On 2014-03-17 10:03:52 -0700, Josh Berkus wrote:
> On 03/17/2014 08:28 AM, Tom Lane wrote:
> > Greg Stark <stark@mit.edu> writes:
> >> The error causes some rows to disappear from indexes resulting in
> >> inconsistent query results on a hot standby depending on whether
> >> indexes are used. If the standby is subsequently activated or if it
> >> occurs during recovery after a crash or backup restore it could result
> >> in unique constraint violations as well.
> >
> > Hm ... "rows disappearing from indexes" might make people think that
> > they could fix or mitigate the damage via REINDEX. That's not really
> > true though is it? It looks to me like IndexBuildHeapScan will suffer
> > an Assert failure in an assert-enabled build, or build a bogus index
> > if not assert-enabled, when it comes across a "heap-only" tuple that
> > has no parent.
>
> First, see suggested text in my first-draft release announcement.
I don't think that text is any better, it's imo even wrong:
"The bug causes rows to vanish from indexes during recovery due to
simultaneous updates of rows on both sides of a foreign key."
Neither is a foreign key, nor simultaneous updates, nor both sides a
prerequisite.
> Second, if a user has encountered this kind of data corruption on their
> master (due to crash recovery), how exactly *do* they fix it?
Dump/restore is the most obvious candidate. The next best thing I can
think of is a noop rewriting ALTER TABLE, that doesn't deal with ctid
chains IIRC, in contrast to CLUSTER/VACUUM FULL.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services