Обсуждение: Error message: cannot identify operator 184

On Thu, Jul 14, 2005 at 01:48:38PM -0700, Chris White (cjwhite) wrote:
> Anybody have any idea why the error message "cannot identify operator
> 184" should be output. Also, what does it mean as to the state of the
> database, do we need to do anything to correct a problem?

Is that the *exact* error message?  Where did you see it?  Can you
tell us anything more about the context in which it was generated?

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

-----Original Message-----
From: Michael Fuhr [mailto:mike@fuhr.org]
Sent: Thursday, July 14, 2005 3:27 PM
To: Chris White (cjwhite)
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Error message: cannot identify operator 184

On Thu, Jul 14, 2005 at 01:48:38PM -0700, Chris White (cjwhite) wrote:
> Anybody have any idea why the error message "cannot identify operator
> 184" should be output. Also, what does it mean as to the state of the
> database, do we need to do anything to correct a problem?

Is that the *exact* error message?  Where did you see it?  Can you tell
us anything more about the context in which it was generated?

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

Sorry, the exact error message is

could not identify operator 184.

I saw it in the syslog error log. At the moment I can't tell what was
going on when it happened.

-----Original Message-----
From: Michael Fuhr [mailto:mike@fuhr.org]
Sent: Thursday, July 14, 2005 3:27 PM
To: Chris White (cjwhite)
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Error message: cannot identify operator 184

On Thu, Jul 14, 2005 at 01:48:38PM -0700, Chris White (cjwhite) wrote:
> Anybody have any idea why the error message "cannot identify operator
> 184" should be output. Also, what does it mean as to the state of the
> database, do we need to do anything to correct a problem?

Is that the *exact* error message?  Where did you see it?  Can you tell
us anything more about the context in which it was generated?

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

On Thu, Jul 14, 2005 at 03:38:19PM -0700, Chris White (cjwhite) wrote:
> Sorry, the exact error message is
>
> could not identify operator 184.
>
> I saw it in the syslog error log. At the moment I can't tell what was
> going on when it happened.

What version of PostgreSQL are you running?  I'll guess 7.4.x, as
that's the only branch that appears to contain the phrase "could
not identify operator" in an error message.  The message appears
in the _bt_getstrategynumber() function in access/nbtree/nbtutils.c;
maybe somebody who's familiar with the btree code can say what's
going on.

What happens if you run the following query?

SELECT 184::regprocedure;

On my 7.4.8 systems the above query returns the following:

  regprocedure
----------------
 oideq(oid,oid)
(1 row)

I'm wondering if the error means you're missing that function, which
is used to compare oid values.  But again, I'm not very familiar with
the btree code, so maybe somebody else can comment.

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

Michael Fuhr <mike@fuhr.org> writes:
> I'm wondering if the error means you're missing that function, which
> is used to compare oid values.

A system without oideq would be so dead in the water it's not funny ---
consider that all the major catalogs are indexed by oid.  It sounds to
me that one particular index's "strategy map" (ie, its in-memory copy
of the relevant indexable-operator info) got trashed.  It's hard to
guess why though.  Could have been a hardware problem (RAM dropped a
bit) or a software problem (wild store that happened to trash this data
structure rather than anything else).

            regards, tom lane