Replica support in Participant

In the current implementation, ACM supports multi-participant with same supported element Type but different participantId, so they need different properties file.

In order to support replica, it needs to support multi-participant using same properties file.

Note: 

• In a scenario of high number of compositions, if participant is restarting it will be slow-down the restarting action: AC-runtime will send a message for each composition primed and instance deployed to the participant.
  To avoid the restarting action, participant needs a database support;
• In a scenario where a participant is stuck in deploying, the instance will be in TIMEOUT and the user can take action like deploy again or undeploy. In that scenario the intermediary-participant has to receive the next message, kill the thread that is stuck in deploying and create a new thread.
• In a scenario where we are increasing the number or participants, could be useful to have different topic name for source and sink. This solution will eliminate the number of useless messages in kafka.
  Example: 
  • for ACM-runtime:
    • sink: POLICY-ACM-PARTICIPANT
    • source: POLICY-ACM-RUNTIME
  • for participant: 
    • sink: POLICY-ACM-RUNTIME
    • source: POLICY-ACM-PARTICIPANT

Solutions

Solution 1: Replicas and Dynamic participantId - still using cache

Changes in Participant intermediary:

• UUID participantId will be generated in memory instead to fetch it in properties file.
• consumerGroup will be generated in memory instead to fetch it in properties file.

Changes in ACM-runtime:

• When participant go OFF_LINE:
  • if there are compositions connected to that participant, ACM-runtime will find other ON_LINE participant with same supported element type;
  • if other ON_LINE participant is present it will change the connection with all compositions and instance;
  • after that, it will execute restart for all compositions and instances to the ON_LINE participant.
• When receive a participant REGISTER:
  • it will check if there are compositions connected to a OFF_LINE participant with same supported element type;
  • if there are, it will change the connection with all compositions and instances to that new registered participant;
  • after that it will execute restart for all compositions and instances changed.
  • Refactor restarting scenario to apply the restarting only for compositions and instances in transition

Issues:

• Participants create randomly participantId and Kafka consumerGroup. This solution has been tested and has the issue to create a new Kafka queue in restarting scenario. 
  During restart scenario, a new consumerGroup is created, that cause some missing initial messages due the creation of new Kafka queue . The result is that to fail to receive messages from ACM to restore compositions and instances.

Solution 2: StatefulSets - still uses cache

Participant replicas can be a kubernetes StatefulSets that consume different properties file with unique consumer groups and unique UUIDs/replica (participants with same UUIDs have different replica number).

The StatefulSet uses the SPRING_CONFIG_NAME environment variable pointing to the spring application properties file unique to each of the participant replica.

Each of the properties file with the names pod-0.yaml, pod-1.yaml is mounted to the volumes. And the SPRING_CONFIG_NAME variable can be set to /path/to/$HOSTNAME.yaml to use the corresponding

properties file. 

By this approach the participant can have multiple replicas to work with a shared data.

env:
- name: HOSTNAME
  valueFrom:
    fieldRef:
      fieldPath: metadata.name

- name: SPRING_CONFIG_NAME
  value: /path/to/${HOSTNAME}.yaml

For example considering the http participant replica, ${HOSTNAME} will be "policy-http-ppnt-0" and "policy-http-ppnt-1" and their 
corresponding properties files with the names "http-ppnt-0.yaml" and "http-ppnt-1.yaml" is volume mounted. 

Note: In a scenario of two participants in replicas (we are calling "policy-http-ppnt-0" and "policy-http-ppnt-1"), ACM-Runtime will assignee as before any composition definition in prime time to specific participant based of supported element definition type. All participants will receive the same messages and store same data, so all participants are synchronized with ACM-R. Into all messages from ACM-R to participants will be present the replica number to indicate what participant will do the job (like prime or deploy). Example ACM-R send deploy message with replica 0, so "policy-http-ppnt-0" save the new instance and deploy and "policy-http-ppnt-1" just save that instance. When "policy-http-ppnt-0" send outProperties, then ACM-R and "policy-http-ppnt-1" receive the message and save that properties. When "policy-http-ppnt-0" has completed the deploy, the send the message and then then ACM-R and "policy-http-ppnt-1" receive the message and save the result. No issue if ACM-R send an undeploy message with replica 1, because all applications are synchronized.

Changes in Participant:

• Register, Status and Unregister message have to contain the replica number
• store data from messages from ACM-R if the participantId is matching and enable the actions only if the participantId and replica number are matching with the message
• store data (outProperties, and action completed) from messages from participants with same participantId and different replica
• implement time-out to stop the process if is running out of time

Changes in ACM-R:

• Save the replica numbers available
• Random replica into any message to participants

Changes in docker/Kubernetes environment

• implement StatefulSets 

Solution 3: Replicas and Database support - no cache

Changes in Participant intermediary:

• Redesign TimeOut scenario: Participant has the responsibility to stop the thread in execution after a specific time.
• Add client support for database (MariaDB or PostgreSQL).
• Add mock database for Unit Tests.
• Refactor CacheProvider to ParticipantProvider to support insert/update, intermediary-participant with transactions.
• Refactor Intermediary to use insert/update of ParticipantProvider.
• Refactor Participants that are using own HashMap in memory (Policy Participant saves policy and policy type in memory)

Changes in Participant:

