Re: Theory about XLogFlush startup failures

> -----Original Message-----
> From: Tom Lane [mailto:tgl(at)sss(dot)pgh(dot)pa(dot)us]
>
> I said:
> > Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
> >> What about PageGetItem ? It seems to be able to touch the item
> >> via HeapTupleSatisfies etc.
>
> > Hmm. Strictly speaking I think you are right, but I'm hesitant to add a
> > bunch of new tests to PageGetItem --- that is much more of a hot spot
> > than PageAddItem, and it'll cost us something in speed I fear.
>
> I wasn't totally comfortable with this (and I'm sure you weren't
> either), but after more thought I still feel it's the right tradeoff.
> Here are a couple of heuristic arguments why we don't need more error
> checking in PageGetItem:

What I have minded is e.g. the following case.
Undoubtedly the page is corrupted(too big offset number).
I'm suspicious if other pages are safe under such a situation.

regards,
Hiroshi Inoue

a part of the report [GENERAL] Database corruption?
by Alvaro Herrera [alvherre(at)atentus(dot)com]

> DEBUG: --Relation delay_171--
> NOTICE: Rel delay_171: TID 15502/4279: OID IS INVALID. TUPGONE 0.
> NOTICE: Rel delay_171: TID 15502/4291: OID IS INVALID. TUPGONE 1.
> NOTICE: Rel delay_171: TID 15502/4315: OID IS INVALID. TUPGONE 1.
> NOTICE: Rel delay_171: TID 15502/4375: OID IS INVALID. TUPGONE 0.
> NOTICE: Rel delay_171: TID 15502/4723: OID IS INVALID. TUPGONE 1.
> NOTICE: Rel delay_171: TID 15502/4771: OID IS INVALID. TUPGONE 0.
> NOTICE: Rel delay_171: TID 15502/4783: OID IS INVALID. TUPGONE 0.
> NOTICE: Rel delay_171: TID 15502/4831: OID IS INVALID. TUPGONE 1.
> NOTICE: Rel delay_171: TID 15502/4843: OID IS INVALID. TUPGONE 0.
> NOTICE: Rel delay_171: TID 15502/4867: InsertTransactionInProgress 0 -
can't shrink relation
> NOTICE: Rel delay_171: TID 15502/4867: OID IS INVALID. TUPGONE 0.
> [a lot similarly looking lines]
> NOTICE: Rel delay_171: TID 15502/6067: OID IS INVALID. TUPGONE 0.
> Server process (pid 22773) exited with status 139 at Sun Oct 21 02:30:27
2001
> Terminating any active server processes...