Storage Container Manager operations in High Availability
When an Ozone cluster has Storage Container Manager (SCM) High Availability (HA) configured, the important SCM operations; for example, managing client requests such as allocating containers, local operations such as destroying pipelines, and processing DataNode updates, are handled differently from a non-HA Ozone cluster.
Client request management
SCM clients are the different Ozone elements that interact with the SCM such as DataNodes, the Ozone Manager (OM) and so on.
On receiving client requests such as allocating a container or a pipeline, the leader SCM performs all the required operations. The leader also performs metadata changes that result from executing the client request, and accordingly updates Ratis. The changes are replicated to the followers through Ratis.
Performing operations local to the SCM
SCM performs local operations such as destroying pipelines, deleting stale DataNodes, and so on, when it stops receiving heartbeats or reports from DataNodes. The followers log the occurrences of these operations and their results. The leader performs any metadata changes that result from these local operations, and accordingly updates Ratis. The changes are replicated to the followers through Ratis.