• Add @EnableJpaRepositories and @EntityScan in Application:
• Application

  @SpringBootApplication
  @EnableJpaRepositories({
      "org.onap.policy.clamp.acm.participant.intermediary.persistence.repository"
  })
  @ComponentScan({
      "org.onap.policy.clamp.acm.participant.sim",
      "org.onap.policy.clamp.acm.participant.intermediary"
  })
  @EntityScan({
      "org.onap.policy.clamp.acm.participant.intermediary.persistence.concepts"
  })
  @ConfigurationPropertiesScan("org.onap.policy.clamp.acm.participant.sim.parameters")
  public class Application {
  
      public static void main(String[] args) {
          SpringApplication.run(Application.class, args);
      }
  }
  
  

• Add db connection in properties file and properties file for tests:

spring:
  security:
    user:
      name: participantUser
      password: zb!XztG34
  mvc:
    converters:
      preferred-json-mapper: gson
  datasource:
    url: jdbc:mariadb://${mariadb.host:localhost}:${mariadb.port:3306}/participantsim
    driverClassName: org.mariadb.jdbc.Driver
    username: policy
    password: P01icY
    hikari:
      connectionTimeout: 30000
      idleTimeout: 600000
      maxLifetime: 1800000
      maximumPoolSize: 10
  jpa:
    hibernate:
      ddl-auto: update
      naming:
        physical-strategy: org.hibernate.boot.model.naming.PhysicalNamingStrategyStandardImpl
        implicit-strategy: org.onap.policy.common.spring.utils.CustomImplicitNamingStrategy
    properties:
      hibernate:
        format_sql: true

spring:
  datasource:
    url: jdbc:h2:mem:testdb
    driverClassName: org.h2.Driver
    hikari:
      maxLifetime: 1800000
      maximumPoolSize: 3
  jpa:
    hibernate:
      ddl-auto: create
    open-in-view: false

• Unit Tests may need some changes

Changes in docker/Kubernetes environment

• Refactor CSIT to support database configuration for participants
• Refactor OOM to support database configuration for participants
• DB Migrator must be added to the helm chart and docker environments. The database schema of the older and newer versions will be different. Cache data added to the db.

Addition of DB Migrator

• Db migrator will alter old version of the db to add new parts of the schema required by this participant change
• Liquibase used for script generation
• Separate image needed for DB Migrator - this will have to be released as a new dependency
• New Job in kubernetes and new service in docker should be added for this migration

Advantages of DB use

• Multiple participant replicas possible - it can deal with messages across many participants
• All participants should have same group-id in kafka
• All should have the same participant-id.

Solution 4: Distributed Cache

Issues:

• Not persistent - if the application that handles cache server restarts - data is lost.
• Approval issues - with Redis, Etcd, Search Engine.

Solution 5: True Participant Replicas

The requirements are:

1. Participants can be replicated, each participant can have an arbitrary number of replicas
2. Composition definitions, instances, element instances and all their data including properties is identical in all participant replicas
3. When anything is changed in one replica, the change is propagated to all the replicas of a participant
4. An operation on a composition element can be sent to any replica of  a participant, which means that for a given element, the deploy could be on replica 1, the update could be on replica 2 and the delete could be on replica 3, as one would expect in any HA solution
5. A single REST operation called on ACM-R will select a participant replica (probably using round robin initially but we could support other algorithms in the future), and use that replica for that operation.
6. The ACM runtime will be made HA (more than one replica of ACM-R will be supported), it will run on a HA postgres.
7. The implementation of change propagation is transparent to participant API users
8. Replicas are "eventually consistent", with consistency typically occurring in 100s of milliseconds

This solution uses a similar approach to that used by Kubernetes when using etcd to implement CRDs and CRs. It implements replication in participants by introducing

1. The concept of participant replicas into ACM-R and the Participant Intermediary
2. A lightweight mechanism for replica update between replicas and ACM-R. Every time a replica changes its data, the change is reported to ACM-R and ACM-R updates all other replicas for that participant

The diagram above depicts data management as implemented today in ACM-R. ACM-R maintains the "source of truth" for all compositions and their elements in the ACM-R database.

1. Composition type data is pushed from ACM-R to the participants and is read-only in the participants
2. Composition element data (state information mainly) is built up by interactions between ACM-R and participants for the various ACM operations (Deploy/Update/Undeploy etc) and ACM-R always maintains the current composition element data in the ACM-R database
3. Composition element properties are pushed by ACM-R to the participant and can be updated in the participant by the participant implementation. When the properties are updated on the participant side, those changes are propagated to the ACM-R database

Therefore today, for the three types of data above, the ACM-R database has the state of the participant. This means that creation of participant replicas is rather trivial. We leave the current data handling mechanism in place, introduce the concept of participant replicas in ACM-R, and introduce data synchronization across participant replicas.

In this solution:

1. Participant Design backward compatibility is preserved, there is no change to the participant intermediary interface for participant implementations
2. Participant Configuration Backward compatibility is preserved, apart from a new "replicas" parameter (optional, default is 1), there is no change to participant configuration
3. ACM-R introduces a new REST API for replica management
4. ACM-R is made HA so that it itself can scale

Optimal Solution:

After analysis, it is clear that the best solution to use is number 3.

• Arbitrary number of participants possible
• DB migrator upgrades older versions
• Restart scenario not applicable anymore. Could be removed.
• Approval not an issue - postgres already used by acm.
• DB will be created automatically - as are required tables.

Older participant versions support (Regression)

• Do they have to upgrade to newest participant version? No, but if they want new functionality - they need to upgrade.