Skip to content

Latest commit

 

History

History
176 lines (128 loc) · 7.78 KB

candidate_relationships.md

File metadata and controls

176 lines (128 loc) · 7.78 KB
Raw

Candidate Relationships

Candidate Relationships serve as a valuable mechanism for creating complex connections between entities within and across different product verticals in the context of our Observability offering.

These relationships play a crucial role in enabling customers to quickly identify issues, resolve incidents, and correlate data across our various product offerings. While our current capabilities for creating relationships are powerful, they have some limitations, especially when dealing with entities where only limited data is available and the entity GUID cannot be determined directly from a single telemetry datapoint.

An illustrative example of these limitations is evident when relating AWS RDS databases. Establishing these relationships requires joining data from the Infrastructure agent, which connects to AWS and retrieves information about provisioned databases, with data from the New Relic agent, which identifies applications connected to databases using their connection URIs. By combining data from both agents, candidate relationships are formed, including the association between an APM application and a database.

In the relationship synthesis, the candidate is combined with the relationship synthesis rules. This integration significantly enhances the Observability experience for our users, offering a more comprehensive understanding of the interconnected entities.

In the next section, we explain the steps needed to define a new Candidate Relationship.

How to create a new Candidate Relationship

To create a new candidate relationship, follow these steps:

  1. Create a <candidate-type>.yml file under the relationships/candidates directory.
    1. In this directory, you'll find all existing candidates, so it's also an excellent source of inspiration for crafting new ones.
  2. Using the current definition of AWSS3BUCKET as an example, the candidate-type should match the filename (AWSS3BUCKET.yml) and the category, which is AWSS3BUCKET.
    1. A recommended practice is to use the entity type as the candidate type where possible.

Example:

category: AWSS3BUCKET
  1. Define one or more lookups for the candidate.
    1. These are rules that specify the entityTypes related to the category. They allow you to target entities of specific types during the lookup process.
    2. For instance, if you have an entity type called INFRA-HOST, and you define a lookup for it, the relationship synthesis process will inspect only entities of type INFRA-HOST to find potential matches for the candidate relationship. This focused approach enhances the precision and efficiency of the relationship synthesis.

Example:

category: AWSS3BUCKET
lookups:
  - entityTypes:
    tags:
    onMatch:
    onMiss:
  1. Fill in the required fields for each lookup. For the AWSS3BUCKET example, the fields are entityTypes, tags, onMatch and onMiss.

Example:

category: AWSS3BUCKET
lookups:
   - entityTypes:
        - domain: INFRA
          type: AWSS3BUCKET
     tags:
        matchingMode: ANY
        predicates:
           - tagKeys: ["aws.bucketName", "aws.s3.BucketName"]
             field: bucketName
     onMatch:
        onMultipleMatches: RELATE_ALL
     onMiss:
        action: CREATE_UNINSTRUMENTED
        uninstrumented:
           type: AWSS3BUCKET

The next section provides detailed explanations of the configurable elements for a candidate relationship definition.

  1. Create a new pull request and make sure all the automatic validations are successfully executed.

  2. Wait for our team to review the pull request and iterate on the feedback.

  3. Once approved, merged to the main branch and a new release is issued, it is available for you to use.

  4. Ensure you also have a matching relationship synthesis rule that utilizes the provided candidate.

  5. Enjoy your new candidate relationship! 🎉

How to configure a new Candidate Relationship

The following table lists the available fields for configuring a new candidate relationship and specifies if they are mandatory or not.

Name Type Required Description
category String Yes Represents the generic concept of the entity category.
lookups List Yes Provides a list of rules for the defined category.
tags Composed Yes Specifies which of the entity tag keys must match the input fields.
onMatch Composed Yes Specifies actions to be executed in case the rule finds a match.
onMiss Composed Yes Specifies actions to be executed in case the rule finds a miss.

Category

The category represents the generic concept of the entity. It is usually more generic than the entityType, as it can encompass multiple types. Think of it as a generic interface.

The category is also used as the uninstrumented type when creating a new entity.

Lookups

Lookups define the rules for the category and specify the entity types related to it. These rules are used during the lookup process to inspect only entities on the specified entity types, allowing for a targeted search for a specific entity type.

Tags

Tags define which entity tag keys must match the input fields. It is possible to define different matching modes.

The tags field is a composed object with the following subfields:

Name Type Required Description
mode String Yes Defines the matching behavior for the specified tags.
predicates Composed Yes Defines the tagKeys from the entities to match with the field value.

(Tags) Mode

  • ANY: All entities that match at least one of the defined predicates will be considered for a relationship.
  • ALL: Entities must match all the specified predicates to be considered for a relationship.
  • FIRST: The first group of entities that match the top-most predicate will be considered for a relationship.

Example:

     tags:
        matchingMode: ANY
        predicates:
           - tagKeys: ["aws.bucketName", "aws.s3.BucketName"]
             field: bucketName

In this case, entities with the tag keys aws.bucketName or aws.s3.BucketName will be matched (using their tag values) with data from the bucketName field.

onMatch

The onMatch section controls the behavior after matching entities. The onMultipleMatches field specifies how to handle multiple matches:

  • RELATE_ALL: Relates the known GUID with all the matched entities.
  • DISCARD_ALL: Assumes a conflict and discards all matches.

Example:

    onMatch:
     onMultipleMatches: RELATE_ALL

onMiss

If the system fails to match with any entity, an uninstrumented entity could be created.

The uninstrumented entity indicates that additional instrumentation can be configured to gain more visibility.

There are two options for handling onMiss:

  • NO_OP: Discards misses instead of generating an uninstrumented entity.
  • CREATE_UNINSTRUMENTED: Generates an uninstrumented entity. When using CREATE_UNINSTRUMENTED, you need to provide the type used for the created entities.

Example:

    onMiss:
      action: CREATE_UNINSTRUMENTED
      uninstrumented:
        type: AWSS3BUCKET # The final domain-type will be UNINSTRUMENTED-AWSS3BUCKET