Operating logging for online recovery in shared memory information systems

• US 7,739,244 B2
• Filed: 10/14/2004
• Issued: 06/15/2010
• Est. Priority Date: 10/14/2004
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Edson, Kirk Meredith, McAuliffe, Mark Lawrence, Verma, Sanjay, Hoang, Chi-Kim
• Primary Examiner(s): Ly; Cheyne D
• Assistant Examiner(s): MENG, JAU SHYA
• Application Number:
  US10/966,285
• Publication Number:
  US 20060085426A1
• Time in Patent Office: 2,070 Days
• Field of Search: 707/103.R, 707/1, 707/2, 707/202, 705/34, 709/226, 714/2
• US Class Current: 707/684
• CPC Class Codes: G06F 11/1438 Restarting or rejuvenating

Abstract

Clients are permitted to directly attach to a client server system. An oplog management system tracks changes as clients make the changes to control structures of the direct connection client server system. The Oplogging system stores changes in oplogs. Then, if a client fails, the Oplogging system can repair the control structures, either by undoing the changes made by the client, or else completing the changes in place of the client; depending upon principles defined.

Claims

1. A computer-based apparatus, comprising:

a system including a server supporting simultaneous access by a first client application and a second client application to managed data of the server, the server including a processor, wherein;

each of the first client application and the second client application operate as a directly connected client model wherein the first client application and the second client application are each linked to an executable server library on the server that supports the first client application and the second client application accessing the managed data via modifying control data of the server,

each of the first client application and the second client application and at least parts of the executable server library are executed as a respective same process,

the managed data is a collection of data items accessed by the first client application and the second client application stored in a main memory on the server,

the control data controls concurrency, permission and access to the managed data, and the control data comprises state information separate from the data items in the managed data and used by critical sections of the executable server library, where a failure or unexpected exit of the first client application or the second client application leave the control data in an inconsistent partially modified state that prevents the executable server library from rolling back the managed data to a consistent state prior to the failure or unexpected exit of the first client application or second client application;

an oplog generator to generate oplogs, the oplogs storing information about changes in the control data while the first client application is operating within the critical sections of the server library but fails to finish all necessary changes to the control data; and

an oplog cleaner process configured to use the oplogs to perform a first operation that undoes the changes in the control data made by the first client application or second client application prior to the failure or unexpected exit of the first client application or second client application and before completing one of the critical sections of the executable server library, wherein undoing the changes in the control data enable the oplog cleaner to then conduct a second roll back operation where the oplog cleaner then uses a second roll back log separate from the oplogs to separately roll back the data items in the managed data to a state prior to the failure or unexpected exit of the first client application or second client application from the executable server library.

2. The computer-based apparatus according to claim 1, wherein the critical sections start when a synchronization primitive is acquired and end when the synchronization primitive is released, and the oplog cleaner process uses the information stored in the oplogs to undo the change in the control data.

3. The computer-based apparatus according to claim 2, wherein:

the oplogs include a partially modified linked list; and

the oplog cleaner process uses the information stored in the oplogs to convert the partially modified link list back into an original non-modified linked list.

4. The computer-based apparatus according to claim 2, wherein the oplogs include:

a first list identifier to identify a first list;

a second list identifier to identify a second list;

an element identifier identifying an element to be moved from the first list to the second list;

a first next element identifier identifying a next element in the first list; and

a second next element identifier identifying a next element in the second list.

5. The computer-based apparatus according to claim 1, further comprising a connection monitor daemon operating on the server and configured to:

grant direct connections between the first client application and the second client application on the terminals with the executable server library on the server;

open sockets with the first client application and the second client application; and

monitor the sockets between the server and the first client application and the second client application.

6. The computer-based apparatus according to claim 5, wherein the connection monitor daemon is operative to:

detect when one of the sockets has been unexpectedly closed; and

launch the oplog cleaner process upon detection of a failure of one of the first client application and the second client application.

7. The computer-based apparatus according to claim 1, wherein the information about the changes to the control data stored in the oplogs includes values associated with table creation, index creation, and cursor operations.

8. The computer-based apparatus according to claim 1, wherein the database system is an in-memory database system.

9. The computer-based apparatus according to claim 1 wherein the information about the changes to the control data stored in the oploas includes instructions that leave tables in an open state and that move elements between lists.

10. The computer-based apparatus according to claim 1, wherein the information stored in the oplogs includes meta-data values associated with a structure of a database containing the data items and state information about currently open client actions performed by the first and second client applications.

