Re: "nointegrity" for flash drives?
Dave Kleikamp <shaggy <at> linux.vnet.ibm.com>
2009-06-11 22:57:03 GMT
On Mon, 2009-06-08 at 09:14 +0100, Peter Grandi wrote:
> Being tempted by a nice 2.5" flash drive, I was wondering
> whether using "nointegrity" as a mount option would mitigate
> repeated writes to the journal area by disposing of the journal.
Yes, it should help in that regard. It will disable writes to the
journal.
> But it is not clear to me what "nointegrity" actually does, and
> whether JFS will behave nicely with it. By nicely here I don't
> mean quick restarts without 'fsck', I mean whether updates will
> be done in a way that would minimize corruption of the filesystem.
Really, there's no guarantee on what happens if the volume isn't
unmounted cleanly. The original motivation for "nointegrity" was to
allow a file system to be restored quickly from a backup (or otherwise
populated in a repeatable fashion). If the system would fail for
whatever reason, it was expected that the volume would be reformatted
and the restore would be restarted from scratch.
> One amusing aspect of flash drives is that not only is a journal
> bad news (especially bad if written in chunks of less than 32KiB)
> but they also should also make 'fsck' a lot faster (really small
> access times).
jfs really wasn't designed with flash drives in mind, and there are
probably better options out there.
> I still value JFS anyhow because of features, stability and
> things like indexing.
(Continue reading)
RSS Feed