Additional message in pg_terminate_backend

On 13 Dec 2025, at 12:44, Roman Khapov <rkhapov@yandex-team.ru> wrote:

Recently I started working on patch for adding additional message from admin
in pg_terminate_backend for one of our greenplum fork. The main idea is that there must be
done little investigation every time you see 'FATAL: terminating connection due to administrator command’ to understand the reason, especially in cases where connection was terminated by another
user. So it was decided to create some new functions, that allows to terminate connection with
additional message.

Overall idea seems good to me.
Keep in mind that Postgres literals are translated into many languages. So text ought to be clear enough for translators to build a sentence that precedes termination reason.

I did POC patches with the next main ideas:
- lets add termReasonStr field in every PGPROC, that field can be used in ProcessInterrupts()
- implementation of pg_terminate_backend/pg_cancel_backend should be accessible from extensions, so lets move it in pg_terminate_backend_impl/pg_cancel_backend_impl and add definitions for it somewhere
- write simple extensions, which defines functions like pg_terminate_backend_msg, that sets termReasonStr and calls pg_terminate_backend_impl

First thing that we need to do is to agree on API of the new feature. We do not need core-extension separation for this.
My vote would be for having pg_cancel_backend(reason text)\pg_terminate_backend(reason text) along with parameterless versions.
32 bytes per PGPROC seems reasonable for a "reason". The patch doesn't seem to take care of cleaning "termReasonStr". Is it done elsewhere?
We have a race condition if many backends cancel same backend. Won't they mess each other's reason?

Thanks!

Best regards, Andrey Borodin.

On 13 Dec 2025, at 08:44, Roman Khapov <rkhapov@yandex-team.ru> wrote:

Recently I started working on patch for adding additional message from admin
in pg_terminate_backend for one of our greenplum fork.

Greenplum already has support for passing a message in the terminate command
doesnt it? Or at least it used to have but perhaps it was ripped out, my
memory is getting a bit fuzzy.

https://github.com/greenplum-db/gpdb-archive/commit/fa6c2d43d675aa05e2d9f797e3008f6fe075ee2c

--
Daniel Gustafsson

On 17 Dec 2025, at 21:02, Daniel Gustafsson <daniel@yesql.se> wrote:

Greenplum already has support for passing a message in the terminate command
doesnt it? Or at least it used to have but perhaps it was ripped out, my
memory is getting a bit fuzzy.

https://github.com/greenplum-db/gpdb-archive/commit/fa6c2d43d675aa05e2d9f797e3008f6fe075ee2c

Well, seems like I missed that patch..

Anyway, now I need same functionality in PostgreSQL, and your patch seems interesting,
especially in part where you keep messages in separated shmem region.

So I adopted your patch, maybe in that form it can be useful for PostgreSQL?

On Sat, 20 Dec 2025 at 14:27, Roman Khapov <rkhapov@yandex-team.ru> wrote:

On 17 Dec 2025, at 21:02, Daniel Gustafsson <daniel@yesql.se> wrote:

Greenplum already has support for passing a message in the terminate command
doesnt it? Or at least it used to have but perhaps it was ripped out, my
memory is getting a bit fuzzy.

https://github.com/greenplum-db/gpdb-archive/commit/fa6c2d43d675aa05e2d9f797e3008f6fe075ee2c

Well, seems like I missed that patch..

Anyway, now I need same functionality in PostgreSQL, and your patch seems interesting,
especially in part where you keep messages in separated shmem region.

So I adopted your patch, maybe in that form it can be useful for PostgreSQL?

