Method and apparatus for improved transaction recovery

Claim 3

The method of claim 1 further comprising the step of applying a plurality of changes recorded in a redo log to a data file, wherein the plurality of changes comprises changes made by transactions that existed in the instance but which had not yet been applied to the data file at the time the instance crashed.

Claim 4

The method of claim 1 wherein the step of determining whether the resource belongs to a particular transaction that was active when the instance crashed comprises checking a container for an identity of an owner of the resource.

Claim 5

The method of claim 3 further comprising the steps of: finding transaction information relating to the owner of the resource based on the identity of the owner; and inspecting the transaction information relating to the owner of the resource to determine whether the transaction information relating to the owner of the resource is marked as dead.

Claim 6

The method of claim 1 wherein the step of undoing one or more changes made by the particular transaction comprises: retrieving a location of a rollback entry for the particular transaction; and executing operations specified in the rollback entry.

Claim 7

The method of claim 1 further comprising the step of running an independent process which undoes changes made by transactions having corresponding transaction information indicating that the transactions are dead.

Claim 8

The method of claim 6 wherein the step of running an independent process includes the step of running an independent process which applies a predefined number of undo records on the transactions having corresponding transaction information indicating that the transactions are dead.

Claim 9

The method of claim 1 wherein the step of finding transaction information that corresponds to one or more transactions that existed in the instance comprises: identifying a rollback segment associated with the instance that crashed; accessing a transaction table in the rollback segment, the transaction table including a plurality of entries each of which corresponds to a plurality of transactions that existed in that instance; and reading a status field in each of the plurality of entries.

Claim 10

The method of claim 1 wherein the step of updating the transaction information to indicate that the one or more transactions that were active when the instance crashed are dead comprises setting a flag in a field of an entry that is associated with a transaction in a transaction table.

Claim 11

The method of claim 1, wherein the step of undoing one or more changes made by the particular transaction includes the step of undoing one or more changes made by the particular transaction to the resource.

Claim 12

The method of claim 10, wherein the particular transaction made changes to a plurality of resources, and the step of undoing one or more changes made by the particular transaction to the resource is performed by undoing only the one or more changes made by the particular transaction to the resource.

Claim 13

The method of claim 1, wherein the step of undoing one or more changes made by the particular transaction includes the step of causing the new transaction to undo one or more changes made by the particular transaction.

Claim 16

The method of claim 13 wherein the step of determining that the resource belongs to the prematurely terminated transaction includes the step of checking a container associated with the resource to identify an owner of the resource.

Claim 17

The method of claim 14 further comprising the steps of: finding transaction information relating to the owner of the resource based on the identity of the owner; and inspecting the transaction information relating to the owner of the resource to determine whether the transaction information relating to the owner of the resource is marked as active.

Claim 18

The method of claim 13 wherein the step of undoing one or more changes made by the prematurely terminated transaction comprises: retrieving a location of a rollback entry for the prematurely terminated transaction; and executing operations specified in the rollback entry.

Claim 19

The method of claim 13, wherein the step of undoing one or more changes made by the prematurely terminated transaction includes the step of undoing one or more changes made by the prematurely terminated transaction to the resource.

Claim 20

The method of claim 17, wherein the prematurely terminated transaction made changes to a plurality of resources, and the step of undoing one or more changes made by the prematurely terminated transaction to the resource is performed by undoing only the one or more changes made by the prematurely terminated transaction to the resource.

Claim 21

The method of claim 13, wherein the step of undoing one or more changes made by the prematurely terminated transaction includes the step of causing the new transaction to undo one or more changes made by the prematurely terminated transaction.

Claim 22

The method of claim 13, further comprising the step of running an independent process which undoes changes made by transactions having corresponding transaction information indicating that the transactions are dead.

Claim 23

The method of claim 20, wherein the step of running an independent process includes the step of running an independent process which applies a predefined number of undo records on the transactions having corresponding transaction information indicating that the transactions are dead.