11. The computer-based apparatus according to claim 2, wherein the synchronization primitive comprises a lock.

12. A computer-based apparatus comprising:

a system including a server supporting simultaneous access by a first client and a second client, the server including a processor, wherein;

each of said first client and said second client operating as a directly connected client model wherein applications executed by the first and second client are linked to an executable library on the server that supports said first and second client applications modifying control data on the server,

the applications on the first and second clients and at least parts of the executable library on the server linked to the applications are executed as a same process,

the server has associated managed data that is a collection of data items accessed by the same process executed by the applications on the first and second clients and the linked part of the executable library on the server,

the control data controls concurrency, permission and access to the managed data on the server, and the control data includes a synchronization primitive configured to manage access of the first client and the second client to the managed data, and

the control data for the server includes critical sections where a failure or unexpected exit of the first client or second client leave the control data in an inconsistent partially modified state;

an oplog generator to generate oplogs for the critical sections of the control data, the oplogs storing information about changes in the control data while the application for said first client is operating within the critical sections of the control data, wherein the oplog generator is further configured to generate managed data logs separately from the oplogs that store changes to the data items in the managed data independently of the changes in the control data stored in the oplogs;

an oplog cleaner process to complete or reverse said critical sections of the control data of the server upon determination that said first client failed before completing said changes to the critical sections of the control data; and

wherein the oplog cleaner process uses the oplogs to perform a first operation that undoes the changes in the control data made by the first client or second client prior to completing said changes to the critical sections of the control data, and

wherein the oplog cleaner, after undoing the changes in the control data, then conducts a second separate roll back operation where the oplog cleaner uses a second managed data log separate from the oplogs to separately roll back the data items in the managed data to a state prior to the first client or second client completing said changes to the critical sections of the control data.

13. The computer-based apparatus according to claim 12, wherein the oplogs include:

a list identifier;

a length for a list identified by the list identifier; and

an operation to be performed iteratively on the list.

14. A method for operating a computer system, comprising:

configuring a client computer to directly attach to a server in the database system such that the database system executes a client application of the client computer and at least a portion of the database system in a single process, wherein;

the client application is linked to an executable server library that supports the client application accessing managed data of the server via modifying control data of the server,

the managed data is a collection of data items accessed by the client application,

the control data controls concurrency, permission and access to the managed data,

the executable server library includes critical sections where a failure or unexpected exit of the client application leaves the control data in an inconsistent partially modified state;

maintaining, via the executable server library, the control data to manage the accessing of the managed data by the client application;

logging changes to the control data while the client application is operating within the critical sections;

separately logging changes to the data items in the managed data;

monitoring for a failure of the client application;

in response to detection of the failure of the client application, undoing the logged changes made to the control data by the client application while operating within the critical sections of the executable server library prior to the failure or unexpected exit of the client application; and

after undoing the logged changes made to the control data, conducting a second separate roll back operation to separately roll back the logged changes made to the data items in the managed data by the client application prior to the failure or unexpected exit of the client application from the executable server library.

15. The method according to claim 14, wherein:

the method further comprises configuring a second client computer to directly attach to the system; and

the repairing the control data includes, repairing the control data in response to detection of the failure of the client application without disconnecting the second client computer from the system.

16. The method according to claim 14, wherein the logging changes to the control data includes:

maintaining a list of open transactions;

associating each one of one or more open transactions in the list of open transactions with a respective cursor; and

associating each of the respective cursors with a respective table in the system, the respective table to be modified as part of the one of the open transactions associated with the respective cursor.

17. The method according to claim 16, wherein, for a particular one of the open transactions, the repairing the control data includes:

removing a first association between the respective cursor and the respective table, upon determination that the first association exists; and

removing a second association between the particular open transaction and the respective cursor, upon determination that the second association exists.

18. The method according to claim 14, wherein the logging changes to the control data includes:

identifying a list in the control data being changed in response to the client application accessing the managed data;

identifying an operation to be performed by the executable server library on an element in the list; and

storing information to enable undoing the operation to be performed by the executable server library on the element.

19. The method according to claim 18, wherein the identifying an operation includes identifying the element to be added to the list.

20. The method according to claim 18, wherein the identifying an operation includes identifying the element to be removed from the list.

21. The method according to claim 20, wherein the repairing the control data includes:

identifying the element to be removed as a hole in the list; and

iteratively moving the hole until the list is continuous.

22. The method according to claim 18, wherein the identifying an operation includes identifying the element to be moved within the list.

23. The method according to claim 18, wherein the repairing the control data includes restoring the element to an original state.

24. The method according to claim 14, wherein the logging changes to the control data includes:

identifying a list in the control data being changed in response to the client application accessing the managed data;

identifying an operation to be performed by the executable server library on elements in the list; and

storing information about the operation.

25. The method according to claim 24, wherein the repairing the control data includes:

identifying elements to which the executable server library has not performed the operation; and

iteratively performing the operation on the identified elements.

