Service Desk Practitioners Forum
Showing results for 
Search instead for 
Did you mean: 

Loading clusters and virtual servers in the CMDB

James Paris
Acclaimed Contributor

Loading clusters and virtual servers in the CMDB

My company has a mixed server environment with stand-alone servers, Sun logical servers, Microsoft clusters, VMWare clusters, and VMware virtual servers.

Keeping track of what kind of server something is (stand alone, cluster, logical, virtual) is starting to be a real hassle...some tools, like SMS and CST can autodiscover some server attributes, but can end up confusing a virtual server with the box it is running on, or mix data about cluster nodes running load balancing for a single application.

Are other companies dealing with this challenge, and what kinds of models are being used to manage this info? We've tried a few models, but so far our server teams have been able to find one they can apply across multiple technologies.
Robert S. Falko
Honored Contributor

Re: Loading clusters and virtual servers in the CMDB


If you are using the CMDB as a database to describe the authorized state of your infrastructure (rather than the actual state of your infrastructure), then the update of the attributes of the CIs should be the result of a control process, rather than the result of a technical autodiscovery tool. If you do not take this approach, then you will have difficulties distinguishing between how a CI is configured, and how it ought to be configured. Furthermore, you will not be able to distinguish between authorized and unauthorized CIs. Finally, you have no means for detecting missing, lost or stolen CIs.

We take we this approach. All the VMWare virtual servers, Solaris domains, AIX lpars etc. are updated as a result of authorized changes that have been successfully completed. We do not try to use autodiscovery tools for this purpose. Clusters are not defined as CIs. Rather, the members of a cluster are related to each other via a specific relation type we created. Similar solution for load balancing. Note that we describe a relation between each application and each each server (logical or otherwise) on which they are installed.

Autodiscovery tools do have a role, however. They are useful for performing the verification and audit activity of configuration management (within the limits of what they are able to reliably detect).

James Paris
Acclaimed Contributor

Re: Loading clusters and virtual servers in the CMDB

Perhaps I dwelt too much on the role of autodiscovery tools. I've attached a jpg of our "Logical Server" model. There is two dashed lines, which are an "or" relationship: either a Logical Server is Hosted by a Physical Server, OR a Logical Server is Part of a Cluster.

I happen to like the model, but it is causing confusion. Some people want to have one CI for a physical server, and not track the logical servers, even though for one physical server there could be over a dozen VMWare servers running on it, and while we may know the name of an application's virtual server, we do not know what physical box it is on. We have frequent incidents of third shift call outs because a cluster node goes down, even though all apps on the cluster successfully failed over to other nodes, making the call out unnecessary. (The fact that these servers are part of the cluster, is not recorded, and the OVO messages do not reflect that the node failure is non-critical.)

Obviously, we have a lot of work to do in the relationship area, and one of the things that stymie the effort is a lack of a model we can either agree to or get management backing for. Thus the request for other models.
//Add this to "OnDomLoad" event