[OpenAFS] Cache partition choice still limited to ext2 on Linux?

Marc Dionne marc.c.dionne@gmail.com
Wed, 7 Nov 2012 09:02:44 -0500


On Wed, Nov 7, 2012 at 5:49 AM, Dirk Heinrichs <dirk.heinrichs@altum.de> wrote:
>
> Am Mittwoch 07 November 2012, 11:38:07 schrieb Lars Schimmer:
>
> > On 2012-11-07 11:09, Dirk Heinrichs wrote:
> > > Hello,
> > >
> > > if I remember right, somebody wrote a few weeks ago, that with
> > > 1.6.x the choices for the cache partition on Linux systems aren't
> > > limited to ext2 only anymore. Is this correct?
> >
> > Correct.
> >
> > > If yes, which filesystems are possible to use (which ones are not)?
> > > Does the cache even need its own partition nowadays?
> >
> > Usual all FS are possible, some are useless (NFS e.g.). Would be
> > interesting to see the cache on OpenAFS itself^^
>
> Didn't think of any network fs at all.

Technically speaking the requirement is that the underlying filesystem
supports the exportfs API, or in other words is exportable via NFS.
That would rule out OpenAFS - not that it would be a good idea or that
it would work for many other reasons.  But most major/popular Linux
filesystems should work.

> > It still needs a partition, but you can loopmount a file for it. Or
> > use memcache.
>
> Last time I used memcache, I had issues with Java applications (Eclipse,
> SQLDeveloper). They brought the system to high load until they were finally
> OOM-killed when run under KDE on a machine with 4G RAM (512M or 1G of which
> set apart for the memcache).

In my (limited) experience with memcache, it doesn't behave very well
if the system is memory contrained and is under pressure.

> > I still would go with ext2 or ext3. But maybe some other fs is a bit
> > faster, never tested.
>
> Thanks a lot. The idea was to setup a btrfs-only VM for testing purposes, and
> use a subvolume for the cache. Not sure about the loop-mount method in this
> case, since you also can't put swapfiles on btrfs.

I have some machines that have used btrfs as a cache for a long time.
It initially exposed a few bugs in the caching code but I'm not aware
of any issues in the 1.6 releases.

Marc