Speculative recovery using storage snapshot in a clustered database

1. A method for recovery in a database, the method comprising:

detecting a failure in a first computing node, the first computing node serving the database in a cluster of computing nodes;

creating, using a processor and a memory, a snapshot of data of the database;

applying, using the processor and the memory, a subset of log entries to the snapshot, the applying modifying the snapshot to result in a modified snapshot;

preserving an access of the first computing node to the data of the database; and

aborting, responsive to receiving a signal of activity from the first computing node during the applying and after a grace period has elapsed, the applying such that the first computing node can continue serving the database in the cluster.

2. The method of claim 1 , further comprising:

receiving the signal of activity from the first computing node during the applying and after the grace period for the signal of activity has elapsed;

determining a level of performance of the first computing node;

completing the applying responsive to the level of performance begin below a threshold; and

taking over the serving of the database from the first computing node using the snapshot such that the first computing node cannot serve the database in the cluster.

3. The method of claim 2 , further comprising:

combining the modified snapshot with a modified data of the database, wherein the modified data of the database results from the first computing node continuing to modify the data of the database after the creating of the snapshot.

4. The method of claim 3 , wherein the combining comprises a reverse copy operation.

5. The method of claim 1 , further comprising:

communicating to a cluster management infrastructure a waiting period within which to allow the preserving.

6. The method of claim 5 , wherein the waiting period is an estimate of time needed to apply the subset of log entries, and wherein the waiting period is changeable during the applying.

7. The method of claim 1 , wherein the preserving comprises:

allowing the first computing node to continue manipulating the data of the database during the applying.

8. The method of claim 1 , wherein the creating and applying occur are a second computing node in the cluster of computing nodes, and wherein the second computing node and the first computing node concurrently serve the database from the cluster for a waiting period.

9. The method of claim 1 , the applying comprising:

selecting the subset of log entries, wherein the log entries comprise transaction information processed by the first computing node after a checkpoint operation and one of (i) up to a time of the failure, and (ii) prior to a time of the failure.

10. The method of claim 1 , the creating comprising:

making a copy of the data at a storage subsystem after the detecting, wherein the storage subsystem is used for storing the data of the database, wherein the making the copy comprises a data replication on the storage subsystem such that a change tracking and disambiguation of access data is resolved on a storage layer that includes the storage subsystem.

11. The method of claim 10 , wherein the making comprises performing a FlashCopy operation and wherein the copy uses a copy on write mode.

12. The method of claim 1 , wherein the failure comprises failure to receive the signal of activity from the first computing node within the grace period.

13. The method of claim 12 , wherein the signal of activity is a heartbeat message.