diff options
Diffstat (limited to 'docs/clamp/acm/design-impl/clamp-runtime-acm.rst')
| -rw-r--r-- | docs/clamp/acm/design-impl/clamp-runtime-acm.rst | 54 |
1 files changed, 45 insertions, 9 deletions
diff --git a/docs/clamp/acm/design-impl/clamp-runtime-acm.rst b/docs/clamp/acm/design-impl/clamp-runtime-acm.rst index a3c22e69..802e9a0e 100644 --- a/docs/clamp/acm/design-impl/clamp-runtime-acm.rst +++ b/docs/clamp/acm/design-impl/clamp-runtime-acm.rst @@ -349,46 +349,61 @@ PARTICIPANT_REGISTER - A participant replica starts and send a PARTICIPANT_REGISTER message with participantId, replicaId and supported Element Types - ACM-runtime collects the message from Message Broker by ParticipantRegisterListener - if not present, it saves participant replica reference with status ON_LINE to DB +- it sends PARTICIPANT_REGISTER_ACK to participant replica PARTICIPANT_PRIME_ACK ++++++++++++++++++++++ - A participant sends PARTICIPANT_PRIME_ACK message in response to a PARTICIPANT_PRIME message - ParticipantPrimeAckListener collects the message from Message Broker -- It updates AC Definition to DB with PRIMED/DEPRIMED as status -- If AC Definition is fully PRIMED, ACM-runtime sends sync message to all participants replica +- It stores the message into the DB +- MessageIntercept intercepts that event and adds a task to handle a monitoring execution in SupervisionScanner +- Monitoring updates AC Definition to DB with PRIMED/DEPRIMED as status +- If AC Definition is fully PRIMED, Monitoring sends sync message to all participants replica PARTICIPANT_STATUS ++++++++++++++++++ -- A participant sends a scheduled PARTICIPANT_STATUS message with participantId, replicaId and supported Element Types +- A participant sends a scheduled PARTICIPANT_STATUS message with participantId, replicaId and supported Element Types. Same message could be used by participant to update OutProperties of an AC instance/AC definition. - ACM-runtime collects the message from Message Broker by ParticipantStatusListener -- if not present, it saves participant replica reference with status ON_LINE to DB +- If not present, it saves participant replica reference with status ON_LINE to DB +- If the message contains OutProperties of an AC instance/AC definition it stores the message into the DB +- MessageIntercept intercepts that event and adds a task to handle a monitoring execution in SupervisionScanner +- Monitoring updates the AC instance/AC definition +- Monitoring sends a sync message to all participants replica AUTOMATION_COMPOSITION_DEPLOY_ACK +++++++++++++++++++++++++++++++++ - A participant sends AUTOMATION_COMPOSITION_DEPLOY_ACK message in response to a AUTOMATION_COMPOSITION_DEPLOY message. It will send a AUTOMATION_COMPOSITION_DEPLOY_ACK - for each AC elements moved to the DEPLOYED state - AutomationCompositionUpdateAckListener collects the message from Message Broker -- It checks the status of all Automation Composition elements and checks if the Automation Composition is fully DEPLOYED -- It updates the AC to DB +- It store the message into the DB - MessageIntercept intercepts that event and adds a task to handle a monitoring execution in SupervisionScanner +- Monitoring checks the status of all Automation Composition elements and checks if the Automation Composition is fully DEPLOYED +- Monitoring updates the AC instance to DB +- If the Automation Composition is fully DEPLOYED Monitoring sends a sync message to all participants replica AUTOMATION_COMPOSITION_STATECHANGE_ACK ++++++++++++++++++++++++++++++++++++++ - A participant sends AUTOMATION_COMPOSITION_STATECHANGE_ACK message in response to a AUTOMATION_COMPOSITION_STATECHANGE message. It will send a AUTOMATION_COMPOSITION_DEPLOY_ACK - for each AC elements moved to the ordered state - AutomationCompositionStateChangeAckListener collects the message from Message Broker -- It checks the status of all Automation Composition elements and checks if the transition process of the Automation Composition is terminated -- It updates the AC to DB +- It store the message into the DB - MessageIntercept intercepts that event and adds a task to handle a monitoring execution in SupervisionScanner +- Monitoring checks the status of all Automation Composition elements and checks if the transition process of the Automation Composition is terminated +- Monitoring updates the AC instance to DB +- If the transition process is terminated, Monitoring sends a sync message to all participants replica Design of monitoring execution in SupervisionScanner **************************************************** Monitoring is designed to process the follow operations: +- to elaborate the messages from participants - to determine the next startPhase in a AUTOMATION_COMPOSITION_DEPLOY message +- to determine the next stage in a AUTOMATION_COMPOSITION_MIGRATION message - to update AC deployState: in a scenario that "AutomationComposition.deployState" is in a kind of transitional state (example DEPLOYING), if all - AC elements are moved properly to the specific state, the "AutomationComposition.deployState" will be updated to that and saved to DB - to update AC lockState: in a scenario that "AutomationComposition.lockState" is in a kind of transitional state (example LOCKING), if all - AC elements are moved properly to the specific state, the "AutomationComposition.lockState" will be updated to that and saved to DB +- to update AC subState: in a scenario that "AutomationComposition.subState" is in a kind of transitional state (example PREPARING), if all - AC elements are moved properly to NONE state, the "AutomationComposition.subState" will be updated to NONE and saved to DB - to delete AC Instance: in a scenario that "AutomationComposition.deployState" is in DELETING, if all - AC elements are moved properly to DELETED, the AC Instance will be deleted from DB - to retry AUTOMATION_COMPOSITION_DEPLOY/AUTOMATION_COMPOSITION_STATE_CHANGE messages. if there is an AC instance with startPhase completed, it will be moved to the next startPhase and retry a broadcast message with the new startPhase -- to send sync message to all participants replica: in scenario where AC instance transition is fully completed +- to retry AUTOMATION_COMPOSITION_MIGRATION messages. if there is an AC instance with stage completed, it will be moved to the next stage and retry a broadcast message with the new stage +- to send sync message to all participants replica: in scenario where AC instance transition is fully completed or OutProperties has been changed The solution Design timeout and reporting for all Participant message dialogues are implemented into the monitoring execution. @@ -396,6 +411,27 @@ The solution Design timeout and reporting for all Participant message dialogues - ThreadPoolExecutor executes the task - set AC instance stateChangeResult in timeout, if ACM-runtime do no receive Act message before MaxWaitMs milliseconds +Producer and Consumer and Scheduling job mechanisms ++++++++++++++++++++++++++++++++++++++++++++++++++++ +To avoid conflicts, for example when an ACM-r pod is receiving a messages to change outProperties of an AC instance and other ACM-r pod is receiving a messages to change the status of an element of the same instance, +Producer and Consumer and Scheduling job mechanisms has been implemented. +With the Producer and Consumer mechanism, any ACM-r replica monitoring could elaborate a message independently from what ACM-r replica has fetched the message. +Listeners of ACM-r will fetch messages from kafka and store to a message FIFO queue stored in DB. + +.. image:: ../images/acm-consumer-producer-messages.png + +Monitoring of a ACM-r replica will fetch messages from that queue. +All Monitoring from different ACM-r pods are synchronized with Scheduling job to avoid to fetch messages related to same AC instance or AC definition at same time. +InstanceIds and a compositionIds are generated with UUID as unique key and could be used as identificationId for messages. + +.. image:: ../images/acm-scheduling-job.png + +Edge case scenario: ACM-r pod has been terminated from Kubernetes +----------------------------------------------------------------- +Full elaboration of a monitoring for an AC instance or for a AC definition is protected by a transaction. +If an ACM-r pod has been terminated from Kubernetes, the monitoring not full completed will be rollback from database. +The monitoring job not completed wil be deleted after 200 seconds. All message of that AC definition or AC instance will processed by other monitoring. + Design of Exception handling **************************** GlobalControllerExceptionHandler |