Duncan posted on Sat, 24 Aug 2013 12:23:18 +0000 as excerpted: > Martin Steigerwald posted on Fri, 23 Aug 2013 14:58:07 +0200 as > excerpted: > >> Am Freitag, 23. August 2013, 12:29:42 schrieb Xavier Bassery: >>> On Fri, 23 Aug 2013 11:38:56 +0200 >>> >>> David Kofler <dkofler92@xxxxxxxxxxxxxx> wrote: [mount-option discussion] One that I omitted in my previous discussion because it's not btrfs specific, but that makes a rather larger difference than normal on btrfs, especially with lots of snapshots, is noatime. Unfortunately, for legacy reasons, that can't be the default, but unless you're running mutt (which needs atime to track read messages except with mbox) or something similar, there really are very few even half-modern apps that require atime, and it really is best to run with noatime, saving the constant access-time update writes. The general kernel default is now relatime, which is a compromise that works reasonably well on most filesystems, but even that tends to trigger way more atime updates and thus de-dupped metadata when dealing with btrfs snapshots than would be optimal. Similarly, limited-write-cycle SSDs will do best with noatime. If you happen to be running btrfs on ssd as I am, well... So just run with noatime unless you are running something (like mutt) that is known to need atimes, and then, preferably limit it to a particular dedicated filesystem, perhaps choosing a filesystem other than btrfs or at least a limited-snapshot btrfs. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman -- 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