26. The method according to claim 25, wherein:

the identifying an operation includes identifying a hole move operation to be performed by the client computer on the elements in the list; and

the iteratively performing the operation on the identified elements includes iteratively moving a hole past the identified elements until the list is continuous.

27. The method according to claim 14, wherein:

the configuring a client computer to directly attach to the server includes configuring the client computer to directly attach to a database system;

the logging changes to the control data includes logging changes to a control structure of the database system in response to the client application accessing the managed data; and

the repairing the control data includes repairing the control structure of the database system while maintaining an availability of the database system.

28. The method according to claim 27, wherein:

the configuring the client computer to directly attach to a database system includes configuring the client computer to directly attach to an in-memory database system;

the logging changes to the control data includes logging changes to a control structure of the in-memory database system in response to the client application accessing the managed data; and

the repairing the control data includes repairing the control structure of the in-memory database system while maintaining an availability of the in-memory database system.

29. An article comprising a computer-readable memory having associated data stored thereon, wherein the data, when accessed, results in a computer performing operations comprising:

configuring a client to directly attach to a server in a system such that a client application of the client and an executable server library execute as a single process, wherein;

the client application is linked to the executable server library, the executable server library to support the client application accessing managed data of the server via modifying control data of the server,

the managed data is a collection of data items accessed by the client application,

the control data controls concurrency, permission and access to the managed data, and the control data includes a synchronization primitive configured to manage access of the client application to the managed data, and

the executable server library includes critical sections where a failure or unexpected exit of the client application leaves the control data in an inconsistent partially modified state;

maintaining, via the executable server library, the control data to manage the accessing of the managed data by the client application;

logging changes to the control data while the client application is operating within the critical sections;

monitoring for a failure of the client application;

undoing the changes made to the control data by the client application while operating within the critical sections of the executable server library prior to the failure or unexpected exit of the client application from the executable server library; and

conducting a second separate roll back operation after undoing the changes made to the control data to separately roll back changes in the data items in the managed data made by the client application prior to the failure or unexpected exit of the client application from the executable server library.

30. The article according to claim 29, wherein:

the operations further comprise configuring a second client to directly attach to the system; and

the repairing the control data includes repairing the control data in response to detection of the failure of the client application without disconnecting the second client from the system.

31. The article according to claim 29, wherein the logging changes to the control data includes:

maintaining a list of open transactions;

associating each one of one or more open transactions in the list of open transactions with a respective cursor; and

associating each of the respective cursors with a respective table in the system, the respective table to be modified as part of the one of the open transactions associated with the respective cursor.

32. The article according to claim 31, wherein, for a particular one of the open transactions, the repairing the control data includes:

removing a first association between the respective cursor and the respective table, upon determination that the first association exists; and

removing a second association between the particular open transaction and the respective cursor, upon determination that the second association exists.

33. The article according to claim 29, wherein the logging changes to the control data includes:

identifying a list in the control data being changed in response to the client application accessing the managed data;

identifying an operation to be performed by the executable server library on an element in the list; and

storing information to enable undoing the operation to be performed by the executable server library on the element.

34. The article according to claim 33, wherein the identifying an operation includes identifying the element to be added to the list.

35. The article according to claim 33, wherein the identifying an operation includes identifying the element to be removed from the list.

36. The article according to claim 35, wherein the repairing the control data includes:

identifying the element to be removed as a hole in the list; and

iteratively moving the hole until the list is continuous.

37. The article according to claim 33, wherein the identifying an operation includes identifying the element to be moved within the list.

38. The article according to claim 33, wherein the repairing the control data includes restoring the element to an original state.

39. The article according to claim 29, wherein the logging changes to the control data includes:

identifying a list in the control data being changed in response to the client application accessing the managed data;

identifying an operation to be performed by the executable server library on elements in the list; and

storing information about the operation.

40. The article according to claim 39, wherein the repairing the control data includes:

identifying elements to which the executable server library has not performed the operation; and

iteratively performing the operation on the identified elements.

41. The article according to claim 40, wherein:

the identifying an operation includes identifying a hole move operation to be performed by the client on the elements in the list; and

the iteratively performing the operation on the identified elements includes iteratively moving a hole past the identified elements in the list until the list is continuous.

42. The article according to claim 29, wherein:

the configuring a client to directly attach to the server includes configuring the client to directly attach to a database system;

the logging changes to the control data includes logging changes to a control structure of the database system in response to the client application accessing the managed data; and

the repairing the control data includes repairing the control structure of the database system while maintaining an availability of the database system.

43. The article according to claim 29, wherein:

the configuring the client to directly attach to a database system includes configuring the client to directly attach to an in-memory database system;

the logging changes to the control data includes logging changes to a control structure of the in-memory database system in response to the client application accessing the managed data; and

the repairing the control data includes repairing the control structure of the in-memory database system while maintaining an availability of the in-memory database system.