Unfortunately, the internals of the Jazz Server are not designed to work
in a cluster. Even if you do get it installed into a cluster, there are
parts of the system that assume that the same instance will serve all
requests and having multiple servers all access the same DB will have
unknown effects. I would strongly recommend *not* running RTC in a
clustered environment.

That said, if you have two machines, with a shared disk between them,
and the Jazz/RTC server is installed onto that shared disk, that it
would probably be possible to create a relatively high available setup.
You can only run one server at a time, but if the primary server was
to stop working, then a second one could be started against the exact
same install (on the shared disk). It would be very important to be
sure that the two servers were never running at the same time, but it
should be possible to set something up to improve availability.

Matt Lavin
Jazz Server Team


zboruvka wrote:

Hi volks,
we need to deploy RTC in at least high availability mode or even
better load balancing as well

I wasn't able to find any useful documentation about that, so I
suppose following:
- both WAS and Tomcat can be deployed in cluster as wel as support
load balancing for HTTP communication (including REST)
- based on fact above RTC server running in WAS or Tomcat can be
clustered and used in load balancing mode

Is asumption above right?
Does anybody have any documentation about that (technical
papers/manuals)?
Do we have such deployments already?

I found only this note in online help: "The Jazz Team Server does
not utilize cluster support in WebSphere Application Server Network
Development and works only in stand-alone mode." but I belive
that we have some workaround..

Thanks a lot!
Zdenek

I don't know the exact time-frame for official 'clustering' support. I
know that supporting larger installations of RTC should be an important
goal of the 1.1 release.

Matt Lavin
Jazz Server Team


zboruvka wrote:

Thanks Matt!
Is cluster support planned for any of 2009 releases?
Zdenek

Hey Zdenek.

Clustering is not likely to be supported next year, but we do intend to deliver a solution for basic high availability. Because we're built on Java EE and relational databases, there are well-known techniques that should get us there. The intention is to support something like an "idle standby" as described for WebSphere Portal here: http://www-10.lotus.com/ldd/portalwiki.nsf/dx/exp-6.1-a-deploying-websphere-portal-express-using-idle-standby

Hope this is helpful, we're working on our own whitepaper like this.

Scott Rich
IBM Jazz Team

Hi!
If I've understood it correctly, you can install the RTC in a clustered environment and have two appservers running at the same time. But you must rely in a properly configured IP sprayer so only one instance should be processing requests at once.

In fact, we could have two possible scenarios using WAS. Both scenarios would have two servers with an active HTTPServer and appserver instance on each machine and an ip sprayer configured as primary/backup in front of the HTTPServers. We can use ND to manage both servers.
- In the first scenario both instances would be part of one WAS cluster and the jazz.war would be deployed to the cluster and to both HTTPServers. The HTTPServer plugin should be configured to use as primary the appserver instance of its own node. In this scenario we would have HA and all requests should be processed on only one instance at once. HA is given by the IP sprayer, the WAS HTTP Plugin and the WAS internal cluster mechanisms.
- In the second scenario, the instances would not be on a cluster and we would deploy the jazz.war two times. Each deploy would use a WAS instance and its correspondent HTTPServer. In this scenario the isolation is greater than the first scenario but HA is only provided by the IP sprayer.

Are both scenarios correct? Which one do you think is better? Do you see any drawback on them?

On the database side, we would use an Oracle RAC database so HA should be transparent to the RTC.

Thanks in advance!

You should not run RTC in a clustered environment. Not all of the code
in RTC is designed to be run across multiple servers right now. Only 1
RTC server should be connected to the RTC database at one time.

-
Matt Lavin
Jazz Server Team


On Mon, 2009-01-05 at 15:58 +0000, d00p wrote:

Hi!
If I've understood it correctly, you can install the RTC in a
clustered environment and have two appservers running at the same
time. But you must rely in a properly configured IP sprayer so only
one instance should be processing requests at once.

In fact, we could have two possible scenarios using WAS. Both
scenarios would have two servers with an active HTTPServer and
appserver instance on each machine and an ip sprayer configured as
primary/backup in front of the HTTPServers. We can use ND to manage
both servers.
- In the first scenario both instances would be part of one WAS
cluster and the jazz.war would be deployed to the cluster and to both
HTTPServers. The HTTPServer plugin should be configured to use as
primary the appserver instance of its own node. In this scenario we
would have HA and all requests should be processed on only one
instance at once. HA is given by the IP sprayer, the WAS HTTP Plugin
and the WAS internal cluster mechanisms.
- In the second scenario, the instances would not be on a cluster and
we would deploy the jazz.war two times. Each deploy would use a WAS
instance and its correspondent HTTPServer. In this scenario the
isolation is greater than the first scenario but HA is only provided
by the IP sprayer.

Are both scenarios correct? Which one do you think is better? Do you
see any drawback on them?

On the database side, we would use an Oracle RAC database so HA should
be transparent to the RTC.

Thanks in advance!