[OpenAFS] best way to recover after HD breakdown?

Jeffrey Hutzelman jhutz@cmu.edu
Wed, 08 Jun 2005 13:07:00 -0400


On Wednesday, June 08, 2005 01:12:02 PM +0200 Frank Burkhardt 
<fbo2@gmx.net> wrote:

> Hi,
>
> On Wed, Jun 08, 2005 at 10:29:04AM +0200, Lars Schimmer wrote:
>
> [snip]
>
>> One big HD in our cell just breaked down.
>> Now we've got some RO copies on 1.2.13 fileserver and some "non existent"
>> volumes, which RW copy were on that HD and no RO were available.
>> In 1.2.13 there was no way to convert RO to RW, right?
>
> There is - but you need a patch. Have a look at
>
>  ftp://instantafs.cbs.mpg.de/instantafs/common/openafs/1.2.13-1.1+rorw
>
> You'll find patched 1.2.13-debian-packages there.
>
>> If there is NO backup of that harddisk, only these RO, which is the best
>> way to recover that data?
>
> vos dump [...] | vos restore [...]

You can do a bit better than that, using 'vos copy'.  I can't remember 
whether this command appeared in any 1.2.x release, but it shouldn't matter 
- unlike convertROtoRW, it doesn't depend on new features in the volserver, 
so you can just use a 'vos' binary from a recent 1.3.x release against your 
existing servers.  Use a command like

vos copy foo.RO <fromserver> <frompart> foo <toserver> <topart> -live

Don't forget the -live; you can't copy from a RO volume without it.

This can considerably faster than dump|restore, because data is copied 
directly from one volserver to the other (or they can be the same server), 
without sending it to your workstation and back.

-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA