[Date Prev] [Date Next] [Prev in Thread] [Next in Thread] [Date Index] [Thread Index]

Re: CMDB: Define "Clusters" of Servers?

** Hi Ray;

Interesting request - and really what you are trying to is a pure CMDB-sim as I say...

I (personally) don't like the term "Cluster" as I immediately think about physical clusters of systems such as fail-over, etc especially for this topic, because your terms are better suited as:

Environment (Development / QA / Production)
Application (Peoplesoft / Oracle)

So if you create a CI (Choose something like "Document" - [Never figured a good use for this anyway :) ] ) and Create something like


Then you can relate Oracle-->Oracle10 and Oracle-->Oracle11
Then you can relate Oracle10 --> Computer_systems / Oracle11 --> Computer_systems

Make the "Relationship Type" "Member of Collection"

However I would also suggest that you can also setup "Applications" in the CMDB just as well to document things from the "application container"...

As far as the "environment" goes I'd use the CI Field "System Role" to capture the environment...

As far as scripting - since they are just standard tables / views just the standard API (Java / Perl / etc) can be utilized to query the tables appropriately and return the data required.


On Mon, Aug 2, 2010 at 11:22 AM, Ray T <cool.develop@gmail.com> wrote:
** We are talking about CMDB (ITSm 7.5) and there is a question of how well we can define "clusters" of Unix servers in CMDB.

The background is that: currently Unix servers have labels assigned to them. For example, an Oracle server box may belong to clusters "oracle", "oracle11g", "production" and another one may belong to "oracle", "oracle10g", "test", "peoplesoft". So clusters are labels that allow us to group servers together so we can view, talk about and manage them in groups. The cluster labels are defined as needed and the mapping of servers to cluster names is in a text file.

There are little utilities that we use to query for this data from the text file. For example, a shell/Perl script can use a clusterquery utility to find out what servers (hostnames) belong to "oracle" cluster, so that it can take actions on those servers. Or I can query to see what hosts are oracle hosts, but not 11g. Or I can list all the clusters a host belongs to. Patch management is one area cluster info is used heavily in.

The question is, if we have the servers as CIs in CMDB, (1) how do we define, represent cluster info (it would be one to many), and (2) how much work is it to write comparable command line tools that we can distribute to folks so that their scripts continue to work (say above 3 queries)? I would think CMDB APIs would be involved. Perl is the language of choice here, although we can also do C.

Has anybody tracked such "cluster" info in CMDB? Is it common to manage such data in CMDB, and to consume data through command line tools like we are thinking?


_attend WWRUG10 www.wwrug.com ARSlist: "Where the Answers Are"_


_attend WWRUG10 www.wwrug.com ARSlist: "Where the Answers Are"_