From: Aaron Straus Subject: Re: [NFS] blocks of zeros (NULLs) in NFS files in kernels >= 2.6.20 Date: Mon, 22 Sep 2008 10:37:15 -0700 Message-ID: <20080922173715.GE12483@merfinllc.com> References: <20080905191939.GG22796@merfinllc.com> <56258A29-95D5-4A8C-A097-014B8FEDFB8F@oracle.com> <20080911184951.GB19054@merfinllc.com> <200809221805.48463.hpj@urpla.net> <1222101322.7615.6.camel@localhost> <20080922170414.GC12483@merfinllc.com> <76bd70e30809221026g7bde774pbffa35881682ea4b@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Trond Myklebust , Hans-Peter Jansen , linux-kernel@vger.kernel.org, Neil Brown , Linux NFS Mailing List To: chucklever@gmail.com Return-path: Received: from quackingmoose.com ([63.73.180.143]:49024 "EHLO penguin.merfinllc.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1751652AbYIVRhQ (ORCPT ); Mon, 22 Sep 2008 13:37:16 -0400 In-Reply-To: <76bd70e30809221026g7bde774pbffa35881682ea4b-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: Hi, On Sep 22 01:26 PM, Chuck Lever wrote: > Why can't you use O_SYNC | O_APPEND ? In our case, some of the writers are not in our control. Another case we see the issue is you spawn a job: job 1> out 2> err & tail -f out err -- Also, we actually do like caching the writes (b/c log files do not need to be checked immediately after being written). We just wish the cache could be written out in-order. If there's no way to make that happen... we can reorganize our file-system, exports, mounts, etc so that the log file directory is mounted with the sync option. However, the problem *might* be annoying and wide-spread enough to warrant a change. Anyway, thanks! Regards, =a= -- =================== Aaron Straus aaron-bYFJunmd+ZV8UrSeD/g0lQ@public.gmane.org