Data management method, node, and system for database cluster

A database cluster comprising two or more dual port solid state disks (SSDs), three or more nodes and a shared storage, wherein each dual port SSD is connected to a unique pair of nodes, each node in the unique pair of nodes is connected to a unique pair of the dual port SSDs, and each node comprises:

• a processor, a communications interface, and a memory capable of mutual communication via a bus, wherein the processor is configured to run executable program code stored in the memory, to:
• acquire a transaction log from a first node of the unique pair of nodes connected to the dual port SSD, wherein the transaction log stores data generated by a first database process at the first node after a checkpoint;
• sending the transaction log to a second of the unique pair of nodes when the first node crashes, wherein the second node
• a) operates a database process according to data obtained from the transaction log, wherein the obtained data is data stored in the first node before the first node crashes and the database process is the same as the first database process, or
• b) directly sends the transaction log to a third node that operates the database process instead of the second node.

A data management system comprising:

• two or more dual port solid state disks (SSDs), three or more nodes and a shared storage, wherein each dual port SSD is connected to a unique pair of nodes, each node in the unique pair of nodes is connected to a unique pair of the dual port SSDs and each of the dual port SSDs is configured to
• acquire a transaction log from a first node of the unique pair of nodes connected to the dual port SSD, wherein the transaction log stores data generated by a first database process at the first node after a checkpoint;
• and
• send the transaction log to a second node of the unique pair of nodes when the first node crashes, wherein the second node
• a) operates a database process according to data obtained from the transaction log, wherein the obtained data is data stored in the first node before the first node crashes and the database process is the same as the first database process, or
• b) directly sends the transaction log to a third node that operates the database process instead of the second node.