At
bootup, linux finds only one chip, and does pseudo-epic. : )

steve

  Grr....

  Although I'm no longer getting lock releases on this table, a daily
vaccum analyze still shows:

NOTICE:  Rel reef: Uninitialized page 492 - fixing
NOTICE:  Rel reef: Uninitialized page 498 - fixing
NOTICE:  Rel reef: Uninitialized page 499 - fixing

  The database is a fledgling mail archive, when a message comes in, a
program parses the message for the relevant information, inserts it into
the table, and it's done.  Since it's in very early development, there are
only very occasional queries on it.  Here's the schema that I used to
create the table....

create sequence reef_sequence increment 2;

create table reef(
    reef_index              int4 default nextval('reef_sequence') primary
key,
    subject                 varchar(256),
    author                  varchar(256),
    author_email          varchar(256),
    message               text,
    date                    date,
    time                    time,
    is_reply                int
);

  Now I know that there are problems with that, like "is_reply" should be a
bool, not an int.  This is a very old project that I'm just starting up
again... go easy on me. : )

steve

Hi Steve,

Have you checked for inserts in the message field larger then 8k?

Cheers Wim.

Steve Wolfe wrote:

>   Grr....
>
>   Although I'm no longer getting lock releases on this table, a daily
> vaccum analyze still shows:
>
> NOTICE:  Rel reef: Uninitialized page 492 - fixing
> NOTICE:  Rel reef: Uninitialized page 498 - fixing
> NOTICE:  Rel reef: Uninitialized page 499 - fixing
>
>   The database is a fledgling mail archive, when a message comes in, a
> program parses the message for the relevant information, inserts it into
> the table, and it's done.  Since it's in very early development, there are
> only very occasional queries on it.  Here's the schema that I used to
> create the table....
>
> create sequence reef_sequence increment 2;
>
> create table reef(
>     reef_index              int4 default nextval('reef_sequence') primary
> key,
>     subject                 varchar(256),
>     author                  varchar(256),
>     author_email          varchar(256),
>     message               text,
>     date                    date,
>     time                    time,
>     is_reply                int
> );
>
>   Now I know that there are problems with that, like "is_reply" should be a
> bool, not an int.  This is a very old project that I'm just starting up
> again... go easy on me. : )
>
> steve
>
> ************

> Have you checked for inserts in the message field larger then 8k?

  Ah..... I see.  I had forgotten that a "text" field only allows up to 8K,
I was under the (false) impression that it was larger.  I suppose that I'll
have to look into a large object for that....

steve