Horton

Re: Re: PANIC: invalid index offnum: 186 when processing BRIN indexes in VACUUM

Alvaro Herrera <alvherre@alvh.no-ip.org>

View thread View on postgr.es
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>, Michael Paquier <michael.paquier@gmail.com>, PostgreSQL mailing lists <pgsql-hackers@postgresql.org>
Date: 2017-11-03T10:04:28Z
Lists: pgsql-hackers 
Tom Lane wrote:
> Alvaro Herrera <alvherre@alvh.no-ip.org> writes:

> > Rather than remove the capability, I'd be inclined to make
> > brin_summarize_new_values summarize the final partial range, and have
> > VACUUM not do it.  Would that be too inconsistent?
> 
> That doesn't really get you out of the problem that this is an abuse of
> the relation extension lock, and is likely to cause issues when people
> try to optimize that locking mechanism.

Right.

> Why is it that the regular technique doesn't work, ie create a placeholder
> tuple and let it get added to by any insertions that happen?

The problem is that we determine relation size (and scan stop point)
before inserting the placeholder tuple, so any relation extension that
occurs after we read the size is not covered by the scan.  The reason we
do this is to avoid calling RelationGetNumberOfBlocks once for each
range, since it's known to be very expensive.

Maybe a solution is to call RelationGetNumberOfBlocks() after the
placeholder tuple has been inserted, for the case where we would be
scanning past end of relation; passing InvalidBlockNumber as stop point
would indicate to do things that way.  I'll try with that approach now.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Commits

  1. Fix BRIN summarization concurrent with extension

  2. Fix corner-case errors in brin_doupdate().

  3. Rewrite PageIndexDeleteNoCompact into a form that only deletes 1 tuple.

  4. Invent PageIndexTupleOverwrite, and teach BRIN and GiST to use it.

  5. Close some holes in BRIN page assignment