Reliable Event Driven Learner Service Architecture

Overview

Learner Service provides APIs related to resources such as Organisations, Users (i.e. learner), Course Batches, Locations, Badges, Pages etc.

For performance reasons Learner Services writes data to Cassandra which is synched in the background to Elasticsearch (for faster read and search).

Learner Service is currently architected to comprise of multiple actors. Broadly, these actors are categorised into foreground and background actors.

Foreground actors execute in the context of an API request to primarily either write information to Cassandra or read information from Elasticsearch. When information is updated in Cassandra a foreground actor sends a message to a background actor for performing sync of information from Cassandra to Elasticsearch. There are reliability concerns with current architecture which can result in information mismatch between Cassandra and Elasticsearch in case Learner Service container is restarted while one or more actors are working on some operations.

The purpose of this document is to propose a more reliable event driven learner service architecture.

Proposed Architecture

There are two types of events that Learner Service generates:

Transactional (generated due to DB update)
Instructional (e.g. manual sync, badge assertion, manual background job run)

Audit Logging

For reliable audit logging, events (on DB updates) should be generated immediately, in order and avoid duplication especially in a multi-node DB installation. Following is a summary of the available alternatives explored.

Approach	Pros	Cons
Cassandra Trigger	Immediately triggered Maintains order (not tested in multi-node scenario)	Deprecated (refer Datastax documentation) Semi-private and subject to change Performance overhead
Use audit logging of Apache Cassandra 4.0	Out of the box product feature which is expected to meet performance and reliability design goals	Not yet released
Change Data Capture	No performance overhead during write	Data not immediately available (requires memtables flush or commit log limit) Every node has a separate commit log (binary format) Consumer needs to handle duplicate events in a multi-node setup
ecAudit (Ericsson Cassandra Audit)	Immediately logged (on the node where the query is executed) Read is straightforward as log is in text format	Correlation required in case of FAILED events Every node has a separate audit log Performance overhead

Samples:

Cassandra Trigger

{{"columnnamename":"firstname", "value":"emp1"},{"columnnamename":"lastname","value":"last1"},"key":"1"}}

Cassandra 4.0 Audit Log

LogMessage: user:anonymous|host:127.0.0.1:7000|source:/127.0.0.1|port:53418|timestamp:1539980764310|type:SELECT|category:QUERY|ks:system_virtual_schema|scope:columns|operation:SELECT * from system_virtual_schema.columns ;

CDC (using a binary file reader)

{"rows":[{"cells":[{"name":"email","time":1552990511988147,"value":"some@gmail.com"},{"name":"firstname","time":1552990511988147,"value":"user101"},{"name":"lastname","time":1552990511988147,"value":"last101"},{"name":"phone","time":1552990511988147,"value":"9176009598"}],"clusteringKey":""}],"key":"101"}

ecAudit

19:53:00.644 - client:'127.0.0.1'|user:'cassandra'|status:'ATTEMPT'|operation:'CREATE ROLE ecuser WITH PASSWORD = '*****' AND LOGIN = true'
19:53:00.653 - client:'127.0.0.1'|user:'cassandra'|status:'FAILED'|operation:'CREATE ROLE ecuser WITH PASSWORD = '*****' AND LOGIN = true'

15:42:41.644 - client:'127.0.0.1'|user:'cassandra'|status:'ATTEMPT'|operation:'INSERT INTO ecks.ectbl (partk, clustk, value) VALUES (?, ?, ?)[1, '1', 'valid']'

Log Reader

A log reader such as logstash or fluentd can be configured for streaming audit logs to an event producer.

Event Producer

Event producer is responsible for constructing a JSON message for each log line and posting to an Event Queue topic.

Sample Transactional Events

Create User (with address and education details)

USER EVENT

{
    "identifier": "3d45fbd8-b911-4cc5-b503-61215902d780",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "userId": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "firstName": {
                "nv": "manzarul haque"
            },
            "email": {
                "nv": "manzarul07@gmail.com"
            },
            "emailVerified": {
                "nv": true
            },
            "phone": {
                "nv": "9888888888"
            },
            "phoneVerified": {
                "nv": true
            },
            "channel": {
                "nv": "ap"
            },
            "rootOrgId": {
                "nv": "123456789"
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional",
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user"
}

ADDRESS EVENT

{
    "identifier": "12345",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "12345"
            },
            "type": { // TODO
                "nv": "usr"
            },
            "userId": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "city": {
                "nv": "Bangalore"
            },
            "zipCode": {
                "nv": "560000"
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "address"
}

EDUCATION EVENT

{
    "identifier": "3d45fbd8-b911-4cc5-b503-61215902d780",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "6789"
            },
            "userId": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "degree": {
                "nv": "B.SC"
            },
            "yearOfPassing": {
                "nv": "2010"
            },
            "addressId": {
                "nv": "1111"
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user_education"
}

ADDRESS EVENT

{
    "identifier": "1111",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "1111"
            },
            "type": { // TODO
                "nv": "edu"
            },
            "userId": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "city": {
                "nv": "Bangalore"
            },
            "zipCode": {
                "nv": "560000"
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "address"
}

