I already have two nodes one with CMS and other with just server processes. If have to add another CMS to Node 2 and cluster it with Node 1 CMS should I make any other changes like any web.xml or any changes to the server’s command line to bind to the cluster and not to CMS for fault tolerance. I read the Admin guide but there is no mention of editing the server processes like Input FRS or Adaptive Job Server or other processes to bind to the clustername instead of server.
All you have to do is update the web.xml cms.clusters area following the guidelines provided within the web.xml comments.
No other changes to the command lines are required as BusinessObjects does not recommend binding to a cluster name anymore, i.e. @
In the past, XI R2, we had to create entries in the windows hosts file on both the servers, i.e. on server 1, add machine name and IP address of server 2, and vice versa on server 2. I haven’t seen any such requirement in the deployment or amdin guide.
I will be clustering my XI 3.1 test environment next week and will let you know of my findings. Would appreciate if you do the same.
I still don’t get this - I have a cluster of 5 servers, and I have 2 boxes running CMS against the same database… I really have to update the web.xml to reflect this?
It seems to work fine with no changes, but am I not running in a redundant fashion? I can see users hitting the system from both CMSs in the sessions tab of CMC…
Just confused, looking for other experiences before I open a case and admit I don’ t really get it … : )
Have you tried shutting down one of the CMS services and see if the load gets shifted to the other? I would try that on both the CMS services one by one to test failover mechanism.
Brent - Any word from SAP on the steps necessary for the CMS clustering setup in XI 3.1?
I have setup a CMS cluster on two nodes in XI 3.1 and didn’t have to make any changes to the web.xml for the java application. I did not rename the cluster as @. I was able to test failover by bringing down both CMS one by one and still able to access the system.
Load balancing the application across different web servers is independent of CMS clustering. You would need a sclaing device such as Net Scaler to redirect traffic to different web servers for load balancing. We have separate web server machines that host nothing but Tomcat and separate intelligene tier machines that have CMSs installed on them. This setup has worked quite well in XI R2 and is holding up well in our migration test bed for XI 3.1
I’m planning a configuration for a client with 2 web app Servers and 2 CMS Servers.
I understand load balancing before/between the 2 web apps, but if a CMS server goes down how does the web app know to redirect traffic to the CMS Server for failover?
Apologies in advance as I’m not familiar with either IIS or tomcat, is it a case of having primary and secondary addresses registered, and/or if one server goes down would my client have to manually intervene and re-direct the web app to the CMS server thats still available?
if you follow the instructions about creating a clusters.txt file with your two CMS servers in that, your cluster will split the work among the two CMSs . if one goes down, the other CMS should take over its work.
Now, in practice, I’m not 100% sure it picks up existing sessions, but the cluster will continue to work normally (the cluster keeps an internal list of servers).
It should be in the admin or deployment guide (I forget)… Look for “cluster”.
The short version is you rename your cluster to something besides the machine name; we’re @BO_Production. Then, when you create CMSs on ‘nodes’, they all become part of this @cluster. The final part of closing that loop is to hit your appserver(s), and wherever you refer to the CMS by its original, machine name, you refer to it by the new @name. There has to be a clusters file with that @name and the individual CMS names / ports to tie that all together.
I’ve gone through the installation guides and things are a lot clearer on clustering, but I have a couple of really stupid questions … but I’ll ask it anyway,
Can I install BOE on a unix machine and the Web App (Tomcat) on a seperate windows machine?
If so, is it a simple case of installing the Web App from the windows product, and follow the windows documentation for the Web App?
I think that you should make the web.xml / cluster file changes, but I am not surprised that this works… I think BO is making registry changes that facilitate this failover.
I changed up my boxes somewhat, and until I made the web.xml changes to reflect my new CMS boxes and names, I had problems. Just a thought…
After renaming my cluster I started experiencing some problems so then re-visited this post.
What I actually discovered is that after installing an out of the box cluster, the InfoViewApp\WEB-INF\classes still actually refers to the specific machines CMS and not the cluster as I first thought.
In other words, for Infoview the load balancing of the CMS which I thought was being carried out by the BO Cluster was actually just by our load balancer in front of our application servers.
Would be nice if they mention this in the Admin guide.
It does reference the web.xml file for platformservices, which I found I needed to change to log into CMC, but frustratingly nothing at all about Infoview.
Is there a specific cluster document in publication by any chance?
There is a section in the Admin (install?) Guide on clustering - I got my info almost verbatim from there.
If you add your cluster name and then create a file with the details in it, you should be good… Just do that for all apps you want load balanced (CMC, Infoview, custom apps, etc).
(BOB member since 2005-08-27)
(BOB member since 2002-08-29)
(BOB member since 2002-09-13)
… but I’ll ask it anyway,
but I just don’t get how this is working? 
