Table replication in a database environment

A computer system that implements a source node, the source node in communication with a replica node, the computer system comprising:

• a processor;
• a memory coupled to the processor;
• one or more computer-readable storage media storing computer-executable instructions that, when executed, cause the computer system to perform operations comprising:
• receiving from a database client a database operation;
• executing the database operation on the at least one database table at the source node;
• sending the database operation to the replica node;
• receiving a commit request for a transaction that comprises the database operation;
• in response to receiving the commit request, assigning a commit identifier to the transaction;
• in response to the receiving the commit request, sending a synchronous prepare commit request to the replica node, wherein, in response to receiving the synchronous prepare commit request from the source node, the replica node assigns a precommit identifier to the transaction, which marks the transaction as in doubt, wherein an in-doubt database transaction is a database transaction that has not been committed or aborted and the precommit identifier determines whether the transaction will be visible to a read operation;
• in response to the receiving the commit request, after assigning the commit identifier to the transaction and after, or currently with, sending the synchronous prepare commit request to the replica node, and without waiting to receive a synchronous prepare commit acknowledgment from the replica node in response to the synchronous prepare commit request:
• initiating writing of a commit log entry for the transaction comprising the database operation to persistent storage, the commit log entry comprising the commit identifier, wherein the synchronous prepare commit request is sent to the replica node prior to finishing writing of the commit log;
• and
• wherein a time taken to complete a replication process is reduced by overlapping the writing of the commit log with the synchronous prepare commit request and the synchronous prepare commit acknowledgement;
• receiving the synchronous prepare commit acknowledgment from the replica node, wherein the prepare commit acknowledgment indicates that the database operation was executed at the replica node;
• and
• after receiving the synchronous prepare commit acknowledgement from the replica node, sending a commit acknowledgement to the database client.

A method, implemented by a replica node comprising a processing unit and memory that implements part of a database replication system, the replica node being in communication with a source node that stores at least one database table, the replica node storing a copy of the at least one database table, comprising:

• receiving, from a database client, a first query associated with a first database transaction;
• determining that the first query depends on a first database operation in the first database transaction;
• determining that the first database operation has not been executed by the replica node;
• by the replica node, delaying execution of the first query pending execution of the first database operation;
• by the replica node, determining that a time associated with the delaying execution exceeds a threshold;
• by the replica node, based on the determining that the time exceeds the threshold, forwarding the first query to the source node for execution by the source node;
• receiving the first database operation asynchronously from the source node;
• executing the first database operation at the replica node on the copy of the at least one database table;
• receiving a first synchronous notification from the source node to prepare to commit the first transaction;
• precommitting the first database transaction at the replica node, the precommitting marking the first database transaction as in-doubt at the replica node and assigning a first precommit ID value to the first database transaction, wherein an in-doubt database transaction is a database transaction that has not been committed or aborted and the first precommit ID value determines whether the first database transaction will be visible to a read operation;
• sending a first synchronous prepare commit acknowledgment to the source node, the first prepare commit acknowledgment indicating that the first database transaction was precommitted by the replica node;
• receiving, from the database client, a second query associated with a second database transaction;
• determining that the second query depends on a second database operation in the second database transaction;
• determining that the second database operation has not been executed by the replica node;
• based on the determining that the second database operation has not been executed by the replica node, holding the second query for execution by the replica node pending execution of the second database operation;
• receiving the second database operation asynchronously from the source node;
• executing the second database operation at the replica node on the copy of the at least one database table;
• in response to executing the second database operation at the replica node, executing the second query, wherein a delay associated with the holding does not exceed the threshold;
• receiving a second synchronous notification from the source node to prepare to commit the second database transaction;
• precommitting the second database transaction at the replica node, the precommitting marking the second database transaction as in-doubt at the replica node and assigning a second precommit ID value to the second database transaction, wherein an in-doubt database transaction is a database transaction that has not been committed or aborted and the second precommit ID value can be used to determine whether the second database transaction will be visible to a read operation;
• and
• sending a second synchronous prepare commit acknowledgment to the source node, the second prepare commit acknowledgment indicating that the second database transaction was precommitted by the replica node.

One or more computer-readable storage media comprising:

• computer-executable instructions that, when executed by a computing system comprising a processor and a memory, the computing system implementing a replica node in communication with a source node, cause the computing system to receive, from a database client, a first query associated with a first database transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to determine that the first query depends on a first database operation in the first database transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to determine that the first database operation has not been executed by the replica node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to by the replica node, delay execution of the first query pending execution of the first database operation;
• computer-executable instructions that, when executed by the computing system, cause the computing system to by the replica node, determine that a time associated with the delaying execution exceeds a threshold;
• computer-executable instructions that, when executed by the computing system, cause the computing system to by the replica node, based on the determining that the time exceeds the threshold, forward the first query to the source node for execution by the source node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to receive the first database operation asynchronously from the source node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to execute the first database operation at the replica node on the copy of the at least one database table;
• computer-executable instructions that, when executed by the computing system, cause the computing system to receive a first synchronous notification from the source node to prepare to commit the first transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to precommit the first database transaction at the replica node, the precommitting marking the first database transaction as in-doubt at the replica node and assigning a first precommit ID value to the first database transaction, wherein an in-doubt database transaction is a database transaction that has not been committed or aborted and the first precommit ID value determines whether the first database transaction will be visible to a read operation;
• computer-executable instructions that, when executed by the computing system, cause the computing system to send a first synchronous prepare commit acknowledgment to the source node, the first prepare commit acknowledgment indicating that the first database transaction was precommitted by the replica node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to receive, from the database client, a second query associated with a second database transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to determine that the second query depends on a second database operation in the second database transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to determine that the second database operation has not been executed by the replica node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to, based on the determining that the second database operation has not been executed by the replica node, hold the second query for execution by the replica node pending execution of the second database operation;
• computer-executable instructions that, when executed by the computing system, cause the computing system to receive the second database operation asynchronously from the source node;
• computer-executable instructions that, when executed by the computing system, cause the computing system to execute the second database operation at the replica node on the copy of the at least one database table;
• computer-executable instructions that, when executed by the computing system, cause the computing system to, in response to executing the second database operation at the replica node, execute the second query, wherein a delay associated with the holding does not exceed the threshold;
• computer-executable instructions that, when executed by the computing system, cause the computing system to receive a second synchronous notification from the source node to prepare to commit the second database transaction;
• computer-executable instructions that, when executed by the computing system, cause the computing system to precommit the second database transaction at the replica node, the precommitting marking the second database transaction as in-doubt at the replica node and assigning a second precommit ID value to the second database transaction, wherein an in-doubt database transaction is a database transaction that has not been committed or aborted and the second precommit ID value can be used to determine whether the second database transaction will be visible to a read operation;
• and
• computer-executable instructions that, when executed by the computing system, cause the computing system to send a second synchronous prepare commit acknowledgment to the source node, the second prepare commit acknowledgment indicating that the second database transaction was precommitted by the replica node.