Hello. My name is Linda, and I work in the training courseware
development department.
To address your question about separating your db servers so that one
is in Building 1, and the other in Bld 2.....
If the network goes down between the buildings, then the db server
processes will elect a quorum (RW database) between the servers in Bld
2. What this means is that the clients in Bld 2 will have access to a
RW database (can change the dbs with command such as kpasswd, pts
creategroup, etc). They will be able to access data provided they can
get to all the volumes down the path to that data (be sure the top of
your tree is replicated with copies on file servers located in Bld 2.
For instance a user with a home volume on a server located in Bld 2
can only access his home directory if all of the volumes starting with
/afs (root.afs) are replicated down to his home volume (which of
course is not replicated :).
As for clients in Bld 1, with AFS 3.4 (with prior releases the pt
server had problems with this situation), the db server there will
become RO, until the network problem is resolved. This means that
with AFS 3.4, the users in this Bld will not be able to change db
information (kpasswd, pts cg will fail), but they will get RO data
that will allow them to klog, and access data. Remember they can only
access data that is located on file servers in Bld 1, and they have to
be able to get to all volumes in the path to the data, so be sure to
replicate the tree with ROs in Bld 1 as well.
Let me know if you have any further questions!
Linda
