Efficient recovery of replicated data items

A method for recovering one or more replicated records, the method comprising:

• recording an entry in a log at an abstract data structure (ADS) layer in a software hierarchy, the entry indicating that an operation was performed on a replicated record, wherein the entry further indicates a desired value of the replicated record;
• recovering the replicated record if the log contains any entries by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record and an uncorrupted copy of the replicated record and replacing the corrupted copy of the replicated record with the uncorrupted copy of the replicated record, wherein the log is maintained at the ADS layer in the software hierarchy that is higher in the software hierarchy than a replicating logical device (RLDev) layer at which the replicated record is recovered;
• determining whether a failure occurred during the operation at least in part by performing a first scan of the log at the ADS layer and inferring an offset to determine whether a data item is potentially corrupted;
• reconciling the replicated record at the RLDev layer by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record if a determination is made that a failure occurred during the operation;
• maintaining the log at the ADS layer, the log storing the plurality of entries associated with changes to a plurality of replicated records for which multiple copies are maintained at the RLDev layer;
• performing a second scan of the log at the ADS layer to recover data associated with the ADS layer;
• and
• recovering the data associated with the ADS layer based on the scan of the log;
• wherein the ADS layer is a higher layer that is an application l ayer, and wherein the RLDev layer is a lower layer that is an operating system layer comprising a replicating device module.

A system for recovering one or more replicated data items, the system comprising:

• a computer;
• and
• a computer program executable by the computer, wherein the computer program comprises computer instructions for:
• maintaining a log at an abstract data structure (ADS) layer in a software hierarchy, the log storing a plurality of entries associated with changes to a plurality of replicated data items for which multiple copies are maintained at a replicating logical device (RLDev) layer below the ADS layer in the software hierarchy;
• determining whether the log contains one or more entries indicating a corruption in a replicated data item at least in part by performing a first scan of the log at the ADS layer and inferring an offset to determine whether an entry indicates a corruption in a replicated data item, wherein the one or more entries further indicate a desired value of the replicated data item;
• and
• recovering the replicated data item if the log contains any such entries by comparing a plurality of copies of the replicated data item to identify a corrupted copy of the replicated data item and an uncorrupted copy of the replicated data item, wherein the log is maintained at the ADS layer;
• identifying a location of the corrupted copy of the replicated data item using an offset, wherein the offset is part of an entry in the log;
• determining whether a failure occurred during an operation;
• reconciling a replicated record at the RLDev layer by comparing the plurality of copies of the replicated record to identify the corrupted copy of the replicated record if a determination is made that the failure occurred during the operation;
• performing a second scan of the log at the ADS layer to recover data associated with the ADS layer;
• and
• recovering the data associated with the ADS layer based on the second scan of the log.

A system for recovering one or more replicated data items, the system comprising:

• a computer;
• and
• a computer program executable by the computer, wherein the computer program comprises computer instructions for:
• recording an entry in a log at an abstract data structure (ADS) layer in a software hierarchy, the entry indicating that an operation was performed on a replicated record, wherein the entry further indicates a desired value of the replicated data item;
• determining whether a failure occurred during the operation at least in part by performing a first scan of the log at the ADS layer and inferring an offset to determine whether a data item is potentially corrupted;
• recovering the replicated record if the log contains any entries by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record and an uncorrupted copy of the replicated record and replacing the corrupted copy of the replicated record with the uncorrupted copy of the replicated record, wherein the log is maintained at the ADS layer that is higher in the software hierarchy than a replicating logical device (RLDev) layer at which the replicated record is recovered;
• identifying a location of the corrupted copy of the replicated data item using an offset, wherein the offset is part of a log entry;
• reconciling the replicated record at the RLDev layer by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record if a determination is made that a failure occurred during the operation;
• maintaining the log at the ADS layer, the log storing a plurality of entries associated with changes to the plurality of replicated records for which multiple copies are maintained at the RLDev layer;
• performing a second scan of the log at the ADS layer to recover data associated with the ADS layer;
• and
• recovering the data associated with the ADS layer based on the second scan of the log.

A computer-readable storage medium comprising computer instructions executable by a computer for:

• determining whether a log contains one or more entries indicating a corruption in a replicated data item, wherein the log is maintained at an application layer that is higher in a software hierarchy than a replicating logical device (RLDev) layer at which the replicated data item is recovered, and wherein determining whether the log contains one or more entries indicating a corruption in a replicated data item comprises performing a first scan of the log at the application layer and inferring an offset to determine whether an entry indicates a corruption in a replicated data item, wherein the one or more entries further indicate a desired value of the replicated data item;
• recovering the replicated data item if the log contains any such entries by comparing a plurality of copies of the replicated data item to identify a corrupted copy of the replicated data item and an uncorrupted copy of the replicated data item and replacing the corrupted copy of the replicated data item with the uncorrupted copy of the replicated data item;
• reconciling a replicated record at the RLDev layer in the software hierarchy by comparing the plurality of copies of the replicated record to identify the corrupted copy of the replicated record if a determination is made that a failure occurred during the operation;
• storing, in the log, a plurality of entries associated with changes to a plurality of replicated data items for which multiple copies are maintained at the RLDev layer;
• performing a second scan of the log at the application layer to recover data associated with the application layer;
• and
• recovering the data associated with the application layer based on the second scan of the log.

A computer-readable storage medium comprising computer instructions executable by a computer for:

• recording an entry in a log at an abstract data structure (ADS) layer in a software hierarchy, the entry indicating that an operation was performed on a replicated record, wherein the entry further indicates a desired value of the replicated record;
• recovering the replicated record if the log contains any entries by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record and an uncorrupted copy of the replicated record and replacing the corrupted copy of the replicated record with the uncorrupted copy of the replicated record, wherein the log is maintained at the ADS layer that is higher in the software hierarchy than a replicating logical device (RLDev) layer at which the replicated record is recovered;
• identifying a location of the corrupted copy of the replicated record using an offset, wherein the offset is part of a log entry;
• determining whether a failure occurred during the operation at least in part by performing a first scan of the log at the ADS layer and inferring an offset to determine whether a record is potentially corrupted;
• reconciling the replicated record at the RLDev by comparing a plurality of copies of the replicated record to identify a corrupted copy of the replicated record if a determination is made that a failure occurred during the operation
• storing, in the log, a plurality of entries associated with changes to a plurality of replicated records for which multiple copies are maintained at the RLDev layer;
• performing a second scan of the log at the ADS layer to recover data associated with the ADS layer;
• and
• recovering the data associated with the ADS layer based on the second scan of the log, wherein the ADS layer is a higher layer that is an application layer, and wherein the RLDev layer is a lower layer that is an operating system layer comprising a replicating device module.