RE: Newly created replication slot may be invalidated by checkpoint

On Friday, January 23, 2026 5:06 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Fri, Jan 23, 2026 at 7:33 AM Zhijie Hou (Fujitsu) <houzj(dot)fnst(at)fujitsu(dot)com>
> wrote:
> >
> > This patch cannot be applied cleanly on backbranches, I can prepare
> > patches for those once the main patch is stable.
> >
>
> Some comments:
> 1.
> + /*
> + * Determine the minimum non-removable LSN by comparing the redo
> + pointer
> + * with the minimum slot LSN.
> + */
> + min_safe_lsn = GetRedoRecPtr();
> + slot_min_lsn = XLogGetReplicationSlotMinimumLSN();
>
> Can we expand these comments a bit to state why we need both RedoRecPtr
> and slot's minimum LSN?

Added some comments for this.

>
> 2.
> +# Verify that while syncing a slot to the standby server, if the WAL
> +before the # remote restart_lsn is at risk of being removed by a
> +checkpoint, the slot # cannot be synced. Otherwise, even if the slot
> +syncing succeeds, it may be # immediately invalidated by the checkpoint.
> +my $primary = $node;
>
> This comment atop the testcase is not very clear. Because, it is testing that
> the slot is synced and is not invalidated. How about:
> "Verify that the synchronized slots won't be invalidated immediately after
> synchronization in the presence of a concurrent checkpoint."?

Thanks, I have changed to use the suggested version.

>
> 3.
> +# Increase the log_min_messages setting to DEBUG2 on both the standby
> +and # primary to debug test failures, if any.
> +my $connstr_1 = $primary->connstr;
>
> Do we need this DEBUG2? I don't think we should add too many DEBUG2
> tests as it increases Log volume.

Removed.

Here are the V4 patches for both HEAD and back branches.

Best Regards,
Hou zj