Different versions of the CMDB have subtly different data models. Syncmappings can support multiple data models with datamodel declarations. CMDB data models are assigned simple integer values:
The mappings performed during CMDB synchronization are specified in syncmapping blocks in TPL. A syncmapping is similar to a pattern, but it triggers from a queued synchronization action, rather than from data being updated in the data store.
The form of a syncmapping is:
syncmapping name version
description
overview
overview_entries
end overview;
[constants
constant_definitions
end constants;]
mapping mapping_source as source_name
mapping_definitions
end mapping;
body
body_details
end body;
end syncmapping;
As with pattern blocks, the name, version, and description are mandatory.
Pattern templates are provided to help you create your own syncmappings.
The overview is required. It contains information about the pattern and the entities it creates. It must contain a tags entry, and can have an optional datamodel entry, as described in Data models.
The overview section can also have an optional overrides entry, which enables you to change the behavior of an existing mapping and to preserve those changes on subsequent TKU updates to the overridden mapping. Non-syncmapping patterns have an "overridden" flag displayed on the pattern management page, syncmappings do not.
Here is an excerpt from an overriding mapping:
...
from CMDB.Host_ComputerSystem *import* Host_ComputerSystem 1.2;
from CMDB.SoftwareInstance_SoftwareServer *import* SoftwareInstance_SoftwareServer 3.2;
// The mapping referred to by the overrides keyword must be imported
syncmapping SoftwareInstance_SoftwareServer_override 3.2
"""
Example override mapping.
"""
overview
tags CMDB, Core_Mapping;
datamodel 0, 1, 2, 3, 4, 5, 6;
overrides SoftwareInstance_SoftwareServer; // The overridden pattern is imported above
end overview;
mapping from Host_ComputerSystem.host as host
traverse Host:HostedSoftware::SoftwareInstance as si_node
traverse Element:Maintainer:Pattern:Pattern *as* si_pattern
end traverse;
...
Note
The mapping referred to by the overrides keyword must be imported. All definitions from the original mapping section such as node kinds, traversals and names must be preserved. It is possible to add additional traversals and names in the overriding mapping.
The mapping section declares the starting point for the mapping, the structure of source data retrieved from the Discovery model, and the target CIs created in the CMDB model. It does not describe how the source data is transformed to the target model — that is performed in the body section.
Each mapping is either a root mapping, meaning that it is invoked by the synchronization of a single root node with the corresponding kind, or an extension mapping, meaning that it extends another mapping at a suitable point.
You should not use DDD directly in mappings, rather, you should copy the required DDD into the inferred model first.
Root mappings have a mapping declaration by using the on keyword:
mapping on node_kind as name
mapping content...
end mapping
For example, this specifies the root mapping for Host nodes:
mapping on Host as host_node
Extension mappings have a mapping declaration by using the from keyword:
mapping from source_scoped_name as name
For example:
mapping from ExampleMapping.host_node as host_node
The source_scoped_name is the name of a source mapping variable from another mapping block, either the source name specified in the mapping declaration, or a traversal name as described below.
Multiple parent mappings are supported, by using more than one "From" expression in the mapping block, for example:
SoftwareInstance_SoftwareServer 4.0
...
mapping from Host_ComputerSystem.host from Cluster.cluster as hosting_system
...
In this example the mapping input can be from a Host or a Cluster node in the (BMC Discovery model). All of the input parameters used in the mapping statement must have corresponding references in the body. An example when the above mapping statement is:
hosting_ci := Cluster.cluster_ci or Host_ComputerSystem.computersystem;
The hosting system (in the CMDB model) is either a BMC_Cluster or a BMC_ComputerSystem. The order matters here; the interpreter takes the first defined object which it finds (checking from left to right).
The source subgraph is declared by using traverse clauses inside the mapping with a syntax similar to traversals in search expressions:
traverse traversal_specification as nodename
traversal contents...
endtraverse
The nodename defined by the traversal can only be used in a for each expression; it cannot be used in any other context.
