Обсуждение: pgsql: doc: remove verbiage about "receiving" data from rep. slots
doc: remove verbiage about "receiving" data from rep. slots The slots are just LSN markers, not something to receive from. Backpatch-through: master Branch ------ master Details ------- https://git.postgresql.org/pg/commitdiff/a554389fb5f96baba834c28bd4fc802fa1c8065f Modified Files -------------- doc/src/sgml/logical-replication.sgml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
Hi, On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > doc: remove verbiage about "receiving" data from rep. slots > > The slots are just LSN markers, not something to receive from. I think this is wrong. Logical slots also preserve resources other than WAL. Greetings, Andres Freund
Hi, On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > doc: remove verbiage about "receiving" data from rep. slots > > > > The slots are just LSN markers, not something to receive from. > > I think this is wrong. Logical slots also preserve resources other than WAL. I also think that the complaint about slots not being something to receive from is simply wrong for logical slots. Greetings, Andres Freund
On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > Hi, > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > The slots are just LSN markers, not something to receive from. > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > I also think that the complaint about slots not being something to receive > from is simply wrong for logical slots. Uh, can you clarify since I can't find details on what they preserve. The reason I liked the new wording is that the slot is only a small part of the receiving path. Almost by definition, the slot is specified as _part_ of the connection parameter --- it is not the connection parameter. Can you think of better wording that would be clearer than what we had? -- Bruce Momjian <bruce@momjian.us> https://momjian.us EDB https://enterprisedb.com Do not let urgent matters crowd out time for investment in the future.
Hi, On 2025-11-14 09:12:33 -0500, Bruce Momjian wrote: > On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > > Hi, > > > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > > > The slots are just LSN markers, not something to receive from. > > > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > > > I also think that the complaint about slots not being something to receive > > from is simply wrong for logical slots. > > Uh, can you clarify since I can't find details on what they preserve. Slots (physical ones can, logical ones always do), prevent row cleanup on the primary. > The reason I liked the new wording is that the slot is only a small part > of the receiving path. Almost by definition, the slot is specified as > _part_ of the connection parameter --- it is not the connection > parameter. I don't know what that has to do with anything. > Can you think of better wording that would be clearer than what we had? I don't see any need to revise the wording of the changed sections. Greetings, Andres Freund
On Fri, Nov 14, 2025 at 09:25:20AM -0500, Andres Freund wrote: > Hi, > > On 2025-11-14 09:12:33 -0500, Bruce Momjian wrote: > > On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > > > Hi, > > > > > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > > > > > The slots are just LSN markers, not something to receive from. > > > > > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > > > > > I also think that the complaint about slots not being something to receive > > > from is simply wrong for logical slots. > > > > Uh, can you clarify since I can't find details on what they preserve. > > Slots (physical ones can, logical ones always do), prevent row cleanup on the > primary. I see your point. How is the attached patch? The link right after the modified text talks only about WAL retention, so I think we need something in this sentence to say we do more than WAL retention for logical replication slots. -- Bruce Momjian <bruce@momjian.us> https://momjian.us EDB https://enterprisedb.com Do not let urgent matters crowd out time for investment in the future.
Вложения
Hi, On 2025-11-14 09:49:02 -0500, Bruce Momjian wrote: > On Fri, Nov 14, 2025 at 09:25:20AM -0500, Andres Freund wrote: > > Hi, > > > > On 2025-11-14 09:12:33 -0500, Bruce Momjian wrote: > > > On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > > > > Hi, > > > > > > > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > > > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > > > > > > > The slots are just LSN markers, not something to receive from. > > > > > > > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > > > > > > > I also think that the complaint about slots not being something to receive > > > > from is simply wrong for logical slots. > > > > > > Uh, can you clarify since I can't find details on what they preserve. > > > > Slots (physical ones can, logical ones always do), prevent row cleanup on the > > primary. > > I see your point. How is the attached patch? The link right after the > modified text talks only about WAL retention, so I think we need > something in this sentence to say we do more than WAL retention for > logical replication slots. It's less wrong, but I don't think this is the right place to add more documentation about what replication slots do. Greetings, Andres Freund
On Fri, Nov 14, 2025 at 09:55:37AM -0500, Andres Freund wrote: > Hi, > > On 2025-11-14 09:49:02 -0500, Bruce Momjian wrote: > > On Fri, Nov 14, 2025 at 09:25:20AM -0500, Andres Freund wrote: > > > Hi, > > > > > > On 2025-11-14 09:12:33 -0500, Bruce Momjian wrote: > > > > On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > > > > > Hi, > > > > > > > > > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > > > > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > > > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > > > > > > > > > The slots are just LSN markers, not something to receive from. > > > > > > > > > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > > > > > > > > > I also think that the complaint about slots not being something to receive > > > > > from is simply wrong for logical slots. > > > > > > > > Uh, can you clarify since I can't find details on what they preserve. > > > > > > Slots (physical ones can, logical ones always do), prevent row cleanup on the > > > primary. > > > > I see your point. How is the attached patch? The link right after the > > modified text talks only about WAL retention, so I think we need > > something in this sentence to say we do more than WAL retention for > > logical replication slots. > > It's less wrong, but I don't think this is the right place to add more > documentation about what replication slots do. Patch applied. I can keep improving it as we get more ideas. -- Bruce Momjian <bruce@momjian.us> https://momjian.us EDB https://enterprisedb.com Do not let urgent matters crowd out time for investment in the future.
