[OpenAFS] Changing the hostname of AFS server
Thu, 4 Dec 2008 13:39:26 -0800
Jaime - Excellent information. Thank you. I've been living in a
vacuum wrt AFS, and it's nice to have a dialog about the capabilities
and possibilities. I've already moved a test volume successfully, but
that was a on a test system. Now I need to understand physical
volumes as they relate to an AFS volume.
On my test machine, I had only one huge filesystem (partition). I
created /vicepa and /vicepb as directories. Unfortunately, when I
attempted to create the AFS volume, it rejected those directories
since they were actually not mounted filesystems. To workaround it, I
actually created a big file and used losetup to create a pseudo
filesystem which could be mounted for testing. It worked, but it was
only test and not for production.
Bottom line seems to be that /vicepXX must be mounted filesystems and
not just directories created under the root.
On Thu, Dec 4, 2008 at 12:15 PM, Jaime Cifuentes <firstname.lastname@example.org> wrote:
> AFS operates under certain structure. It requires it's own partitions to
> store the AFS data, and these are the /vicepx. These partitions, to the best
> of my knowledge, could be as large as the FS Type permits. However I thinks
> there is a max size AFS permits, which escapes my mind at the moment. So,
> you do have to have dedicated filesystems for AFS, and you do need to name
> them /vicepx. I believe AFS supports up to 256 partitions on a given server.
> Second thing I want to mention is that if you already created the new AFS
> Fileservers, you can just "move" volumes from the old servers to the new
> one, without disrupting the current AFS file structure or data accesibility.
> If the new AFS servers are both Fileservers and Database servers, then the
> only thing you need to do is to move the data and once the old servers are
> empty, you can just sunset them. You can even run a foreground script to
> move volumes. I would do something like: (with an "admin" id, or admin
> vos listvol -server <server name> -partition <part_name> |
> grep -v .backup |grep -v .readonly |
> while read vol_name rest
> vos move -id <vol name or id> -fromserver <from server name> -frompartition
> <from part name> -toserver <to server name> -topartition <to part name>
> Make sure you do add replicas of those .readonly volumes to another server
> once finished. BTW, these .readonly volumes are removed with "vos remsite",
> and added to a different server with "vos addsite"
> For a command reference. visit: http://www.openafs.org/doc/index.htm
> I hope this helped a bit
> Jaime Cifuentes
> DCE/DFS-AFS Support, Hitachi
> San Jose, CA
> 408 717-7998
> "dave first" <email@example.com>
> "dave first" <firstname.lastname@example.org>
> Sent by: email@example.com
> 12/04/2008 10:59 AM
> "David R Boldt" <firstname.lastname@example.org>, email@example.com
> Re: [OpenAFS] Changing the hostname of AFS server
> Thank you for responding.
> Actually, I do have the server up and running under OpenAFS on Linux.
> However, I did this under the quickly departing tutelage of the admin
> who has left, and don't really understand some nuances about the
> structure yet.
> For example, the old Transarc AFS server has distinct and separate
> files systems for each mount point:
> This made sense back in the day. But, now, as disks volumes can be
> quite large, and on Linux one RAID volume for a disk can be more than
> adequate to accommodate all the data we require to be held on AFS. In
> other words, there is a 1:1 correspondence of volume to physical
> So, the question then becomes, can I just use directories under root (
> / ), when I move the old data volume to the new server, or must I keep
> the same physical structure for a seamless move?
> From what I can see, I need to keep the same structure, unless there
> is a way to map a volume to a subdirectory so that the user need not
> be disturbed by the new physical location of the actual data.
> I'm sure I must be missing some information about this.... Also, is
> this an appropriate question for the openafs-info list, or should I be
> asking this on the openafs-user list? Which would be more
> On Thu, Dec 4, 2008 at 8:55 AM, David R Boldt <firstname.lastname@example.org> wrote:
>> I empathize with your situation, about seven years ago I started
>> down the path of learning AFS had set up the first cell in USGS.
>> My expertise is nowhere near that of the developers who
>> regularly contribute to this forum, but I would be happy to contribute
>> what small bits of wisdom I've managed to accumulate.
>> My understanding is that AFS is largely oblivious to hostnames
>> and uses IP addresses for all internal data structures.
>> Not sure from your description whether you have already set up
>> a new AFS fileserver on your Linux box or not.
>> "dave first" <email@example.com>
>> Sent by: firstname.lastname@example.org
>> 12/03/2008 04:50 PM
>> [OpenAFS] Changing the hostname of AFS server
>> I am new to AFS. I have inherited an old, crusty, but functional
>> Transarc AFS cluster of three servers running on Solaris 5.8.
>> Unfortunately, the hardware has become unstable. I have installed
>> OpenAFS server on Linux which communicates with the old AFS service.
>> My goal is to migrate all the AFS services to the new Linux OpenAFS
>> server which would then become the primary server. I have a few
>> hurdles to overcome before I can do that. One hurdle is that my
>> knowledge of AFS is minuscule, but growing. I've barely scratched the
>> My question for today is: How integrated into OpenAFS is the server's
>> hostname? Our Linux server name needs to be changed - easy enough to
>> do on Linux, but what needs to be updated in AFS? My assumption is
>> that as long as I touch all the correct config files, and restart the
>> service, all will be happy. Is that correct?
>> Since this server also talks with the old AFS server, I would need to
>> update the configuration on that as well.
>> Any advise or pointers would be most welcome! I am sure I'm not the
>> first one to travel this migration path - so any of you who have "been
>> there, done that" please chime in!
>> OpenAFS-info mailing list
> OpenAFS-info mailing list