+CREATE OR REPLACE FUNCTION
+  pg_terminate_backend_msg(pid integer, timeout int8 DEFAULT 0, msg text DEFAULT '')
+  RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS 'pg_terminate_backend_msg'
+  PARALLEL SAFE;

I don't think we need to create a function with a name other than
`pg_terminate_backend`. I also do not think we need
pg_terminate_backend_msg as a wrapper to another function - all of
this can be a single function, accepting different number of params,
exampli gratia "bt_index_check"

```
Datum
bt_index_check(PG_FUNCTION_ARGS)
{
Oid indrelid = PG_GETARG_OID(0);

...

if (PG_NARGS() >= 2)
args.heapallindexed = PG_GETARG_BOOL(1);
if (PG_NARGS() >= 3)
args.checkunique = PG_GETARG_BOOL(2);
....

PG_RETURN_VOID();
}

```

--
Best regards,
Kirill Reshke

Hi Roman,

As pointed out by Kirill, there is no reason to create
pg_terminate_backend_msg or pg_cancel_backend_msg. You can simply use
the existing functions and expand them to use one extra parameter, e.g.

CREATE OR REPLACE FUNCTION
pg_terminate_backend(pid integer, timeout int8 DEFAULT 0, msg text
DEFAULT '')
RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS
'pg_terminate_backend'
PARALLEL SAFE;

Here I don't think we need to check PG_NARGS, since the function calls
will always have a default value. Something like this perhaps:

Datum pg_terminate_backend(PG_FUNCTION_ARGS)
{
int pid;
int timeout; /* milliseconds */
char *msg;

pid = PG_GETARG_INT32(0);
timeout = PG_GETARG_INT64(1);
msg = text_to_cstring(PG_GETARG_TEXT_PP(2));

return pg_terminate_backend_internal(pid, timeout, msg);
}

stpncpy() -> strlcpy()

Documentation is missing -- assuming the feature design is already solid.

Add backend_msg.c to meson.build.

I rebased the patch (it was failing for quite some time) with some
suggestions. Feel free to remove them and revert to your v2 if you disagree.

I'll review the rest of code in the next days.

Best, Jim

On 01/02/2026 01:08, Jim Jones wrote:

As pointed out by Kirill, there is no reason to create
pg_terminate_backend_msg or pg_cancel_backend_msg. You can simply use
the existing functions and expand them to use one extra parameter

Since the message's size is limited to BACKEND_MSG_MAX_LEN, shouldn't
you use it to limit msg at pg_terminate_backend[_msg]()? Something like:

Datum pg_terminate_backend(PG_FUNCTION_ARGS)
{
int pid;
int timeout; /* milliseconds */
char msg[BACKEND_MSG_MAX_LEN];

pid = PG_GETARG_INT32(0);
timeout = PG_GETARG_INT64(1);
text_to_cstring_buffer(PG_GETARG_TEXT_PP(2), msg, sizeof(msg));

return pg_terminate_backend_internal(pid, timeout, msg);
}

Hi Jim!
Thanks for your review and rebase!

Since the message's size is limited to BACKEND_MSG_MAX_LEN, shouldn't
you use it to limit msg at pg_terminate_backend[_msg]()? Something like:

The message is truncated inside BackendMsgSet function, so I see a little
point in truncating it at pg_terminate_backend..

Also, changing the stpncpy() to strlcpy() breaks the logic
of returning result length of the message and NOTICE message about it,
so I reverted this change. But this note make me think about adding test
to truncation logic, so I added it in v4.

--
Best regards,
Roman Khapov

On 03/02/2026 08:52, Roman Khapov wrote:

The message is truncated inside BackendMsgSet function, so I see a little
point in truncating it at pg_terminate_backend..

Thanks for the update!

