On Wed, Feb 26, 2014 at 6:10 AM, David Champion <dgc@xxxxxxxxxxxx> wrote:
> I have a 1.6 TB collecton of 8 million files in CephFS, distributed up
> to 8-10 directories deep. (Never mind why - this design decision is out
> of my hands and not in scope.) I need to expose this data on multiple
> application servers. For the sake of argument, let's say I'm exposing
> files from CephFS via Apache. (This is true but it's not the only use
> case.)
>
> This is pretty slow to search, but so be it. However, if someone wants
> to copy the entire body of files, that's something I should be able to
> speed up. I see two options:
>
> 1. Place the 8m files in a disk image. Mount the disk image (read-only)
> to provide access to the 8m files, and allow copying the disk image to
> accelerate read of the enture dataset.
>
> 2. Put the 8m files in an RBD, and mount that instead. I guess if it's
> RO I can map it to multiple heads -- true?
Should be fine.
>
> Questions:
>
> q1. CephFS has a tunable for max file size, currently set to 1TB. If
> I want to change this, what needs to be done or redone? Do I have to
> rebuild, or can I just change the param, restart services, and be off?
What version are you running? It varies on whether that's set at FS
creation time or (on new enough code, but I don't remember when
off-hand) can be set via the cli ("ceph mds set max_file_size
<size_in_bytes>".
> q2. Sounds fine, except then the only access to the RBD raw blocks is
> via the block dev in /dev. I expose the CephFS mount to users, but not
> /dev. Is there a way to map the RBD as a pseudo-file within the CephFS
> mount? If not, then perhaps I'm looking at a bind/loopback mount of
> /dev/rbd/rbd into the user-visible namespace?
No, definitely no mapping of RBD into CephFS. It's a completely
different data format.
-Greg
Software Engineer #42 @ http://inktank.com | http://ceph.com
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