On 2025-11-14 10:46:30 -0500, Bruce Momjian wrote: > On Fri, Nov 14, 2025 at 09:55:37AM -0500, Andres Freund wrote: > > On 2025-11-14 09:49:02 -0500, Bruce Momjian wrote: > > > On Fri, Nov 14, 2025 at 09:25:20AM -0500, Andres Freund wrote: > > > > Hi, > > > > > > > > On 2025-11-14 09:12:33 -0500, Bruce Momjian wrote: > > > > > On Fri, Nov 14, 2025 at 09:05:07AM -0500, Andres Freund wrote: > > > > > > Hi, > > > > > > > > > > > > On 2025-11-14 09:02:12 -0500, Andres Freund wrote: > > > > > > > On 2025-11-14 13:56:06 +0000, Bruce Momjian wrote: > > > > > > > > doc: remove verbiage about "receiving" data from rep. slots > > > > > > > > > > > > > > > > The slots are just LSN markers, not something to receive from. > > > > > > > > > > > > > > I think this is wrong. Logical slots also preserve resources other than WAL. > > > > > > > > > > > > I also think that the complaint about slots not being something to receive > > > > > > from is simply wrong for logical slots. > > > > > > > > > > Uh, can you clarify since I can't find details on what they preserve. > > > > > > > > Slots (physical ones can, logical ones always do), prevent row cleanup on the > > > > primary. > > > > > > I see your point. How is the attached patch? The link right after the > > > modified text talks only about WAL retention, so I think we need > > > something in this sentence to say we do more than WAL retention for > > > logical replication slots. > > > > It's less wrong, but I don't think this is the right place to add more > > documentation about what replication slots do. > > Patch applied. I can keep improving it as we get more ideas. I'm rather baffled by the process here. You pushed a patch, without any review, which also wasn't previously posted publicly, that made things worse. Then, after you got pushback on that change, you posted another patch, that got mildly negative feedback. Then you also pushed that change.
On Fri, Nov 14, 2025 at 12:14:15PM -0500, Andres Freund wrote:
> > Patch applied. I can keep improving it as we get more ideas.
>
> I'm rather baffled by the process here. You pushed a patch, without any
> review, which also wasn't previously posted publicly, that made things
> worse. Then, after you got pushback on that change, you posted another patch,
> that got mildly negative feedback. Then you also pushed that change.
Well, we talked privately on chat and I gave my reasons why I thought
the additional detail was helpful. When you stopped replying, I thought
that meant you were happy with the change, or had no ideas for
improvement. Obviously my assumption was wrong. As I stated, I am open
to continue improving it.
To reiterate what I said on private chat, our only docs about
replication slots are here:
https://www.postgresql.org/docs/devel/warm-standby.html#STREAMING-REPLICATION-SLOTS
It mostly talks about retaining WAL, and I cleaned that up to specify
the cases where only physical replication was involved. (You can see
yesterday's commits in the output.) Logical replication does more than
just retain WAL, and since we were only referencing the above
WAL-retention chapter from the subscription chapter:
https://www.postgresql.org/docs/devel/warm-standby.html#STREAMING-REPLICATION-SLOTS
I added a mention of "track relevant transaction activity". If that is
wrong, what is the right wording that adds details beyond WAL retention?
I personally was confused about replication slots, so I assumed these
changes will help others.
--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com
Do not let urgent matters crowd out time for investment in the future.