You might have a point there. One issue I see is with UTF8 character
boundaries. Calculating len like this can lead to invalid characters,
for instance:

postgres=# SELECT
pg_terminate_backend(pg_backend_pid(),0,repeat('🐘',1000));

NOTICE: message is too long, truncated to 127
FATAL: terminating connection due to administrator command:
🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘🐘�
server closed the connection unexpectedly
This probably means the server terminated abnormally

You might wanna take a look at other alternatives, such as pg_mbcliplen,
for instance (pseudocode):

len = strlen(msg);
if (len >= sizeof(slot->msg))
len = pg_mbcliplen(msg, len, sizeof(slot->msg) - 1);

memcpy(slot->msg, msg, len);
slot->msg[len] = '\0';

Best, Jim

On 3 Feb 2026, at 12:52, Roman Khapov <rkhapov@yandex-team.ru> wrote:

<v4-0001-message-in-pg_terminate_backend-and-pg_cancel_bac.patch>

Some notes on this version:

1. Did you mean BackendMsgShmemSize()?
size = add_size(size, BackendStatusShmemSize());

2. In docs:

<function>pg_cancel_backend</function> ( <parameter>pid</parameter> <type>integer</type>, <parameter>message</parameter> <type>test</type> <literal>DEFAULT</literal> <literal>''</literal> )

Did you mean <type>text</type>?

3. Windows build failed [0]https://github.com/x4m/postgres_g/runs/62314358734

4. In src/include/utils/misc/backend_msg.c identification is backend_msg.h

Best regards, Andrey Borodin.

[0]: https://github.com/x4m/postgres_g/runs/62314358734

Hi again Jim, Andrew!

Thanks for another round on review, updated the patch according to comments.

Also, fix `make check` by updating pg_proc.data with new functions,
similar to uuid7 way: defining _msg versions of the functions

--
Best regards,
Roman Khapov

Hi Roman

On 03/02/2026 13:28, Roman Khapov wrote:

Thanks for another round on review, updated the patch according to comments.

Also, fix `make check` by updating pg_proc.data with new functions,
similar to uuid7 way: defining _msg versions of the functions

Why did you decide to revert the pg_proc.dat and system_functions.sql
changes from v3? In v3 I thought that adding a msg DEFAULT '' to the
corresponding function in system_functions.sql ...

--- a/src/backend/catalog/system_functions.sql
+++ b/src/backend/catalog/system_functions.sql
@@ -400,7 +400,12 @@ CREATE OR REPLACE FUNCTION
   PARALLEL SAFE;

 CREATE OR REPLACE FUNCTION
-  pg_terminate_backend(pid integer, timeout int8 DEFAULT 0)
+  pg_cancel_backend(pid integer, msg text DEFAULT '')
+  RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS 'pg_cancel_backend'
+  PARALLEL SAFE;
+
+CREATE OR REPLACE FUNCTION
+  pg_terminate_backend(pid integer, timeout int8 DEFAULT 0, msg text
DEFAULT '')
   RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS
'pg_terminate_backend'
   PARALLEL SAFE;

... and updating pg_proc.dat accordingly would do the trick

diff --git a/src/include/catalog/pg_proc.dat
b/src/include/catalog/pg_proc.dat
index 5e5e33f64f..47aa30d716 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6724,10 +6724,11 @@

 { oid => '2171', descr => 'cancel a server process\' current query',
   proname => 'pg_cancel_backend', provolatile => 'v', prorettype => 'bool',
-  proargtypes => 'int4', prosrc => 'pg_cancel_backend' },
+  proargtypes => 'int4 text', proargnames => '{pid,msg}',
+  prosrc => 'pg_cancel_backend' },
 { oid => '2096', descr => 'terminate a server process',
   proname => 'pg_terminate_backend', provolatile => 'v', prorettype =>
'bool',
-  proargtypes => 'int4 int8', proargnames => '{pid,timeout}',
+  proargtypes => 'int4 int8 text', proargnames => '{pid,timeout,msg}',
   prosrc => 'pg_terminate_backend' },
 { oid => '2172', descr => 'prepare for taking an online backup',
   proname => 'pg_backup_start', provolatile => 'v', proparallel => 'r',

My rationale is based on the header in system_function.sql:

...
* src/backend/catalog/system_functions.sql
*
* This file redefines certain built-in functions that are impractical
* to fully define in pg_proc.dat. In most cases that's because they use
* SQL-standard function bodies and/or default expressions...

