[OpenAFS] A question about volume replication

[Derrick J Brashear]

On Tue, 12 Jun 2001, Hartmut Reuter wrote:

> wangkun@legend.com wrote:
> > 
> > I have three servers machines in my cell,one first server,two addtional
> > server.All the three are both file server and database server.
> > I made an examination on volume replication to see what could happen by
> > replicating some volumes on all the three machines. In my oppinion:
> >       when the first server corrupts,AFS system can run normally;
> >       when the machine with the original R/W volume corrputs,other readonly
> > copys can work proper           although they refuse any modification.
> > But nothing desirable happened!I saw that:
> >       when the first server corrupts,AFS system dies.
> >       when the machine with the original R/W volume corrputs,you can not
> > see the directory on which the R/W volume has mounted,saying nothing of
> > access it or its readonly copys.
> > I must make some mistake and need help.Thanks in advance.
> 
> As I wrote in my previous reply: in order to get the RO volume this must
> be mounted in another RO-volume and this back the whole path to /afs.
> This makes sense only for volumes with a constant contents. in order to
> get the RW-volume you either must have mounted it in another RW volume
> or with the -rw option. There is no mechanism that switches from RW to
> RO in the case the RW volume is gone.

The RW isn't supposed to become "gone".

Also, the RW may have had updates since the RO was cloned, so just
silently making it happen might mean you didn't notice data loss, for
instance

-D