[OpenAFS] Migration

Christopher Allen Wing wingc@engin.umich.edu
Tue, 23 Mar 2004 12:08:29 -0500 (EST)


As previous posters have written, you'll want to think about separating
the data into different AFS volumes based on type.

Use read-only/replicated AFS volumes where appropriate.

For data that will be written, figure out who the 'owners' and 'groups'
should be.

You will want to coordinate the user names with the numeric IDs.

"Home directories" should be treated according to AFS standards, e.g.:

	1. You will probably want all files in the home directories to
	   have the same uid/gid (this may require changing the uid/gid
	   when you copy it over)

	2. The root of the home directory should have an ACL like:

		owner all
		system:anyuser l

	3. All subdirectories should only have:

		owner all

	4. Areas of a 'home directory' that want to be public/shared need
	   to be handled carefully.



Some unix permissions can not be expressed in AFS, for instance:

	anchor% ls -l
	total 0
	-rw-------    1 wingc    staff           0 Mar 23 11:46 file1
	-rw-r--r--    1 wingc    staff           0 Mar 23 11:46 file2


This will not work in AFS, because 'readability' is a per-directory
attribute, not a per-file attribute! (This is something to watch out for,
because if you set the ACL so that 'file2' is world-readable, then 'file1'
will also be world-readable)


We moved a bunch of stuff from local storage into AFS last year, so we
might have a script lying around for reference.


-Chris Wing
wingc@engin.umich.edu




On Mon, 22 Mar 2004, Matthew Turk wrote:

> Hi there!  At my installation site, we're currently in the process of
> migrating from an NFS mounted 'storage' system to OpenAFS.  The storage
> system uses about 800gigs of a total of 1.5 terabytes of space, and I was
> unable to find any information on migration.
>
> Is there a preferred migration path?  A way of importing existing file
> systems into OpenAFS, with perhaps mapped usernames (or even not --
> resetting the ACLs isn't difficult, as the file structure is quite
> coherently structured.)
>
> Thanks!
>
> mjt
> Penn State Astronomy Dep't