Horton

Re: Minimal logical decoding on standbys

bdrouvotAWS <bdrouvot@amazon.com>

View thread
From: "Drouvot, Bertrand" <bdrouvot@amazon.com>
To: Andres Freund <andres@anarazel.de>, Robert Haas <robertmhaas@gmail.com>
Cc: 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: 2022-07-06T13:30:37Z
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.

Hi,

On 10/28/21 11:07 PM, Andres Freund wrote:
> Hi,
>
> On 2021-10-28 16:24:22 -0400, Robert Haas wrote:
>> On Wed, Oct 27, 2021 at 2:56 AM Drouvot, Bertrand <bdrouvot@amazon.com> wrote:
>>> So you have in mind to check for XLogLogicalInfoActive() first, and if true, then open the relation and call
>>> RelationIsAccessibleInLogicalDecoding()?
>> I think 0001 is utterly unacceptable. We cannot add calls to
>> table_open() in low-level functions like this. Suppose for example
>> that _bt_getbuf() calls _bt_log_reuse_page() which with 0001 applied
>> would call get_rel_logical_catalog(). _bt_getbuf() will have acquired
>> a buffer lock on the page. The idea that it's safe to call
>> table_open() while holding a buffer lock cannot be taken seriously.
> Yes - that's pretty clearly a deadlock hazard. It shouldn't too hard to fix, I
> think. Possibly a bit more verbose than nice, but...
>
> Alternatively we could propagate the information whether a relcache entry is
> for a catalog from the table to the index. Then we'd not need to change the
> btree code to pass the table down.

Looking closer at RelationIsAccessibleInLogicalDecoding() It seems to me 
that the missing part to be able to tell whether or not an index is for 
a catalog is the rd_options->user_catalog_table value of its related 
heap relation.

Then, a way to achieve that could be to:

- Add to Relation a new "heap_rd_options" representing the rd_options of 
the related heap relation when appropriate

- Trigger the related indexes relcache invalidations when an 
ATExecSetRelOptions() is triggered on a heap relation

- Write an equivalent of RelationIsUsedAsCatalogTable() for indexes that 
would make use of the heap_rd_options instead

Does that sound like a valid option to you or do you have another idea 
in mind to propagate the information whether a relcache entry is for a 
catalog from the table to the index?

Regards,

-- 
Bertrand Drouvot
Amazon Web Services: https://aws.amazon.com