From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans Reiser Subject: Re: IO randomly blocked for >1 minute while disk writes still in 2.6.16.1 + 2.6.16-reiser4-1 Date: Thu, 30 Mar 2006 13:13:47 -0800 Message-ID: <442C4A0B.6070607@namesys.com> References: <20060330023909.GA7547@rvalles.homedns.org.> <442BD204.4030402@namesys.com> <442C32A0.8030200@dermichi.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <442C32A0.8030200@dermichi.com> List-Id: Content-Type: text/plain; charset="us-ascii" To: Michael Weissenbacher Cc: reiserfs-list@namesys.com, Edward Shishkin , vs Michael Weissenbacher wrote: >Hi, > > >>>of the movie. Now, just to know the format, I played it again, and as it >>>began, it blocked again... not even knoppix is downloading on the >>>background. >>> >>> >I just wanted to note that i observed that same bahavior. I tried >various things, like changing the kernel from preemption to no >preemtion, trying the various IO-Schedulers and trying 2.6.15, 2.6.16 >and 2.6.16-mm1. In the end i found out that there was a MAJOR >performance gain by adding the noatime mount option. Before i added that >I had to wait up to 20 minutes sometimes and now it doesn't stall more >than a few seconds, but it still does. Seems like reiser4 has very big >performance problems with the writing of atimes. > > I think this got fixed in recent versions, vs please comment. >Another problem i had was specific to the reiser4-for-2.6.16-1.patch.gz >patch. I got a kernel crash two times now, but wasn't yet able to log >the error message anywhere because it would not let me write to the >reiser4 partition after the error. It never corrupted any data though, >fsck.reiser4 always reported a consistent fs. I'll try to trigger that >problem again and make another post. > >