Am I missing something?

Thanks

Best, Jim

Hi all,

Thank you for the continued work on this patch.

Prafulla Ranadive and I reviewed this patch through the patch review
workshop. Overall, we find that the proposed feature is useful. We
have reviewed all five versions and have the following observations.

We agree with Jim's point of view that the v3 approach is better. v5
introduced separate _msg overloads (OID 8223 for
pg_cancel_backend_msg, OID 8222 for pg_terminate_backend_msg), which
adds catalog bloat and forces callers to use a different function
name:

+{ oid => '8223', descr => 'cancel a server process\' current query',
proname => 'pg_cancel_backend', provolatile => 'v', prorettype => 'bool',
proargtypes => 'int4 text', prosrc => 'pg_cancel_backend_msg' },
+{ oid => '8222', descr => 'terminate a server process',
proname => 'pg_terminate_backend', provolatile => 'v', prorettype => 'bool',
proargtypes => 'int4 int8 text', proargnames => '{pid,timeout,message}',
prosrc => 'pg_terminate_backend_msg' },

V3 keeps one OID per function and stays backward-compatible. v6
restores the v3 approach.

--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6727,19 +6727,14 @@

 { oid => '2171', descr => 'cancel a server process\' current query',
   proname => 'pg_cancel_backend', provolatile => 'v', prorettype => 'bool',
-  proargtypes => 'int4', prosrc => 'pg_cancel_backend' },
-{ oid => '8223', descr => 'cancel a server process\' current query',
-  proname => 'pg_cancel_backend', provolatile => 'v', prorettype => 'bool',
-  proargtypes => 'int4 text', prosrc => 'pg_cancel_backend_msg' },
+  proargtypes => 'int4 text', proargnames => '{pid,message}',
+  proargdefaults => '{""}'.
+  prosrc => 'pg_cancel_backend' },
 { oid => '2096', descr => 'terminate a server process',
-  proname => 'pg_terminate_backend', provolatile => 'v', prorettype => 'bool',
-  proargtypes => 'int4 int8', proargnames => '{pid,timeout}',
-  proargdefaults => '{0}',
-  prosrc => 'pg_terminate_backend' },
-{ oid => '8222', descr => 'terminate a server process',
   proname => 'pg_terminate_backend', provolatile => 'v', prorettype => 'bool',
   proargtypes => 'int4 int8 text', proargnames => '{pid,timeout,message}',
-  prosrc => 'pg_terminate_backend_msg' },
+  proargdefaults => '{0,""}',
+  prosrc => 'pg_terminate_backend' },

--- a/src/backend/catalog/system_functions.sql
+++ b/src/backend/catalog/system_functions.sql
@@ -378,6 +378,16 @@ BEGIN ATOMIC
 END;

+CREATE OR REPLACE FUNCTION
+  pg_cancel_backend(pid integer, message text DEFAULT '')
+  RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS 'pg_cancel_backend'
+  PARALLEL SAFE;
+
+CREATE OR REPLACE FUNCTION
+  pg_terminate_backend(pid integer, timeout int8 DEFAULT 0, message
text DEFAULT '')
+  RETURNS boolean STRICT VOLATILE LANGUAGE INTERNAL AS 'pg_terminate_backend'
+  PARALLEL SAFE;
+

The current approach scans all MaxBackends slots by PID twice. Since
pg_signal_backend() already calls BackendPidGetProc(pid) and has the
PGPROC* in hand, the PGPROC index into allProcs[ ] is the ProcNumber,
which is also the direct index into BackendMsgSlots[]. So v6 changes
the signature to BackendMsgSet(ProcNumber procno, ...) and the call in
pg_signal_backend() passes GetNumberFromPGProc(proc), making the
lookup O(1).

