From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | "Ben-Nes Michael" <miki(at)canaan(dot)co(dot)il> |
Cc: | "Bruno Wolff III" <bruno(at)wolff(dot)to>, "postgresql" <pgsql-general(at)postgresql(dot)org> |
Subject: | Re: Q about transactions |
Date: | 2003-04-14 14:10:13 |
Message-ID: | 7488.1050329413@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
"Ben-Nes Michael" <miki(at)canaan(dot)co(dot)il> writes:
> When I want to insert a branch to the tree I need:
> 1. select lft, rgt from the table to find where to insert
> 2. using update I open a gap in the tree ( offset all the lft, rgt above the
> location by two )
> 3. using insert I add a new branch.
> so "select for updates" seem inappropriate as I select only one row as I
> understood it lock only this row.
> Read Committed is also bad as if the second transaction will use select
> before the first transaction ran the update the select will be useless.
> so it seems that Serializable isolation is the solution, or am I wrong ?
Yeah. Use serializable mode, and be prepared to cope with "can't
serialize" errors (a retry loop around the whole transaction is the
standard answer to that).
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | scott.marlowe | 2003-04-14 15:17:41 | Re: Index information and log disable... |
Previous Message | Bob Kline | 2003-04-14 12:22:53 | Re: Upgrade to Red Hat Linux 9 broke PostgreSQL |