On Tue, Jan 16, 2007 at 12:40:40PM +0100, Achim Stumpf wrote:
> [root_at_clusty4 ~]# cfs listvol /coda/mytest.de/
>  Status of volume 7f000000 (2130706432) named "/"
>  Volume type is ReadWrite
>  Connection State is Connected
>  Reintegration age: 4294967295 sec, hogtime 4294967.295 sec
>  Minimum quota is 0, maximum quota is unlimited
>  Current blocks used are 2
>  The partition has 8191888 blocks available out of 8211208
> 
> I stopped with your advice after the client setup. I haven't created any 
> volumes. So I wonder now, if the /coda/mytest.de/ is now replicated over 
> those three servers or not?

That is simple,

    cfs whereis /coda/mytest.de

I think that the default setup created only a single replica, since the
volume got created when we set up the SCM before we had the other
servers up.

If it only has a single replica, then the most reliable method is to
remove the existing root volume and create a new one,

    purgevol_rep /

    createvol_rep / clusty1.mytest.de clusty2.mytest.de clusty3.mytest.de

The client may complain a bit once it notices, so you may have to
restart venus with -init.

> It seems to be the same partition (blocks available). The superuser is 
> able to store files on /coda/mytest.de/ (the root directory). I am 
> confused a bit about that. Is the /coda/mytest.de/ also replicated?

It is replicated, but probably only on a single site. There is some code
that should allow us to grow or shrink existing replicas, but it is not
always as reliable when we grow from a single replica since those
volumes don't have resolution logs associated with their directories.

> Is it the best way to setup only volumes under 
> /coda/mytest.de/<myvolume-mountpoint> and not to put files directly there?

Actually having a replicated root is useful for clients that are
starting up when your current SCM is unavailable, or when it's harddisk
dies.

All servers will have a copy of all the volume and user information, so
even with a dead harddrive you can recover. Any server can become SCM by
changing /vice/db/scm and restarting the auth2 and update daemons on all
servers in your Coda realm.

Jan