Replication agent state should be persisted during scale up publish action

shinesolutions / aem-orchestrator

Java application for orchestrating AEM infrastructure created using aem-aws-stack-builder

Apache License 2.0

12 stars 9 forks source link

Replication agent state should be persisted during scale up publish action #33

Open cliffano opened 6 years ago

cliffano commented 6 years ago

On scale up publish action, the replication agent is created for the new pair, but it starts with an empty queue. This means we're losing the state of the replication agent on the originating publish instance.

Replication agent state should be treated just like the publish and publish-dispatcher pair's state, along with the publish instance's repository state (via EBS volume).

cliffano commented 6 years ago

First step to this effort is to investigate an AEM endpoint (or perhaps an internal API) which either:

copies a replication agent
copies queue state from one replication agent to another
retrieves queue from one replication agent, sets it on another

Once the API is identified, it should then be added to swagger-aem and published via aem4j.

cliffano commented 5 years ago

After further discussions, this ticket will require a capability from AEM side which should solve the problem with copying the state from a source replication agent to a destination replication agent. So this AEM feature will be looked at by @epari113 .

One that ticket is done, it will then have to be integrated into AEM OpenCloud first, and then the endpoints called by AEM Orchestrator.

cliffano commented 3 years ago

Clearing assignees since we're potentially going to switch to a new approach to require healthy publish instance candidate to have an empty replication agent. That way, creating an empty replication agent for the new publish instance during scale up action is the right thing to do.

I will update this ticket with confirmation whether we're definitely going to switch to that approach.