Elton Yang writes:
 > braam_at_cs.cmu.edu wrote:
 > 
 > > Elton Yang writes:
 > >  > Hi,
 > >  >       I'm a teacher in Peking Univ. of China, and I'm trying to build a
 > >  > distribute computing environment in our lab using CODA.
 > >  > I think CODA is really wonderful  system. Now I encounter sever
 > >  > problems, and I want your help eagerly:
 > >  >
 > >  > 1)I had created a replication volume among several servers,  several
 > >  > days later we want to add a new node  into the
 > >  >     replication group, i.e. we want the replication volume also created
 > >  > and appeared on new node. How can I do that?
 > >
 > > We don't think this is possible yet.  We hope to add this as a feature before too long.
 > >
 > 

 > This is really bad news for us. Anyway, thank you very much.

You can make changes to the code and fix it.

 > 
 > >
 > >  >
 > >  > 2)Sometimes our SCM may crash, is it possible to  prepare some nodes as
 > >  > subtitutes of SCM then when SCM is down one of the
 > >  >     subtitute can take place of it?
 > >
 > > Possibly, but wouldn't it be better for the SCM to be up and not crash?  Ours has been up for 6 months or so.
 > >
 > 

 > Because we are making a cluster environment with fault tolerance, every node could be down,
 > including  SCM, so we need a solution to this. Would you kindly give me a more detail steps
 > to subtitute SCM (when it's down)?

The SCM is distinguished in that it is the RW node for the
VLDB/VSGDB/VRDB etc. Otherwise it is a normal server.  All other
servers fetch these databases from the SCM using the updateclnt.

Every cluster member which is up will have to get the new scm hostname
into the /vice/db/scm file.  Then the update client on these members
needs to be restarted.  After that, this member will fetch all
database files from the new scm.

 > 
 > >
 > > > If possible. could you tell me the
 > >  > detail steps?
 > >  >     And also, when the old SCM recovered, how to deal with it, I think
 > >  > it can be a ordinary server, but how to configure it?
 > >
 > > It can be an ordinary server and can be treated just like any other server.  I don't understand the question.
 > >
 > 

 > When the old SCM recoveres and becomes a member server, is it
nessary to modify the config files
 > on it?

You should also make the recovered SCM talk to the new SCM - in the
same way as for any other server.

 > 
 > >
 > > Peter
 > 
 > I've another question(sorry), is there any method or command to force a server to integrate with
 > other replicated servers immidiately after the reconnection? The auto-integration seems to delay
 > for several minutes.  

I think you mean resolution.  It is done on demand.  To force it, do
ls -lR in the affected directories from a client.

- Peter -

Thank you for your answer!
 > 
 > Best Regards,
 > 
 > Elton
 > 
 >