Re: BUG #18213: Standby's repeatable read isolation level transaction encountered a "nonrepeatable read" problem

Yes, the existing "ProcArrayGetReplicationSlotXmin" function has met the functional requirements and I have updated it in the attachment.

Best regards,

Changhong Fei

------------------ Original ------------------
From: "zhihuifan1213" <zhihuifan1213(at)163(dot)com&gt;;
Date:&nbsp;Wed, Nov 29, 2023 03:50 PM
To:&nbsp;"费长红"<feichanghong(at)qq(dot)com&gt;;
Cc:&nbsp;"pgsql-bugs"<pgsql-bugs(at)lists(dot)postgresql(dot)org&gt;;
Subject:&nbsp;Re: 回复： BUG #18213: Standby's repeatable read isolation level transaction encountered a "nonrepeatable read" problem

"费长红" <feichanghong(at)qq(dot)com&gt; writes:

&gt; -------------------------------------------------------------------------------------------------------------------------
&gt;
&gt;&nbsp; *&nbsp; 起个啥名好呢&nbsp;
&gt;&nbsp;&nbsp;&nbsp; feichanghong(at)qq(dot)com&nbsp;
&gt;&nbsp;
&gt; Indeed, I simply implemented and verified the solution. In the above test, the "create index" command on RW will hang
&gt; until the transaction on standby is committed or aborted.

I think this is acceptable since its behavior is same as we wait for the
transactions in primary.

&gt; In addition, even if there is no select query on the standby, RW's "create index" command may wait for a period of time,
&gt; which affected by the wal_receiver_status_interval parameter.

The "wait" happens at the very last of index building, so the building
stages gives an enough time for standby to exceeds the xmin, so in a
real case, I think the overhead will be pretty low.

&gt; You can see attachment for the patch.

I take a look at the that, function GetReplicationSlotXmin should be not
needed. You can use ProcArrayGetReplicationSlotXmin directly.

After searching the caller of WaitForOlderSnapshots, I think this bug
probably has impact on "deteach partition concurrently" / "reindex
concurrently" features.&nbsp; All of them needs the same fix.

--
Best Regards
Andy Fan