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: Tomas Vondra <tomas.vondra@2ndquadrant.com>, Tom Lane <tgl@sss.pgh.pa.us>
Cc: Michael Paquier <michael.paquier@gmail.com>, PostgreSQL mailing lists <pgsql-hackers@postgresql.org>
Date: 2017-11-03T13:12:27Z
Lists: pgsql-hackers

Attachments

Alvaro Herrera wrote:

> 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.

Yeah, I think this approach results in better code.  The attached patch
implements that, and it passes the test for me (incl. calling
brin_summarize_new_values concurrently with vacuum, when running the
insert; the former does include the final page range whereas the latter
does not.)

Tomas Vondra wrote:

> FWIW this patch fixes the issue for me - I can no longer reproduce the
> bitmapscan vs. seqscan result discrepancies (even with the extra UPDATE
> phase).

Thanks for testing!  This confirms that the issue was correctly
identified.  Would you try the current patch, which is better than the
other one?  It's significantly different that I think it invalidates
prior testing.

Thanks!

-- 
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