Re: BUG #19616: pgoutput sends stream abort ('A') to clients that did not enable streaming

Поиск
Список
Период
Сортировка
Искать
От
Andrey Rachitskiy
Тема
Re: BUG #19616: pgoutput sends stream abort ('A') to clients that did not enable streaming
Дата
в 07:56:41
Msg-id
CAB8bMiuRPBfvMxsf-gUay_XKtaT0xKR9T0LNRaGmfok7YhMHtA@mail.gmail.com
Список
Дерево обсуждения
BUG #19616: pgoutput sends stream abort ('A') to clients that did not enable streaming PG Bug reporting form <noreply@postgresql.org>
Dear Kuroda-san!

Thanks for the review.

I addressed both comments in the attached v3. 


пт, 14 авг. 2026 г. в 08:03, Hayato Kuroda (Fujitsu) <kuroda.hayato@fujitsu.com>:
Dear Tyler, Andrey,
(Adding Sawada-san in CC)

Good catch and thanks for the report. I confirmed that your reproducer causes an
Assert failure for the debug build. See:

```
(gdb) bt
#0  ReorderBufferMaybeMarkTXNStreamed (rb=0x3385430, txn=0x33a55a8)
    at ../postgres/src/backend/replication/logical/reorderbuffer.c:2154
#1  0x0000000000973d05 in ReorderBufferTruncateTXN (rb=0x3385430, txn=0x33a5440,
    txn_prepared=false) at ../postgres/src/backend/replication/logical/reorderbuffer.c:1677
#2  0x00000000009740ec in ReorderBufferCheckAndTruncateAbortedTXN (rb=0x3385430, txn=0x33a5440)
    at ../postgres/src/backend/replication/logical/reorderbuffer.c:1816
#3  0x0000000000977aa1 in ReorderBufferCheckMemoryLimit (rb=0x3385430)
    at ../postgres/src/backend/replication/logical/reorderbuffer.c:3985
#4  0x0000000000972345 in ReorderBufferQueueChange (rb=0x3385430, xid=696, lsn=24970056,
    change=0x33a8be0, toast_insert=false)
...
```

Few comments for the code:

```
+ * A top-level transaction is always marked.  A subtransaction is marked only
+ * when it has changes and its top-level transaction is already streamed.
```

The last sentence can be "its top-level transaction is already marked as streamed."

```
+       /*
+        * A subtransaction is marked only when it has changes, and only when its
+        * top-level transaction has already been marked as streamed.  We never
+        * stream XIDs of empty subxacts, and we must not send an abort for an XID
+        * the downstream has never heard of.
         *
-        * We do it this way because of aborts - we don't want to send aborts for
-        * XIDs the downstream is not aware of. And of course, it always knows
-        * about the top-level xact (we send the XID in all messages), but we
-        * never stream XIDs of empty subxacts.
+        * The top-level check matters because ReorderBufferTruncateTXN is also
+        * used to discard already-aborted transactions at eviction, where the
+        * top-level xact is not streamed.  Marking a subxact there would make a
+        * later abort emit stream_abort to a client that never enabled streaming.
         */
-       if (rbtxn_is_toptxn(txn) || (txn->nentries_mem != 0))
+       if (txn->nentries_mem != 0 && rbtxn_is_streamed(rbtxn_get_toptxn(txn)))
```

I feel the code comment might be too detail: second paragraph is not needed
for me.

Best regards,
Hayato Kuroda
FUJITSU LIMITED



--
Regards,
Rachitskiy Andrey
В списке pgsql-bugs по дате отправления
От: Zsolt Parragi
Дата:
FAQ