Horton

RE: Conflict detection for update_deleted in logical replication

Zhijie Hou (Fujitsu) <houzj.fnst@fujitsu.com>

View thread View on postgr.es
From: "Zhijie Hou (Fujitsu)" <houzj.fnst@fujitsu.com>
To: shveta malik <shveta.malik@gmail.com>
Cc: Amit Kapila <amit.kapila16@gmail.com>, vignesh C <vignesh21@gmail.com>, Masahiko Sawada <sawada.mshk@gmail.com>, Nisha Moond <nisha.moond412@gmail.com>, "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2025-05-26T07:16:39Z
Lists: pgsql-hackers 
On Fri, May 23, 2025 at 2:45 PM shveta malik wrote:

> 
> Thanks you for v31 patch-set. Please find few comments on patch001:

Thanks for the comments.

> 
> 1)
> 
> wait_for_local_flush:
> 
> + if (data->last_recv_time &&
> + TimestampDifferenceExceeds(data->flushpos_update_time,
> +    data->last_recv_time, WalWriterDelay))
> + {
> + XLogRecPtr writepos;
> + XLogRecPtr flushpos;
> + bool have_pending_txes;
> +
> + /* Fetch the latest remote flush position */
> + get_flush_position(&writepos, &flushpos, &have_pending_txes);
> +
> + if (flushpos > last_flushpos)
> + last_flushpos = flushpos;
> +
> + data->flushpos_update_time = data->last_recv_time;
> + }
> 
> We should only get new flush-position, if 'last_flushpos' is still
> lesser than 'data->remote_lsn'. Since 'last_flushpos' is also updated
> by 'send_feedback' and we do not update 'data->flushpos_update_time'
> there, it is possible that we have latest flush position but still
> TimestampDifferenceExceeds gives 'true', making it re-read the flush
> position unnecessarily.
> 
> Having said that, I think the correct way will be to move
> 'flushpos_update_time' out of RetainConflictInfoData() similar to
> last_flushpos. Let it be a static variable, then we can update it in
> send_feedback as well.

I added the check to update the flush when last_flushpos is behind.

But I tend to avoid adding more static variables if possible.
The flushpos_update_time is only used to prevent fetching flush position too
frequently which is specific to RCI logic. And even if in some cases, we might
have already updated the flush position in send_feedback(), I feel it's not a
big issue as the new flush position update logic is only used when applying
changes in a loop where send_feedback() is rarely invoked.

> 4)
> Everywhere we are using variable name as 'data', it is a very generic
> name. Shall we change it to 'conflict_info_data' or
> 'retain_conf_info_data'?

I changed them to rci_data which is shorter.

Best Regards,
Hou zj

Commits

  1. Fix intermittent BF failures in 035_conflicts.

  2. Resume conflict-relevant data retention automatically.

  3. Fix intermittent test failure introduced in 6456c6e2c4.

  4. Fix Coverity issue reported in commit a850be2fe.

  5. Add test to prevent premature removal of conflict-relevant data.

  6. Post-commit review fixes for 228c370868.

  7. Add max_retention_duration option to subscriptions.

  8. Detect and report update_deleted conflicts.

  9. Preserve conflict-relevant data during logical replication.

  10. Integrate FullTransactionIds deeper into two-phase code

  11. Improve checks for GUC recovery_target_timeline

  12. Prevent excessive delays before launching new logrep workers.

  13. Keep WAL segments by slot's last saved restart LSN

  14. Rework some code handling pg_subscription data in psql and pg_dump

  15. Use generateClonedIndexStmt to propagate CREATE INDEX to partitions.