[Unionfs] Using UnionFS on many many filesystems
Erez Zadok
ezk at cs.sunysb.edu
Tue Mar 25 17:21:48 EDT 2008
In message <6F79B9DB472D0C4F82EC633CBDEDAAE9618627 at elessar.cc.w2k.vt.edu>, "Fabian, Andy" writes:
> Are there any hard limits on the number of filesystems that can be
> joined concurrently into a single UnionFS filesystem? Are there any
> performance characteristics that drop more than linearly as the number
> of source filesystems increase?
>
> I'm thinking of using UnionFS in a backup system, to join potentially
> hundreds of incremental backups together. Is there any reason this
> shouldn't work?
Andrew,
Unionfs 2.x has one internal limit of 128 branches in can handle. This
limit can be changed in fs/unionfs/union.h:
/* maximum number of branches we support, to avoid memory blowup */
#define UNIONFS_MAX_BRANCHES 128
The reason for the limit, as the comment above states, is to avoid an
unbounded memory consumption. Imagine someone inserting more and more
branches into an existing union -- eventually you'll run out of physical
kernel memory.
So if you think you need more branches, try to change the above macro and
recompile and let me know.
Erez.
More information about the unionfs
mailing list