Techniques for synchronous access to database replicas

A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to execute a process for providing consistency in read-after-write operations in a multi-replica relational database system (RDS), the process comprising:

• receiving a read transaction to read data from a database (DB) replica of the multi-replica RDS;
• obtaining counter values of DB replicas in the multi-replica RDS, wherein a counter value of a DB replica in the multi-replica RDS determines a current version of recent data stored in the respective DB replica;
• determining a write replica out of the DB replicas in the multi-replica RDS;
• determining if there is an available read replica, wherein a read replica is a replica other than the determined write replica having counter value at least equals to a counter value of the write replica;
• waiting, when there is no available read replica, a preconfigured time window to allow all replicas to synchronize with the write replica;
• and
• reading the data from a read replica when the preconfigured time window has elapsed.

A controller for providing consistency in read-after-write operations in a multi-replica relational database system (RDS), comprising:

• a processing circuitry;
• and
• a memory, the memory containing instructions that, when executed by the processing circuitry, configure the controller to:
• receive a read transaction to read data from a database (DB) replica of the multi-replica RDS;
• obtain counter values of DB replicas in the multi-replica RDS, wherein a counter value of a DB replica in the multi-replica RDS determines a current version of recent data stored in the respective DB replica;
• determine a write replica out of the DB replicas in the multi-replica RDS;
• determine if there is an available read replica, wherein a read replica is a replica other than the determined write replica having counter value at least equals to a counter value of the write replica;
• wait, when there is no available read replica, a preconfigured time window to allow all replicas to synchronize with the write replica;
• and
• read the data from a read replica when the preconfigured time window has elapsed.