Scalable log-based continuous data protection for distributed databases

• US 10,423,493 B1
• Filed: 12/21/2015
• Issued: 09/24/2019
• Est. Priority Date: 12/21/2015
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Vig, Akshat, Pol, Parikshit Shivajirao, Subramanian, Subramanian Sankara, Pokkunuri, Rama Krishna Sandeep, Loganathan, Rajaprabhu Thiruchi, Chandrasekharan, Harini
• Primary Examiner(s): Vital, Pierre M
• Assistant Examiner(s): Ho, Andrew N
• Application Number:
  US14/977,472
• Time in Patent Office: 1,373 Days
• Field of Search: 707685
• CPC Class Codes: G06F 11/0757 by exceeding a time limit, i.e. time-out, e.g. watchdogs
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery
• CPC Class Codes: G06F 11/2094 Redundant storage or storage space G06F11/2056 takes precedence
• CPC Class Codes: G06F 11/3006 where the computing system is distributed, e.g. networked systems, clusters, multiprocessor systems multiprogramming arrangements G06F9/46; allocation of resources G06F9/50
• CPC Class Codes: G06F 11/3055 Monitoring arrangements for monitoring the status of the computing system or of the computing system component, e.g. monitoring if the computing system is on, off, available, not available error or fault processing without redundancy G06F11/0703; error detection or correction by redundancy in data representation G06F11/08; error detection or correction by redundancy in operation G06F11/14; error detection or correction by redundancy in hardware G06F11/16
• CPC Class Codes: G06F 16/1815 Journaling file systems
• CPC Class Codes: G06F 16/2322 using timestamps
• CPC Class Codes: G06F 2201/84 Using snapshots, i.e. a logical point-in-time copy of the data

Abstract

In response to determining that continuous data protection is to be enabled for a particular table of a database service, a service component verifies that automated transmission of change records of the table to a log-structured journal has been configured. A given change record comprises a before-image and an after-image associated with a committed database write, and is assigned a unique sequence number. In response to a determination to restore the table as of a specified point in time, a restore record set is identified from the journal with respect to a selected snapshot of the table. The restore record set includes change records which are not represented in the snapshot and are to be represented in the restored table. A restore result table is created using the selected snapshot and the restore record set.

Claims

1. A system, comprising:

a plurality of computing devices of a distributed non-relational database service of a provider network;

wherein the plurality of computing devices are configured to;

receive a request from a client via a programmatic interface to enable continuous data protection for a particular table to provide for point-in-time restore of the particular table, wherein at least a portion of the particular table is stored at one or more storage devices using an overwrite-in-place protocol;

in response to the request from the client to enable continuous data protection, initiate;

(a) transmission of change records of the particular table to a log-structured journal at which a write-once protocol is implemented, wherein a particular change record includes;

one or more before-values for one or more records modified in the particular table by a write operation directed to the particular table,

one or more after-values for the one or more records modified in the particular table by the write operation directed to the particular table, and

a respective sequence number indicative of an order in which the change records are processed by a journal manager of the log-structured journal, and

(b) generation of a sequence of snapshots of the particular table, wherein a particular snapshot of the sequence of snapshots represents write operations performed on the particular table up to a corresponding snapshot request time, and wherein the particular snapshot is stored at a service other than the distributed non-relational database service;

in response to determining that a point-in-time restore operation with a targeted restoration time of the particular table is to be performed,

identify, with respect to a selected snapshot of the sequence of snapshots, a restore record set stored in the log-structured journal, wherein a starting sequence number of the change records in the restore record set corresponds to a first change record which is not represented in the selected snapshot, and wherein an ending sequence number of the change records in the restore record set corresponds to a second change record which was inserted into the log-structured journal no later than the targeted restoration time;

initiate application of changes indicated in the restore record set to the selected snapshot to generate a restore-result table; and

provide an indication of the restore-result table to the client.

