Received: from ns1.engrs.org (ns1.engrs.org [216.54.29.10]) by hub.org (8.10.1/8.10.1) with ESMTP id e7MHTAT50995 for ; Tue, 22 Aug 2000 13:29:11 -0400 (EDT) Received: (from uucp@localhost) by ns1.engrs.org (8.9.3/8.9.3) id NAA05356; Tue, 22 Aug 2000 13:42:10 -0400 Received: from UNKNOWN(216.54.52.157), claiming to be "jupiter.greatbridge.com" via SMTP by ns1.engrs.org, id smtpdmt8Iqa; Tue Aug 22 17:42:04 2000 Received: (from janwieck@localhost) by jupiter.greatbridge.com (8.9.3/8.9.3) id NAA01610; Tue, 22 Aug 2000 13:31:31 -0500 From: Jan Wieck Message-Id: <200008221831.NAA01610@jupiter.greatbridge.com> Subject: Re: Continuous inserts... In-Reply-To: from Stephan Szabo at "Aug 22, 2000 08:59:29 am" To: Stephan Szabo Date: Tue, 22 Aug 2000 13:31:31 -0500 (EST) CC: Joerg Hessdoerfer , pgsql-sql@postgresql.org X-Mailer: ELM [version 2.4ME+ PL68 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Archive-Number: 200008/232 Stephan Szabo wrote: > Wierd, I've not seen that behavior really, although I've never > done time sensitive stuff. It might be the time before the > shared cache updates? Not sure really. If you do the rule > inline with your inserts (rather than a second transaction) > does it still wait? Just jumping in not having followed the discussion. But... The rules applied to a table by the rewriter are taken out of the relation descriptor that is returned by heap_open() or heap_openr(). I haven't looked at the code, but pg_class only has a boolean telling if a class has rules or not. Could it be that adding more rules (or dropping just a few instead of all) doesn't update the pg_class tuple, thus the syscache for the table isn't invalidated and other backends continue to use the old information instead of rescanning pg_rewrite? Jan > > Stephan Szabo > sszabo@bigpanda.com > > On Tue, 22 Aug 2000, Joerg Hessdoerfer wrote: > > > Hi! > > > > At 08:18 18.08.00 -0700, you wrote: > > [...] > > > > >I didn't try with vacuum, I just did a table lock and that > > >seemed to still hang the inserts with two tables, so I figured > > >maximum safety was adding the third table. If it works with two > > >that's much cooler. Was this with real data or just a small test > > >set? > > > > It was a test set ... ~20000 records, *BUT* I found that postgres > > decides when it starts to use the rule - means, if you do continous > > inserts on the table and create the rule, there's a varying time until > > the rule applies. In my first tests, I re-connected the DB very often, > > and the the change seemed immediate. > > > > Any ideas on how to 'promote' the rules faster?!? > > > -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #