Horton

Re: Minimal logical decoding on standbys

Robert Haas <robertmhaas@gmail.com>

View thread
From: Robert Haas <robertmhaas@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>, Alvaro Herrera <alvherre@2ndquadrant.com>, Ibrar Ahmed <ibrar.ahmad@gmail.com>, Amit Khandekar <amitdkhan.pg@gmail.com>, fabriziomello@gmail.com, tushar <tushar.ahuja@enterprisedb.com>, Rahila Syed <rahila.syed@2ndquadrant.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2022-12-14T17:48:08Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Reduce the log level in 035_standby_logical_decoding.pl.

  2. 035_standby_logical_decoding: Add missing waits for replication

  3. For cascading replication, wake physical and logical walsenders separately

  4. Handle logical slot conflicts on standby

  5. Support invalidating replication slots due to horizon and wal_level

  6. Prevent use of invalidated logical slot in CreateDecodingContext()

  7. Replace replication slot's invalidated_at LSN with an enum

  8. Pass down table relation into more index relation functions

  9. Assert only valid flag bits are passed to visibilitymap_set()

  10. Remove unused _bt_delitems_delete() argument.

  11. Add xl_btree_delete optimization.

On Wed, Dec 14, 2022 at 12:35 PM Andres Freund <andres@anarazel.de> wrote:
> >  typedef struct xl_heap_prune
> >
> > I think this is unsafe on alignment-picky machines. I think it will
> > cause the offset numbers to be aligned at an odd address.
> > heap_xlog_prune() doesn't copy the data into aligned memory, so I
> > think this will result in a misaligned pointer being passed down to
> > heap_page_prune_execute.
>
> I think the offset numbers are stored separately from the record, even
> though it doesn't quite look like that in the above due to the way the
> 'OFFSET NUMBERS' is embedded in the struct. As they're stored with the
> block reference 0, the added boolean shouldn't make a difference
> alignment wise?
>
> Or am I misunderstanding your point?

Oh, you're right. So this is another case similar to
xl_btree_reuse_page. In heap_xlog_prune(), we access the offset number
data like this:

                redirected = (OffsetNumber *)
XLogRecGetBlockData(record, 0, &datalen);
                end = (OffsetNumber *) ((char *) redirected + datalen);
                nowdead = redirected + (nredirected * 2);
                nowunused = nowdead + ndead;
                nunused = (end - nowunused);
                heap_page_prune_execute(buffer,

redirected, nredirected,
                                                                nowdead, ndead,

nowunused, nunused);

This is only safe if the return value of XLogRecGetBlockData is
guaranteed to be properly aligned, and I think that there is no such
guarantee in general. I think that it happens to come out properly
aligned because both the main body of the record (xl_heap_prune) and
the length of a block header (XLogRecordBlockHeader) happen to be
sufficiently aligned. But we just recently had discussion about trying
to make WAL records smaller by various means, and some of those
proposals involved changing the length of XLogRecordBlockHeader. And
the patch proposed here increases SizeOfHeapPrune by 1. So I think
with the patch, the offset number array would become misaligned.

No?

-- 
Robert Haas
EDB: http://www.enterprisedb.com