On 09/05/2011 06:29 PM, Hugo Mills wrote: > On Mon, Sep 05, 2011 at 12:25:21PM -0400, Christoph Hellwig wrote: >> On Mon, Sep 05, 2011 at 06:23:23PM +0200, Tomasz Chmielewski wrote: >>>> That's because dpkg is known for using (f)sync very heavily. btrfs >>>> honours the sync request in all cases, so it's much much slower than >>>> ext3, which doesn't. >>> >>> Hmm, is it really the case with ext3/ext4 (ignoring fsync in some cases)? >>> >>> Sounds like a bug in ext3/ext4 then. >>> >>> Is it documented anywhere where ext3/ext4 would just silently ignore fsync? >> >> They don't. Unteil recently ext3 and reiserfs would not flush the >> disk caches unless enabled by a mount option, but even that has recently >> been fixed. > > Apologies for disseminating my misunderstanding, then. > > Hugo. > I'm wondering if this may be the very same unacceptable behavior as discussed earlier also in the following message threads: btrfs always writes something after sync news://news.gmane.org:119/CABVOS09hntXYUsCJBg+byEGC5y53j4DeT6b49Rx1AaRpqFV54A@xxxxxxxxxxxxxx Re: Writes in idle/How to debug filesystem activity news://news.gmane.org:119/4E5C8A20.7030604@xxxxxxxx Re: Applications using fsync cause hangs for several seconds every few minutes news://news.gmane.org:119/4E41A6A4.9070702@xxxxxxx Btrfs slowdown news://news.gmane.org:119/CAO47_-9BLKWUGDEuzaLqHSq9tZkAUaO8FMQEy1pPk9A2Hb+5AQ@xxxxxxxxxxxxxx PLEASE TEST: Everybody who is seeing weird and long hangs news://news.gmane.org:119/4E36C47E.70309@xxxxxxxxxx and this: curious writes on mounted, not used btrfs filesystem http://thread.gmane.org/gmane.comp.file-systems.btrfs/10840 and probably this: 2.6.37: Multi-second I/O latency while untarring http://www.spinics.net/lists/linux-btrfs/msg08516.html Hasn't the slowdown been fixed yet? For those experiencing the slowdown issue btrfs is entirely useless... (I had to stop using btrfs because of these frequent slowdowns, while btrfs was doing most likely needless disk writes for multiple seconds, I couldn't work at all) Can someone look into this if this is not fixed? There are numerous analysis reports in the above threads, without any constructive result. Thanks -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
