| From: | Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | Vadim Mikheev <vmikheev(at)reveredata(dot)com>, Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp>, Manfred Koizar <mkoi-pg(at)aon(dot)at>, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Nested transactions: low level stuff |
| Date: | 2003-03-21 15:53:05 |
| Message-ID: | 200303211553.h2LFr5i29478@candle.pha.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Tom Lane wrote:
> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > Do we want UNDO just for subtransactions?
> > That was pretty easily defeated, though I made an argument that you
> > could do UNDO pretty cheaply when you have WAL ensuring crash recovery.
>
> That argument was what got us into the early-7.1 WAL bloat problems.
> I don't think it's "pretty cheap" to have to hold the entire WAL for the
> length of your longest-running transactions.
With my idea, you wouldn't have to keep WAL around. Each backend would
keep a list of tids or the relid (if lots of rows are changed) in local
memory and UNDO on subtransaction abort.
--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Bruce Momjian | 2003-03-21 15:54:18 | Re: Roadmap for FE/BE protocol redesign |
| Previous Message | Tom Lane | 2003-03-21 15:49:54 | Re: Roadmap for FE/BE protocol redesign |