Claim 24

The method of claim 13, wherein the premature termination of the transaction is caused by a crash of an instance of a database, and the step of starting execution of a new transaction after the premature termination of the transaction and before all updates made by the prematurely terminated transaction have been undone includes the step of starting execution of a new transaction after the crash of the instance of the database and before all updates made by the prematurely terminated transaction have been undone.

Claim 25

A computer-readable medium having stored thereon sequences of instructions for recovering after a crash of an instance of a database, wherein one or more transactions were active when the instance crashed, the sequences of instructions including instructions which, when executed by a processor, cause the processor to perform the steps of: finding transaction information that corresponds to one or more transactions that existed in the instance; determining, based on the transaction information, whether the transactions that existed in the instance were active when the instance crashed; if one or more transactions were active when the instance crashed, then updating the transaction information to indicate that the one or more transactions that were active when the instance crashed are dead; making the database available to users after updating the transaction information and before undoing all of the updates performed by the one or more transactions that were active when the instance crashed; executing a new transaction; when the new transaction requests access to a resource, determining that the resource belongs to a particular transaction that was active when the instance crashed; and in response to determining that the resource belongs to the particular transaction that was active when the instance crashed, undoing one or more changes made by the particular transaction.

Claim 26

A computer-readable medium having stored thereon sequences of instructions for recovering after a crash of an instance of a database, wherein one or more transactions were active when the instance crashed, the sequences of instructions including instructions which, when executed by a processor, cause the processor to perform the steps of: finding transaction information that corresponds to one or more transactions that existed in the instance; determining, based on the transaction information, whether the transactions that existed in the instance were active when the instance crashed; if one or more transactions were active when the instance crashed, then updating the transaction information to indicate that the one or more transactions that were active when the instance crashed are dead; making the database available to users after updating the transaction information and before undoing all of the updates performed by the one or more transactions that were active when the instance crashed; executing a new transaction; when the new transaction requests access to a resource, determining that the resource belongs to a particular transaction that was active when the instance crashed; and in response to determining that the resource belongs to the particular transaction that was active when the instance crashed, undoing one or more changes made by the particular transaction.

Claim 27

The computer-readable medium of claim 23 wherein the step of undoing one or more changes made by the particular transaction includes the step of causing the new transaction to apply a plurality of changes recorded in a redo log to a data file, wherein the plurality of changes comprises changes made by transactions that existed in the instance but which had not yet been applied to the data file at the time the instance crashed.

Claim 28

The computer-readable medium of claim 23 wherein the step of determining that the resource belongs to a particular transaction that was active when the instance crashed comprises checking a container for an identity of an owner of the resource.

Claim 29

The computer-readable medium of claim 25 further containing instructions which, when executed by the processor, would cause the processor to perform the steps of: finding transaction information relating to the owner of the resource based on the identify of the owner; and inspecting the transaction information relating to the owner of the resource to determine whether the transaction information relating to the owner of the resource is marked as dead.

Claim 30

The-computer-readable medium of claim 23 wherein the step of undoing one or more changes made by the particular transaction comprises: retrieving a location of a rollback entry for the particular transaction; and executing operations specified in the rollback entry.

Claim 31

The computer-readable medium of claim 23 wherein the step of undoing one or more changes made by the particular transaction comprises running an independent process which undoes one or more changes made by the particular transaction.

Claim 32

The computer-readable medium of claim 23 wherein the step of undoing one or more changes made by the particular transaction includes the step of applying a predefined number of undo records on the particular transaction.

Claim 33

The computer-readable medium of claim 23 wherein the step of finding transaction information that corresponds to one or more transactions that existed in the instance comprises: identifying a rollback segment associated with the instance that crashed; accessing a transaction table in the rollback segment, the transaction table including a plurality of entries each of which corresponds to a plurality of transactions that existed in that instance; and reading a status field in each of the plurality of entries.

Claim 34

The computer-readable medium of claim 23 wherein the step of updating the transaction information to indicate that the one or more transactions that were active when the instance crashed are dead comprises setting a flag in a field of an entry that is associated with a transaction in a transaction table.