| From: | Jelte Fennema-Nio <postgres(at)jeltef(dot)nl> |
|---|---|
| To: | Jacob Champion <jacob(dot)champion(at)enterprisedb(dot)com> |
| Cc: | Tatsuo Ishii <ishii(at)postgresql(dot)org>, tgl(at)sss(dot)pgh(dot)pa(dot)us, hlinnaka(at)iki(dot)fi, peter(at)eisentraut(dot)org, pgsql-hackers(at)postgresql(dot)org, Dave Cramer <davecramer(at)gmail(dot)com> |
| Subject: | Re: BackendKeyData is mandatory? |
| Date: | 2025-07-03 06:37:47 |
| Message-ID: | CAGECzQSxGPVY8U1Wy8hMG=yeeVk35bP3Qr-Kkww4wZy=uFunyw@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Thu, 3 Jul 2025 at 02:03, Jacob Champion
<jacob(dot)champion(at)enterprisedb(dot)com> wrote:
> Okay, wait -- JDBC was _copying_ our weird behavior? Why? Does
> something depend it in the wild?
I'm pretty sure there was no intent behind this, but it's simply
because our weird behaviour is the simplest from the client
implementation side with a fixed 4-byte key length:
1. Store two ints in a struct/class, these default to 0.
2. When you receive a BackendKeyData during startup update the two ints
3. When client wants to cancel, send a CancelRequest with the two ints.
So if step 2 is missed, then you send this all zero message. PgBouncer
has the same behaviour.
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Shinya Kato | 2025-07-03 06:40:07 | Re: Extend COPY FROM with HEADER <integer> to skip multiple lines |
| Previous Message | shveta malik | 2025-07-03 06:34:33 | Re: Using failover slots for PG-non_PG logical replication |