With HA - i.e Oplog set to RF=2
1. A guest vm writes data on a Nutanix datastore
2. The data is transferred to a CVM on the same node.
3. The CVM/stargate process writes data to its oplog and ensures its persistence, and in parallel issues a write to the remote CVMs/stargate process oplog
for replication purpose.
3a.Local ACK from the oplog is received after persisting the data locally.
3b.The CVM waits for an ACK from the remote oplog.
4. After ACK is received by stargate from remote CVM,response is sent to the application.
At this point we have completed the IO from an application perspective.
Without HA - i.e Oplog set to RF=1
1. A guest vm writes data on a Nutanix datastore
2. The data is transferred to a CVM on the same node.
3. The CVM/stargate process writes data to its oplog and ensures its persistence.
3a.Local ACK from the oplog is received.
4 After the local ACK is received by stargate, the response is sent to the application
At this point we have completed the IO from an application perspective.
For the data that was written to Fusion IO we apply ILM function.
5. The CVM writes the data async to the backend storage with desired replication i.e RF=2
Without HA - i.e Oplog set to RF=1 and data set to RF=1
1. A guest vm writes data on a Nutanix datastore
2. The data is transferred to a CVM on the same node.
3. The CVM/stargate process writes data to its oplog and ensures its persistence.
3a.Local ACK from the oplog is received.
4 After the local ACK is received by stargate, the response is sent to the application
At this point we have completed the IO from an application perspective.
