Ray,

There are three types of reconciliation

Reconciliation Identity -- where you are looking for CIs/relationships in 
different datasets that are the same object.  You control and specify what 
fields are compared in the reconciliation identity rules

Reconciliation merge -- where you are combining different CIs/relationships 
from different datasets into a single CI.  The ONLY field that is used in this 
operation to select CIs and relationships is the reconciliation ID.  Entities 
with matching reconciliation IDs are merged according to your precedence rules.

Reconciliation compare -- where you are comparing two datasets to look for 
differences between CIs and relationships.  The ONLY field that is used in this 
operation to select CIs and relationships is the reconciliation ID.  Entities 
with matching reconciliation IDs are compared.


I suspect you are talking about the Merge operation.  So, the only field used 
is the reconciliation ID.  That means that only CIs and relationships that have 
been through the reconciliation identify process will have reconciliation IDs 
to be able to be merged.

Since data is merging and you are getting results, there must be reconciliation 
IDs present.  Make sure that those rules are properly linking the common items 
and that the identity rules are not giving the same item in different datasets 
different IDs -- that would give you duplicates like you indicate may be 
happening when you say a SECOND record is created.  The problem here would be 
that the reconciliation identity rules may not be complete enough and you need 
to review and maybe revise those rules.


If you have data sources that are sparse, the attributes don't default, they 
are just NULL.  If the field is a required field, it cannot be NULL.  You would 
either need to allow it to be NULL by not being required or supply a value.

There is a configuration option that says to "ignore null values" in the 
precedence definition.  That would mean that if dataset A and B both had a copy 
of the CI but you defined A as the winner in precedence, values of attributes 
from A would win.  BUT, if an attribute was NULL, this setting says to ignore 
it and move to the next in line and get the value from the next item (and so on 
until it gets a non-null value).

You can set the precedence at the class level or at the attribute level.  So, 
you can different attributes win from different datasets or you can have the 
entire set win from one dataset over another.

From what you are describing, you are not using the "ignore null values" (note 
that I don't remember if it is called exactly that, but it is close to that).  
That would fix the overwrite of data with NULL values.  You may have a load 
issue with defaults flowing in rather than NULL values for missing data.  
Missing data should be NULL in the CMDB not defaulted.  And, you may have a 
reconciliation identify problem where the rules there -- and you pick what 
fields are tested for matching because you specify the search criteria for 
determining the match of CIs and relationships.  That would require a review of 
the rules there and fixing them as needed so that the same object in multiple 
discovery datasets get the same reconciliation ID so that they are merged 
and/or compared correctly.


Then, there is one final area around Normalization.  You may get slightly 
different spellings or flavors or whatever of values.  Normalization allows you 
to say that if you get any of the flavors, assign a single common value -- like 
if company is Microsoft, Microsoft Corp, MSFT, ..  then set the field to 
Microsoft) so that you get consistent data.  This is run BEFORE any of the 
reconciliation steps above so that when you try and identify, you don't 
mismatch on different flavors of the same thing.  This is all about data 
cleanliness.  One more thing to consider if the identify is not linking items 
due to this type of data inconsistency.


I hope this has given you (and others) some understanding of the process and 
some ideas to look at.  And, maybe an option or two that you may not have been 
aware of that may help your situation.

Doug Mueller

-----Original Message-----
From: Action Request System discussion list(ARSList) 
[mailto:[email protected]] On Behalf Of Ray Gellenbeck
Sent: Thursday, January 22, 2015 10:54 AM
To: [email protected]
Subject: CMDB Reconciliation - Dataset Precedence & Comparisons

Caveat:  yes, I know how to set/edit Dataset precedence in the Reconciliation 
tool.

My question is a bit more general:  is there a published list of what 
attributes are compared during reconciliation?

Client has multiple source data sets.  Sometimes a subset of records in one of 
the data sets is not as complete as desired (long story, don't go there).

They will then run a manual CSV of those records and import them into a new 
dataset and set it higher precedence than the original so that matches go into 
Asset and over-write the less-complete set's records.

The problem is that they use spoon and leave a lot of fields default, resulting 
in items like Type or Item not exactly matching.  You know what happens 
then...instead of over-writing the Asset record, a SECOND record is created in 
Asset.  No bueno.

So the customer came back with the question "ok, so where is the list of all 
the attributes that are compared so I can ensure we have exact matches or we 
are filling in the blanks?"  Short of manually setting precendence rules for 
each attribute, I don't have the answer myself.

Anybody got a handy "default" list for BMC.Core or BMC.ComputerSystem 
attributes compared?  I'm sure I"m overlooking something simple...blame it on 
the flu...and thanks in advance.

_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org "Where the Answers 
Are, and have been for 20 years"

_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org
"Where the Answers Are, and have been for 20 years"

Reply via email to