2. The system as recited in claim 1, wherein to identify the restore record set, the plurality of computing devices initiate an examination of a particular time marker record inserted in the log-structured journal, wherein the particular time marker record (a) indicates a timestamp derived from a clock time of a particular node of the distributed non-relational database service at which a portion of the particular table is stored, and (b) does not include a pre-image or a post-image associated with a write operation directed to the particular table.

3. The system as recited in claim 1, wherein determining that the point-in-time restore operation is to be performed comprises receiving a first restore request from the client, wherein the plurality of computing devices are configured to:

receive a request to generate automated snapshots of the particular table, wherein the request indicates one or more triggering conditions for initiating an automated snapshot;

store a first automated snapshot of the particular table in accordance with the request to generate automated snapshots;

prepare, in response to a second restore request associated with the particular table, a second restore-result table using (a) the first automated snapshot and (b) a different restore record set identified from the log-structured journal.

4. The system as recited in claim 1, wherein the plurality of computing devices are configured to:

schedule an operation to start creating a particular snapshot of the sequence of snapshots based at least on part on determining a change in the size of the log-structured journal since a previous snapshot of the sequence was generated.

5. The system as recited in claim 1, wherein the plurality of computing devices are configured to:

receive a change log read request via a stream-oriented programmatic interface from the client; and

transmit, to the client, contents of one or more change records of the log-structured journal.

6. A method, comprising:

performing, by one or more nodes of a database service of a provider network, wherein the one or more nodes are implemented at one or more computing devices;

determining that continuous data protection for a particular table of the database service is to be enabled;

in response to determining that continuous data protection is to be enabled, initiating;

(a) transmission of change records of the particular table to a log-structured journal, wherein a particular change record includes;

one or more before-values for one or more records modified in the particular table by a write operation directed to the particular table, and

one or more after-values for the one or more records modified in the particular table by the write operation directed to the particular table, and

(b) generation of one or more snapshots of the particular table;

in response to determining a target restoration time of the particular table,

identifying, with respect to a selected snapshot of the one or more snapshots, a restore record set stored in the log-structured journal, wherein a starting sequence number of the change records in the restore record set corresponds to a first change record which is not represented in the selected snapshot, and wherein an ending sequence number of the change records in the restore record set corresponds to a second change record which was inserted into the log-structured journal no later than the target restoration time; and

initiating a generation of a restore-result table, wherein the generation of the restore-result table includes applying changes indicated in the restore record set to the selected snapshot.

7. The method as recited in claim 6, further comprising:

adding, to the log-structured journal by a journal manager, one or more time marker records, wherein a particular time marker record of the one or more time marker records indicates a timestamp;

wherein said identifying the restore record set comprises examining the particular time marker record.

8. The method as recited in claim 7, wherein the one or more nodes comprise a first table storage node at which at least a portion of the particular table is stored, the method further comprising:

determining the timestamp by the first table storage node based at least in part on (a) a time marking interval parameter and (b) a local clock time of the first table storage node; and

transmitting, to the journal manager by the first table storage node, a message comprising the timestamp.

9. The method as recited in claim 8, wherein the first table storage node is a member of a replication group established for the particular table, wherein the replication group comprises a second table storage node, the method further comprising:

computing, based at least in part on a trusted time source, a local clock drift associated with the first table storage node; and

in response to determining that the local clock drift exceeds a threshold, failing over one or more functions of the first table storage node to the second table storage node.

10. The method as recited in claim 6, wherein determining said target restoration time comprises receiving a first restore request indicating the targeted restoration time at the one or more nodes, the method further comprising performing, by the one or more nodes:

receiving a request to generate automated snapshots of the particular table, wherein the request indicates one or more triggering conditions for initiating an automated snapshot;

storing a first automated snapshot of the particular table in accordance with the request to generate automated snapshots; and

preparing, in response to a second restore request associated with the particular table, a second restore-result table using (a) the first automated snapshot and (b) a different restore record set obtained by examining the log-structured journal.

