Method, device, and system for peer-to-peer data replication and method, device, and system for master node switching

A method for peer-to-peer data replication, the method comprising:

• receiving, by a subscriber node, a commit redo record attached with a commit sequence number (CSN) from a master node, wherein the CSN attached in the commit redo record is used for indicating the sequence of a write operation within the master node;
• updating a CSN of the subscriber node to the CSN attached to the commit redo record, when an absolute value of the CSN attached to the commit redo record minus the CSN of the subscriber node is 1, wherein the CSN of the subscriber node is set to 0 during an initialization;
• replaying, by the subscriber node, the commit redo record to replicate data of a write operation happened on a master node from the master node to the subscriber node, wherein the subscriber node is connected to the master node and other subscriber nodes connected to the master node;
• and
• pushing, by the subscriber node, the commit redo record to the other subscriber nodes, to replicate the data of the write operation happened on the master node to the other subscribers.

A method for master node switching, the method comprising:

• performing, by a subscriber node among at least two subscriber nodes, when a master node is down, data synchronization among the subscriber node and the other subscriber nodes until the data synchronization is finished or a pre-configured time period for the data synchronization is reached, wherein the subscriber nodes and the master node are in a same cluster;
• transmitting, by the subscriber node which is acted as a new master node, a pull request to the remaining subscriber nodes which are not elected to be the new master node, after the subscriber node is selected from the subscriber nodes to be the new master node based on a highest commit sequence number (CSN) of the subscriber nodes;
• and
• receiving, by the subscriber node which is acted as the new master node, data from the remaining subscriber nodes which are not elected to be the new master node.

A method for master node switching, the method comprising:

• performing, by a subscriber node among at least two subscriber nodes, when a master node is down, data synchronization among the subscriber node and the other subscriber nodes until the data synchronization is finished or a pre-configured time period for the data synchronization is reached, wherein the subscriber nodes and the master node are in a same cluster;
• electing, by the subscriber node together with the other subscriber nodes, a subscriber node to be a new master node from the subscriber nodes based on a highest commit sequence number (CSN) of the subscriber nodes;
• receiving, by the subscriber node which is not elected to be the new master node, a pull request issued by the new master node;
• and
• transmitting, by the subscriber node which is not elected to be the new master node, data in the subscriber node to a new master node in response to the pull request transmitted by the new master node.

A master node, comprising:

• memory;
• and
• a processor coupled to the memory, wherein the processor is configured to:
• generate a first commit redo record attached with a commit sequence number (CSN), when a first write operation happens on a first physical entity in the master node, wherein the master node is preset with at least two physical entities including the first physical entity and a second physical entity and the master node is connected to at least two subscriber nodes including a first subscriber node and a second subscriber node, wherein the CSN attached in the first commit redo record is used for indicating the sequence of the first write operation within the master node, and
• push the first commit redo record attached with the CSN to a first subscriber node corresponding to the first physical entity based on mapping relationships between the physical entities and the subscriber nodes, wherein the first commit redo record is used for replicating data of the first write operation from the master node to the first subscriber node corresponding to the first physical entity, and then to the second subscriber node through pushing, by the first subscriber node, the received first commit redo record, to the second subscriber node.

A subscriber node, comprising:

• memory;
• and
• a processor coupled to the memory, wherein the processor is configured to:
• receive a commit redo record attached with a commit sequence number (CSN) from a master node, wherein the CSN attached in the commit redo record is used for indicating the sequence of a write operation within the master node,
• update a CSN of the subscriber node to the CSN attached to the commit redo record, when an absolute value of the CSN attached to the commit redo record minus the CSN of the subscriber node is 1, wherein the CSN of the subscriber node is set to 0 during an initialization,
• replay the commit redo record to replicate data of a write operation happened on a master node from the master node to the subscriber node, wherein the subscriber node is connected to the master node and other subscriber nodes connected to the master node, and
• push the commit redo record to the other subscriber nodes, to replicate the data of the write operation happened on the master node to the other subscribers.

A subscriber node, comprising:

• memory;
• and
• a processor coupled to the memory, wherein the processor is configured to:
• when a master node is down, perform data synchronization among the subscriber node and the other subscriber nodes until the data synchronization is finished or a pre-configured time period for the data synchronization is reached, wherein the subscriber nodes and the master node are in a same cluster,
• transmit a pull request to the remaining subscriber nodes which are not elected to be a new master node, after the subscriber node is selected from the subscriber nodes to be the new master node based on a highest commit sequence number (CSN) of the subscriber nodes, and
• receive data from the remaining subscriber nodes which are not elected to be the new master node.

A subscriber node, comprising:

• memory;
• and
• a processor coupled to the memory, wherein the processor is configured to:
• when a master node is down, perform data synchronization among the subscriber node and the other subscriber nodes until the data synchronization is finished or a pre-configured time period for the data synchronization is reached, wherein the subscriber nodes and the master node are in a same cluster,
• together with the other subscriber node, elect a subscriber node to be a new master node from the subscriber nodes based on a highest commit sequence number (CSN) of the subscriber node,
• when the subscriber node is not elected to be the new master node, receive a pull request issued by the new master node, and
• when the subscriber node is not elected to be the new master node, transmit data in the subscriber node to a new master node in response to the pull request transmitted by the new master node.