You can also define a relationship relname for a traverse clause inside the mapping. The first is the name of the relationship, and the second is the name of the node.
traverse traversal_specification as relname, nodename
traversal contents...
end traverse
In each case the relname and nodename defined by the traversal can only be used in a for each expression; they cannot be used in any other context. You can access the attributes on a relationship by using the name you assigned to the relationship, followed by a dot and the attribute as specified in the syncmapping.
The initial source node and the results of traversals can be filtered with where clauses, specified before the as token. where clauses in mapping blocks use the same subset of search where clauses as trigger conditions in pattern blocks.
mapping on node_kind where condition as name
mapping from source_scoped_name where condition as name
traverse traversal_specification where condition as name
However, you cannot use search in a syncmapping. CMDB synchronization takes a graph of connected nodes in BMC Discovery and transforms them into a graph of connected CIs in the CMDB. If the where clause contains a search there is no guarantee of any such connection.
As the subgraph is processed in the body, target CIs are specified. The mapping block contains declarations of the CIs that are mapped, in the form:
name -> CI_class;
For example:
computersystem -> BMC_ComputerSystem;
Targets are specified within the traversal structure. For example, part of the mapping of virtual machines is as follows:
mapping from Host_ComputerSystem.host where virtual defined as host
traverse ContainedHost:HostContainment:HostContainer:SoftwareInstance as vm_si
vse -> BMC_VirtualSystemEnabler;
traverse RunningSoftware:HostedSoftware:Host:Host as containing_host
containing_cs -> BMC_ComputerSystem;
*end* *traverse*;
*end* *traverse*;
end mapping;
In some circumstances, a number of nodes in the Discovery model must be grouped together to construct a single CI in the CMDB model. This is declared in the mapping with a group block. The form of a group block is
traverse traversal_specification as traversal_name
group group_name
group contents...
expand group as expansion_name
expansion contents...
end* *expand*;
end* *group*;
end* *traverse*;
The declaration indicates that nodes from the containing traversal will be grouped together (according to rules specified in a group block in the body), and then the group will be expanded to the individual group members. The expand is not required if there is no need to process the individual nodes within the group.
Syncmapping body
The body of a syncmapping is responsible for implementing the mapping described in the mapping block. The majority of language features and functions available in pattern body blocks are permitted, except that functions in the following namespaces are not available since they are only appropriate for patterns that perform discovery and construct the Discovery model.
- discovery
- inference
- mail
- model
Additionally, user-defined functions are not supported.
NoteAny date/time fields should be passed as datetime objects rather than formatted datetime strings.
Body execution
The body of a syncmapping is executed at a time that depends upon the mapping source definition.
The body of a root mapping (specified with mapping on) is executed at the time the root node is scheduled for synchronization.
The body of an extension mapping (specified with mapping from) that extends the source node of another mapping is executed when the body of the extended mapping completes.
The body of an extension mapping that extends a traversed-to node of another mapping is executed each time the associated for each loop (see below) completes.
When a node in the Discovery data store is marked as destroyed, only the root mapping's body is executed, and the target root CI is scheduled for deletion in the CMDB. When the delete is synchronized with the CMDB, the root CI and all the related CIs previously created by the mapping are deleted. For best performance during deletion, root mappings should not perform any traversals or other time-consuming activities.
CIs and relationships in the CMDB are specified with functions in the sync namespace, similar to those in the model namespace used within pattern blocks.
Any CMDB class can be specified with a function call of the form
value := sync._BMC_ClassName_(key := _some_key_, _attributes..._);
Any class name can be specified. Specifying a class that is not defined in the CMDB results in a runtime error. The key attribute must be set, and is used to populate the ADDMIntegrationId attribute in the CMDB. Any other attribute name can be set; attributes that are not defined in the CI class are ignored.
The result of the function must be assigned to a target CI name specified in the mapping block. The class specified in the function must be the same as the one specified in the mapping or a subclass of it.
