Обсуждение: ...
Hello all,
Just so this problem is logged in the system, I am making an overly vague
bug report. I have no nice method of repeating this as it requires a large
database with somewhat broken RI to run for a long time before running VACUUM
FULL ANALYZE;
I get:
dropsites=# vacuum full analyze;
WARNING: Rel pg_type: TID 12/46: InsertTransactionInProgress 27217603 - can't
shrink relation
WARNING: Rel pg_type: TID 12/47: InsertTransactionInProgress 27217603 - can't
shrink relation
WARNING: Rel pg_attribute: TID 152/1: InsertTransactionInProgress 27217603 -
can't shrink relation
WARNING: Rel pg_attribute: TID 152/2: InsertTransactionInProgress 27217603 -
can't shrink relation
...
...
And so on. The source code says "this should never happen".
This took place on the following system:
dropsites=# select version();
version
---------------------------------------------------------------------
PostgreSQL 7.3.2 on i386-unknown-freebsd4.6, compiled by GCC 2.95.4
(1 row)
[dropsites@m sendpage]$ uname -a
FreeBSD ***.****.com 4.9-PRERELEASE FreeBSD 4.9-PRERELEASE #0: Wed Sep 17
11:18:42 PDT 2003 ***@****.com:/usr/obj/usr/src/sys/WAY2EASY i386
Any more details you may request and I shall do my best to deliver. I am
currently trying to get this to happen on -STABLE, but to no avail.
Thanks all.
--
Wade Klaver
Wavefire Technologies Corporation
GPG Public Key at http://archeron.wavefire.com
/"\ ASCII Ribbon Campaign .
\ / - NO HTML/RTF in e-mail .
X - NO Word docs in e-mail .
/ \ -----------------------------------------------------------------
Wade Klaver <archeron@wavefire.com> writes:
> dropsites=# vacuum full analyze;
> WARNING: Rel pg_type: TID 12/46: InsertTransactionInProgress 27217603 - can't
> shrink relation
> WARNING: Rel pg_type: TID 12/47: InsertTransactionInProgress 27217603 - can't
> shrink relation
> WARNING: Rel pg_attribute: TID 152/1: InsertTransactionInProgress 27217603 -
> can't shrink relation
> WARNING: Rel pg_attribute: TID 152/2: InsertTransactionInProgress 27217603 -
> can't shrink relation
> ...
> ...
> And so on. The source code says "this should never happen".
7.4 has downgraded those messages to NOTICEs, and says
/*
* This should not happen, since we hold exclusive
* lock on the relation; shouldn't we raise an error?
* (Actually, it can happen in system catalogs, since
* we tend to release write lock before commit there.)
*/
If you saw it on anything besides a system catalog, then I'm interested,
but otherwise not to worry.
BTW, you really should update to 7.3.4.
regards, tom lane