Re: [HACKERS] make async slave to wait for lsn to be replayed

From: Kartyshov Ivan <i(dot)kartyshov(at)postgrespro(dot)ru>
To: pgsql-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: [HACKERS] make async slave to wait for lsn to be replayed
Date: 2023-02-28 10:10:47
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

The main purpose of the feature is to achieve
read-your-writes-consistency, while using async replica for reads and
primary for writes. In that case lsn of last modification is stored
application. We cannot store this lsn inside database, since reads are
distributed across all replicas and primary.

Lots of proposals were made how this feature may look like.
I aggregate them into the following four types.

1) Classic (wait_classic_v1.patch)
advantages: multiple events, standalone WAIT
disadvantages: new words in grammar

WAIT FOR  [ANY | ALL] event [, ...]
BEGIN [ WORK | TRANSACTION ] [ transaction_mode [, ...] ]
    [ WAIT FOR [ANY | ALL] event [, ...]]
where event is one of:
LSN value
TIMEOUT number_of_milliseconds

2) After style: Kyotaro and Freund (wait_after_within_v1.patch)
advantages: no new words in grammar, standalone AFTER
disadvantages: a little harder to understand

AFTER lsn_event [ WITHIN delay_milliseconds ] [, ...]
BEGIN [ WORK | TRANSACTION ] [ transaction_mode [, ...] ]
    [ AFTER lsn_event [ WITHIN delay_milliseconds ]]
START [ WORK | TRANSACTION ] [ transaction_mode [, ...] ]
    [ AFTER lsn_event [ WITHIN delay_milliseconds ]]

3) Procedure style: Tom Lane and Kyotaro (wait_proc_v1.patch)
advantages: no new words in grammar,like it made in
pg_last_wal_replay_lsn, no snapshots need
disadvantages: a little harder to remember names
SELECT pg_waitlsn(‘LSN’, timeout);
SELECT pg_waitlsn_infinite(‘LSN’);
SELECT pg_waitlsn_no_wait(‘LSN’);

4) Brackets style: Kondratov 
advantages: only one new word in grammar,like it made in VACUUM and
REINDEX, ability to extend parameters without grammar fixes
WAIT (LSN '16/B374D848', TIMEOUT 100);
BEGIN WAIT (LSN '16/B374D848' [, etc_options]);

Below I provide the implementation of patches for the first three types.
I propose to discuss this feature again/


Ivan Kartyshov
Postgres Professional:
The Russian Postgres Company

Attachment Content-Type Size
wait_after_within_v1.patch text/x-diff 28.1 KB
wait_classic_v1.patch text/x-diff 46.3 KB
wait_proc_v1.patch text/x-diff 16.2 KB


Browse pgsql-hackers by date

  From Date Subject
Next Message Jeevan Ladhe 2023-02-28 10:19:18 Re: Make some xlogreader messages more accurate
Previous Message qinghao huang 2023-02-28 09:56:00 Maybe we can remove the type cast in typecache.c