CMDB classes are assumed to be in the BMC.CORE namespace. To refer to a class in a different namespace, provide a namespace parameter to the function call:
value := sync._BMC_ClassName_(key := _some_key_, namespace := "My.NameSpace", _attributes..._);
The namespace must be a literal string — it cannot be constructed at runtime.
The subgraph of data in the Discovery model is transformed into a subgraph of CIs in the target CMDB model. Most of the CIs belong to a single target subgraph, but some are shared by more than one subgraph. An example is the BMC_IPConnectivitySubnet CI that is shared by all the computers on a particular subnet. For deletion to work correctly, the system must know that such CIs are shared. This is achieved by calling the function in the sync.shared namespace:
value := sync.shared._BMC_SharedClassName_(key := _some_key_, _attributes..._);
Similarly, it is sometimes necessary to specify a relationship to a CI that is not part of the target subgraph. An example is to relate the BMC_ComputerSystem for a physical host to the one for a virtual host — the two CIs belong to different subgraphs. External CIs are specified with a function in the sync.external namespace:
value := sync.external._BMC_ClassName_(key := _some_key_ _[, namespace := "_NAMESPACE_" ]);
The key must be specified, and namespace must be specified if required. No other attributes can be set.
It is not an error if a CI with the specified key does not exist in the CMDB. In that situation, the CI and any relationships to it are simply ignored.
As the mapping is processed, the CIs are specified in a tree traversal across the graph. To refer to a CI specified in a different branch of the tree, it can be specified with a function in the sync.crossref namespace:
value := sync.crossref._BMC_ClassName_(key := _some_key_ _[, namespace := "_NAMESPACE_" ]);
The key must be specified, and namespace must be specified if required. No other attributes can be set.
It is a runtime error to specify a cross-reference to a CI that is not fully specified elsewhere within the mapping.
TokenId is an attribute that, in some circumstances, aids the reconciliation of CIs populated by multiple data sources. BMC Discovery sets TokenId depending on the Node kind.
For most Nodes, TokenId is one of the forms:
key
or
ADDM:%key%
or
ADDM:%hashed_key%
For databases, TokenId is a database host found as a result of all possible combinations of where the database might be running:
%hosting_ci.TokenId%:%database_node.type%:%si_instance%:%database_node.instance%
and
%hosting_ci.Name%:%database_node.type%:%si_instance%:%database_node.instance%
For the Fibre Channel Node, TokenId is of the form:
WWN:%wwpn%
For physical hosts, TokenId is of the form:
%host.hostname%:%host.dns_domain%
If the domain name is unavailable, then
%host.hostname%
For virtual hosts that are running in the cloud, TokenId contains instance_id.
- For Alibaba, TokenId is of the form:
%alibaba_instance_id%
- For Azure, TokenId is of the form:
%azure_vm_id%
- For AWS, TokenId is of the form:
%aws_instance_id%
- For GCP, TokenId is of the form:
%gce_instance_id%
- For IBM, TokenId is of the form:
%instance_id%
- For OCI, TokenId is of the form:
%ocid%
For virtual hosts that are running in the hypervisor, TokenId contains UUID.
- For AHV, TokenId is of the form:
AHV-UUID:%uuid%
- For Hyper-V, the UUID is only available on the physical machine, so TokenId is only set for virtual machines that have been successfully linked to their hosting physical machines:
HYPERV-ID:%vm_uuid%
- For KVM (including RedHat Enterprise Virtualization), TokenId is of the form:
KVM-ID:%uuid%
- For VMware, TokenId is of the form where each letter represents a hexadecimal digit:
VI-UUID:%uuid%
- For Xen (including Oracle VM), TokenId is of the form:
XEN-ID:%uuid%
The following table shows how TokenId looks like for different Node Kinds:
| | |
|---|
| | |
| | |
| BMC_LogicalSystemComponent | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | %hosting_ci.TokenId%:%database_node.type%:%si_instance%:%database_node.instance% |
%hosting_ci.Name%:%database_node.type%:%si_instance%:%database_node.instance% |
| BMC_HardwareSystemComponent | |
| | |
| | |
| | %host.hostname%:%host.dns_domain% |
|
ALIBABA-ID:%alibaba_instance_id% |
|
|
|
|
|
|
|
|
|
|
| | |
| | |
| | |
| | |
| BMC_LogicalSystemComponent | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| BMC_HardwareSystemComponent | |
Relationships are specified with functions in the sync.rel namespace:
sync.rel._BMC_RelName_(Source := source_val,
Destination := dest_val,
Name := "RELNAME" [, ... ]);
The first two parameters must be Source and Destination. Any other attributes can also be set; Name is not required, but it is conventionally always set.
