Hi,

I'm working on a project to provide CAS on our campus.  We're hoping to 
cluster two instances of CAS 3.3.5 on Tomcat 6 behind a load balancer.
In reading through the list archives and the wiki pages, it seems there 
are a number of different ways to cluster CAS.  I'm hoping to get some 
advice regarding these different options:

1. Tomcat session replication vs Spring Web Flow "request state"

Configuring Web Flow to use repository-type "client" seems simpler
to implement than clustering the two tomcat servers with multicast.
Are there any trade-offs or disadvantages to doing it this way?

2. JpaTicketRegistry and table locking issues

In our test deployment (one server) we're using the JpaTicketRegistry.
I've read the past discussions on the table locking issues that some 
sites have experienced.  Are these common to all implementations, or 
are there some databases that don't suffer from this problem?  We're 
considering both a single Oracle database and multiple Postgres 
databases with replication via pgpool.  Has anyone had success using 
pgpool with CAS?  If so, do you still use c3p0, or do you use pgpools 
own connection pooling?

3. Ticket Registry Cleaners

Is it a good idea to run the Ticket Registry Cleaner on both servers in
the cluster, or just one?  Or is it better to remove this altogether and
run external SQL deletes against the database to remove expired tickets?

Any advice is appreciated.  Thanks in advance...

Sandy


--
Sandy Gordon
Unix Systems Administrator
University of Victoria Libraries
[email protected] / 250-721-8266


-- 
You are currently subscribed to [email protected] as: 
[email protected]
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-user

Reply via email to