-int BackendMsgSet(pid_t pid, const char *msg)
+int BackendMsgSet(ProcNumber procno, const char *msg)
 {
    BackendMsgSlot      *slot;
    int                 len;
@@ -94,35 +94,26 @@ int BackendMsgSet(pid_t pid, const char *msg)
    if (msg == NULL || msg[0] == '\0')
        return 0;

-   for (int i = 0; i < MaxBackends; ++i)
-   {
-       slot = &BackendMsgSlots[i];
-
-       if (slot->pid == 0 || slot->pid != pid)
-           continue;
-
-       SpinLockAcquire(&slot->lock);
-
-       if (slot->pid != pid)
-       {
-           SpinLockRelease(&slot->lock);
-           break;
-       }
+   slot = &BackendMsgSlots[procno];

Also, fixed a stale file-header path in backend_msg.c:

--- a/src/backend/utils/misc/backend_msg.c
+++ b/src/backend/utils/misc/backend_msg.c
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * src/include/utils/misc/backend_msg.c
+ * src/backend/utils/misc/backend_msg.c

Removed dead NULL checks on msg in pg_cancel_backend_msg() and
pg_terminate_backend_msg() because text_to_cstring() never returns
NULL.

pid = PG_GETARG_INT32(0);
msg = text_to_cstring(PG_GETARG_TEXT_PP(1));
- if (msg == NULL) {
- PG_RETURN_BOOL(false);
- }

Furthermore, the else before the fallback ereport(FATAL) in the
terminate path is missing in v5. Without the else, the fallback
ereport is unreachable when BackendMsgIsSet() is true because
ereport(FATAL) never returns, but it misleads readers into thinking
the generic message is always emitted. Adding an else is also
consistent with the pg_cancel approach. v6 adds else to make the
intent explicit.

-           ereport(FATAL,
-                   (errcode(ERRCODE_ADMIN_SHUTDOWN),
-                    errmsg("terminating connection due to
administrator command")));
+           else
+               ereport(FATAL,
+                       (errcode(ERRCODE_ADMIN_SHUTDOWN),
+                        errmsg("terminating connection due to
administrator command")));
        }
    }

The _internal helper functions (pg_cancel_backend_internal,
pg_terminate_backend_internal) each have only one call site. What is
the intent behind keeping them separate, or can they be inlined
directly into pg_cancel_backend() and pg_terminate_backend()?

Andrey noted that:

We have a race condition if many backends cancel same backend.
Won't they mess each other's reason?

The spinlock in BackendMsgSet protects the write, but there is no
atomicity between writing the slot and delivering the signal. Another
backend can overwrite the slot between one writer's write and the
target reading it. Would it be valuable to define the behavior
explicitly, for example last-writer-wins, or document the limitation?

Andrey also raised this about translation:

Keep in mind that Postgres literals are translated into many languages.
So text ought to be clear enough for translators to build a sentence
that precedes termination reason.

The current pattern:

errmsg("terminating connection due to administrator command: %s", msg)

The colon-append pattern breaks in languages where the reason clause
is structured differently. Any suggestions on how to approach this, or
how similar cases are handled elsewhere in the codebase?

Also, parallel workers receive the generic message as they hit the
IsBackgroundWorker branch in ProcessInterrupts(), which bypasses
BackendMsgGet(). This is fine since parallel workers are ephemeral,
but should it be documented in the docs?

Patch attached.

--

Best Wishes,
Warda Bibi

Hi Warda and Prafulla

On 05/04/2026 21:27, warda Bibi wrote:

Prafulla Ranadive and I reviewed this patch through the patch review
workshop. Overall, we find that the proposed feature is useful.

Thanks for the nice review!

Roman, are you still planning to work on this? The patch has been
failing for several weeks.

https://cirrus-ci.com/github/postgresql-cfbot/postgresql/cf%2F6331

Best, Jim

Hi!

Thanks Warda for review!

Yes, I am planning to continue work on this patch, will send new version soon.

--
Best regards,
Roman Khapov