gjournal and TRIM: A safe combination?

[Adam Vande More]

On Thu, Oct 22, 2015 at 8:15 AM, Julien Cigar <jcigar at ulb.ac.be> wrote:

> On Wed, Oct 21, 2015 at 02:38:16PM -0500, Brandon J. Wandersee wrote:
> >
> > Julien Cigar writes:
> > > Correct me if I'm wrong but, unless there is a bug in Soft Updates
> (SU),
> > > your FS will, in theory, never be corrupted with SU only.
> >
> > If I read this right, then, would gjournal be overkill for typical
> > desktop/laptop use?  UFS2 is perhaps the only filesystems I've ever
> > encountered in common use that doesn't have data journaling enabled by
> > default, so that's thrown me off somewhat...
>
> It depends of your needs, but for a typicaly desktop use it seems a bit
> overkill to me.
>
> Contrary to other filesystems UFS doesn't need journaling to avoid
> _corruption_ as SU is an alternate approach to maintain metadata
> integrity.
>
> One important thing to remember is that neither SU, or SU+J, or any
> journaled metadata prevents data loss: corruption != data loss. This is
> where gjournal (and data=journal on EXT-like FS) comes in as datas and
> metadatas are written in the journal _before_, so you avoid corruption
> and data loss.
>

gjournal nor data=journal nor COW prevent data loss.  They simply ensure
both meta-data and data are consistent.  In general though they do reduce
the risk of data loss at a high cost in journaled scenarios.

-- 
Adam