The Impacted and ImpactWeight attributes can be used to create impact relationships. They specify the impact direction and percentage, respectively. For example, an ImpactWeight of 25 might be appropriate when representing an impact relationship between a BMC_ComputerSystem and BMC_Cluster, where there are four computers in a cluster. The value assigned to Impacted must be the Source or Destination CI appearing in the same definition.
sync.rel.BMC_Component(
Source := computersystem,
Destination := cluster_ci,
Name := "CLUSTEREDSYSTEM",
Impacted := cluster_ci,
ImpactWeight := cluster_rel.impact_weight
);
NotePrior to TPL 1.10, multiple relationship definitions were needed to represent a single impact relationship in the CMDB. This was because the way in which impact is represented in the CMDB depends on the data model in effect (see below). In tpl 1.10 and later, only one definition is required.
One of the main activities performed in the body is to iterate over the nodes reached through the traversals specified in the mapping block. A for each loop is used to iterate over the named nodes:
for each source_node do
...
end for
The nesting structure of for each loops in the body must match the nesting structure of the traverse expressions in the mapping block.
A for each loop is required even if the corresponding traversal is expected to reach just one node. There is no other way to access the state of the traversed-to node.
When the mapping block specifies a group, there must be a corresponding group block in the body. The group block will always be inside a for each block, either directly within a single syncmapping body or in an extended source syncmapping.
A group block takes the form:
for each traversed_to_node do
...
ident := group_identifier;
group group_name with ident do
...
for each expand_name do
...
end for;
end group;
end for;
The grouping is evaluated in two phases. In the first phase, every iteration of the surrounding for each loop is executed. The nodes are grouped according to the identifier provided to the group expression. After all the iterations of the for each loop, the group block is executed once for each group. If the group declaration in the mapping block contains an expand declaration, there should be a corresponding for each loop in the body.
The group content is executed in a context based on an arbitrary member of the group. Any local variables from the surrounding loop will therefore be valid for a member of the group, but there is no guarantee that it will be the same group member each time a particular group is processed.
Different versions of the CMDB have subtly different data models. Syncmappings can support multiple data models with datamodel declarations. CMDB data models are assigned simple integer values:
| | |
|---|
| | HasImpact and ImpactDirection attributes are set as appropriate. |
| | Only to be used with legacy SIM version 7.4. BMC_Impact relationships with Name “ImpactOnly” are created. |
| | No impact details are set by BMC Discovery. They may be set by Impact Normalization in the CMDB. |
| | BMC_Impact relationships with name “IMPACT” are created. |
| | BMC_Impact relationships with name “IMPACT” are created. |
A syncmapping can limit itself to a particular set of data models with a datamodel declaration in the overview:
overview
tags Some_tags;
*datamodel* 3, 4; Only CMDB 7.6.x
end overview
The body of a syncmapping can further modify its behavior for different data models with a datamodel block. The datamodel block only executes if the data model in effect matches the declaration:
ci := sync.BMC_Thing(key := my_key, ...);
datamodel 2, 3 do
sync.rel.BMC_Dependency(
Source := ci,
Destination := other_ci,
Name := "DEPENDENCY_NAME"
);
end datamodel*;
The data model, in effect, is not chosen automatically. After you configure CMDB synchronization, the data model is selected. This is described in Setting up a CMDB synchronization connection.