11. The method as recited in claim 6, further comprising performing, by the one or more nodes:

scheduling an operation to start creating a particular snapshot of the one or more snapshots based at least on part on determining a change in the size of the log-structured journal since a previous snapshot of the one or more snapshots was generated.

12. The method as recited in claim 6, further comprising performing, by the one or more nodes:

determining a restoration response time goal for the particular table; and

scheduling an operation to start creating a particular snapshot of the one or more snapshots based at least on part on the restoration response time goal.

13. The method as recited in claim 6, further comprising performing, by the one or more nodes:

receiving a change log read request via a stream-oriented programmatic interface from a client; and

transmitting, to the client, contents of one or more change records of the log-structured journal.

14. The method as recited in claim 6, further comprising performing, by the one or more nodes:

in response to receiving a request via a programmatic interface to disable continuous data protection for the particular table,

terminating transmission of the change records of the particular table; and

terminating generation of snapshots of the particular table.

15. The method as recited in claim 6, wherein the provider network comprises a first data center and a second data center, wherein at least a portion of the particular table is stored at the first data center, the method further comprising performing, by the one or more nodes:

receiving a restore request indicating (a) the target restoration time for the particular table and (b) a geographical preference with respect to a location of the restore-result table; and

selecting, based at least in part on the geographical preference, one or more storage destinations located in the second data center for the restore-result table.

16. The method as recited in claim 6, wherein the database service implements a first data model, the method further comprising performing, by the one or more nodes:

in response to receiving a point-in-time restore request indicating that the particular table is to be restored at a second database service at which a different data model is implemented, initiating a generation of a second restore-result table at the second database service, wherein the generation of the restore-result table includes applying, to the selected snapshot, changes indicated in a different restore record set identified with respect to a second target restoration time indicated in the point-in-time restore request.

17. A non-transitory computer-accessible storage medium storing program instructions that when executed on one or more processors:

determine that continuous data protection is to be enabled for a particular table of a database service;

in response to determining that continuous data protection is to be enabled, verify that automated transmission of change records of the particular table to a log-structured journal has been configured, wherein a particular change record includes;

one or more before-values for one or more records modified in the particular table by a write operation directed to the particular table, and

one or more after-values for the one or more records modified in the particular table by the write operation directed to the particular table;

in response to determining a target restoration time of the particular table,

identify, with respect to a selected snapshot of one or more snapshots of the particular table, a restore record set stored in the log-structured journal, wherein a starting sequence number of the change records in the restore record set corresponds to a first change record which is not represented in the selected snapshot, and wherein an ending sequence number of the change records in the restore record set corresponds to a second change record which was inserted into the log-structured journal no later than the target restoration time; and

initiate generation of a restore-result table, wherein the generation of the restore-result table includes application of changes indicated in the restore record set to the selected snapshot.

18. The non-transitory computer-accessible storage medium as recited in claim 17, wherein to identify the restore record set, the instructions when executed on the one or more processors initiate an examination of a particular time marker record inserted in the log-structured journal, wherein the particular time marker record indicates a timestamp.

19. The non-transitory computer-accessible storage medium as recited in claim 17, wherein the instructions when executed on the one or more processors:

in response to determining that continuous data protection is to be enabled for the particular table, verify that the particular table has been configured for automated snapshot generation.

20. The non-transitory computer-accessible storage medium as recited in claim 17, wherein the database service is implemented at a provider network which comprises a plurality of geographically dispersed sub-units including a first sub-unit and a second sub-unit, wherein at least a portion of the particular table is stored in the first sub-unit, and wherein the instructions when executed on the one or more processors:

select, based at least in part on a client-indicated preference, the second sub-unit as a destination at which at least a portion of the restore-result table is to be stored.

21. The non-transitory computer-accessible storage medium as recited in claim 17, wherein the instructions when executed on the one or more processors:

provide, before the restore-result table has been fully populated, an indication of a status of the restore-result table via a programmatic interface to a client.