Received: from sss.sss.pgh.pa.us (sss.pgh.pa.us [209.114.166.2]) by hub.org (8.9.3/8.9.3) with ESMTP id NAA66213 for ; Sun, 24 Oct 1999 13:32:12 -0400 (EDT) (envelope-from tgl@sss.pgh.pa.us) Received: from sss.sss.pgh.pa.us (localhost [127.0.0.1]) by sss.sss.pgh.pa.us (8.9.1/8.9.1) with ESMTP id NAA03818; Sun, 24 Oct 1999 13:31:07 -0400 (EDT) To: "Albert REINER" cc: PostgreSQL-SQL Subject: Re: [SQL] Can VACUUM, but cannot do VACUUM ANALYZE In-reply-to: Your message of Sun, 24 Oct 1999 17:18:46 +0200 <19991024171846.B27105@frithjof> Date: Sun, 24 Oct 1999 13:31:06 -0400 Message-ID: <3816.940786266@sss.pgh.pa.us> From: Tom Lane "Albert REINER" writes: > I'm still using Linux kernel 2.0.35 (SuSE 5.3), but planning to > upgrade the kernel. I also built Postgres 6.5.1 with locale support. Hmm. I recall several different reports of weird problems seen with concurrently running backends that went away after an upgrade from kernel 2.0.35 or earlier to 2.0.36 or later. Apparently, some significant bugs in Linux's handling of SysV semaphores were corrected at 2.0.36 --- and Postgres depends heavily on semaphores. However, I wouldn't really expect that to be a problem unless you were running several backends concurrently. It wasn't clear if you had other things going on besides the VACUUM or not. I forgot to mention yesterday: if you track down the vacuum error message as I suggested, and it's something about "can't destroy lock file", then you just need to update to 6.5.2 and the problem should go away. If it's something else then more investigation will be needed. regards, tom lane