Update User

{
    "identifier": "3d45fbd8-b911-4cc5-b503-61215902d780",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "userId": {
                "nv": "3d45fbd8-b911-4cc5-b503-61215902d780"
            },
            "firstName": {
                "nv": "manzarul haque"
            },
            "profileSummary": {
                "nv": "Test to check for update"
            },
            "framework": {
                "id": {
                    "nv": "NCFCOPY"
                },
                "gradeLevel": {
                    "nv": [
                        "Kindergarten"
                    ]
                },
                "subject": {
                    "nv": [
                        "English"
                    ]
                },
                "medium": {
                    "nv": [
                        "Hindi"
                    ]
                },
                "board": {
                    "nv": [
                        "NCERT"
                    ]
                }
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional",
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user"
}

Add User to Org

{
    "identifier": "12345",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "12345"
            },
            "userId": {
                "nv": "877c28e7-5b8b-4586-a999-ff768d66562c"
            },
            "organisationId": {
                "nv": "0125345903028387842"
            },
            "roles": {
                "nv": [
                    "ORG_ADMIN"
                ]
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional",
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user_org"
}

Create Course Batch

{
    "identifier": "778899",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "courseId": {
                "nv": "do_112233"
            },
            "identifier": {
                "nv": "445566"
            },
            "batchId": {
                "nv": "778899"
            },
            "batchStartDate": {
                "nv": "2019-03-08"
            },
            "status": {
                "nv": 0
            }
        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "CourseBatch"
}

Add Participant to Course Batch

{
    "identifier": "1234",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "id": {
                "nv": "1234"
            },
            "batchId": {
                "nv": "0127015498299555847"
            },
            "userId": {
                "nv": "983c4665-fa2e-42a3-a236-e40a07833d55"
            },
            "courseId": {
                "nv": "do_1010"
            }

        }
    },
    "operationType": "UPSERT",
    "eventType": "transactional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user_courses"
}

Sample Instructional Events

Sync

{
    "identifier": "8a3c5eb4-7355-4e99-8802-8ba96a8c257c",
    "ets": 1548833319693,
    "event": {
        "properties": {
            "objectType": {
                "nv": "user"
            },
            "ids": {
                "nv": ["08cbc0b2-040d-4fc5-89b8-4a002778744b", "3a68c372-4d96-4133-99d1-de4491e86a38"]
            }
        }
    },
    "operationType": "sync",
    "eventType": "instructional"
    "userId": "ANONYMOUS",
    "createdOn": "2019-01-30T12:58:39.693+0530",
    "objectType": "user"
}

Event Queue

Available options for event queue.

Approach	Pros	Cons
Kafka	High throughput (100k msg / sec) Low latency Fault tolerant Durability Horizontal scale is easy	Message states not supported
RabbitMQ	Vertical scale is easy Message states supported (consumed, acknowledged, unacknowledged)	Low throughput (20k msg / sec)

Event Consumers

Following consumers are envisaged currently for Learner Service. More consumers can be added (if needed) in future as per requirement.

Consumer	Event Type	Description
De-Duplication	Transactional	Remove duplicate events arising from a multi-node DB installation with replication enabled
Indexer	Transactional	Generic consumer which can sync data for following resources. User Organisation Course Batch Location
Sync	Instructional	Generic consumer to manually trigger sync of one or more resources of a particular type (e.g. User, Organisation, Location etc.)
Badge Assertion	Instructional	Custom consumer to update assertion information in Badgr and LP (content) / DB (user).
Background Job	Instructional	Generic consumer to manually trigger a particular background job.

Indexer Mapping

{
    "user": {
        "index": "user",
        "key": "id",
        "visibilityEnabled": true
    },
    "user_education": {
        "index": "user",
        "path": "education",
        "key": "userid",
        "visibilityEnabled": true
    },
    "user_job_profile": {
        "index": "user",
        "path": "jobProfile",
        "key": "userid",
        "visibilityEnabled": true
    },
    "user_org": {
        "index": "user",
        "path": "organisations",
        "key": "userid",
        "visibilityEnabled": true
    },
    "address": {
        "types": {
            "usr": {
                "index": "user",
                "path": "address",
                "key": "userid",
                "visibilityEnabled": true
            },
            "edu": {
                "index": "user",
                "path": "education.address",
                "key": "userid",
                "visibilityEnabled": true
            },
            "job": {
                "index": "user",
                "path": "jobProfile.address",
                "key": "userid",
                "visibilityEnabled": true
            },
            "org": {
                "index": "org",
                "path": "address",
                "key": "orgid" // TODO
            },
        }
    },
    "organisation": {
        "index": "org",
        "key": "id"

    },
    "course_batch": {
        "index": "cbatch",
        "key": "id"
    },
    "user_courses": {
        "index": "usercourses",
        "key": "id"
    },
    "location": {
        "index": "location",
        "key": "id"
    },
    "visibility": {
        "index": {
            "user": {
                "table": "user",
                "column": "profilevisibility",
                "indexType": "profilevisibility",
                "key": "id"
            }
        }
    }
}

Action Points

1) Test Cassandra Triggers and ecAudit with different data types (e.g. list, map, blob etc.)

Please refer Cassandra Audit Logging for Different Types

2) Can Cassandra Trigger be applied per column?

No (refer Apache Cassandra Trigger documentation)

3) Performance comparison of Cassandra Triggers and ecAudit

Please refer Performance Comparison of Cassandra Audit Log Options.

4) Drop data in user_education, user_job_profile and address tables based on discussion with Rahul and Rayulu

5) Rename indexType in mapping to ____ ?

index

Appendix

Learner Service Data Models Comparison