Zone storage—quickly returning to a state of consistency following an unexpected event

• US 10,552,341 B2
• Filed: 02/17/2017
• Issued: 02/04/2020
• Est. Priority Date: 02/17/2017
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Baptist, Andrew D., Kaczmarek, Joseph M., Motwani, Manish, Viraraghavan, Praveen, Volvovski, Ilya
• Primary Examiner(s): Rones, Charles
• Assistant Examiner(s): Krieger, Jonah C
• Application Number:
  US15/435,385
• Publication Number:
  US 20180239701A1
• Time in Patent Office: 1,082 Days
• Field of Search: 711135
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery
• CPC Class Codes: G06F 12/0813 with a network or matrix configuration
• CPC Class Codes: G06F 12/0868 Data transfer between cache memory and other subsystems, e.g. storage devices or host systems
• CPC Class Codes: G06F 12/12 Replacement control
• CPC Class Codes: G06F 12/121 using replacement algorithms
• CPC Class Codes: G06F 2212/1021 Hit rate improvement
• CPC Class Codes: G06F 2212/1032 Reliability improvement, data loss prevention, degraded operation etc
• CPC Class Codes: G06F 2212/154 Networked environment

Abstract

Systems and Methods for data storage in a distributed storage network are disclosed. Unexpected errors can adversely affect consistency of both the content of a write (including the slice data), and the synchronicity between the written slices and metadata structures. To maintain consistency between these data structures, a sequencing of the order of writes and flushes to the memory devices for the different data structures may be enforced as follows: First: Slice content data is first written to the volatile memory (e.g. a cache memory) of a DS unit; Second: the Slice content data stored in volatile memory is “flushed” to a non-volatile bin (which bin is associated with a group of physical memory blocks in non-volatile memory); Third: after the flush of the slice content data to the bin (i.e. data is durable on the media device): metadata relating to the data is written.

Claims

1. A method of storing data comprising:

writing error encoded slice content data to a volatile memory;

flushing the error encoded slice content data to a bin in a non-volatile memory after determining the error encoded slice content data was completely written to the volatile memory, wherein the bin is associated with a group of physical memory blocks in the non-volatile memory;

discovering a recoverable inconsistency in the bin, where the bin contains one or more of the encoded slice content data that is not referenced by either a metadata catalog or a metadata journal;

initiating a recovery process to recover missing metadata by performing a scan of metadata for a safe point marker, wherein the discovering a recoverable inconsistency in the bin occurs from the safe point marker to a write pointer or last append point of the bin;

writing a first missing metadata entry including missing metadata corresponding to the error encoded slice content data, after flushing the error encoded data slice content data to the bin, wherein the missing metadata is determined by recognition of bin entries in bin files that are not in the metadata journal including pointers to the bins or not in the metadata catalog, including an encoded slice name and size information; and

writing a second missing metadata entry including the missing metadata corresponding to the bin after flushing the error encoded slice content data to the bin.

2. The method of claim 1, wherein the error encoded slice content data is flushed to the bin in the non-volatile memory using a pattern of append only writes.

3. The method of claim 1, wherein the step of writing missing metadata corresponding to the error encoded slice content data and the step of writing missing metadata corresponding to the bin are performed sequentially.

4. The method of claim 1, wherein the step of writing missing metadata corresponding to the error encoded slice content data and the step of writing missing metadata corresponding to the bin are performed in parallel.

5. The method of claim 1, wherein the error encoded slice content data is written to the volatile memory using an atomic write structure.

6. The method of claim 5, wherein the atomic write structure includes one or more of:

defined headers, length indicators, defined footers or data checksums.

7. The method of claim 1, further comprising the steps of:

writing second error encoded slice content data to the volatile memory;

determining that the error encoded second slice content data was incompletely written to the volatile memory; and

marking an area of the volatile memory associated with the error encoded second slice content data as invalid.

8. The method of claim 1, further comprising updating the safe point marker after the step of writing the missing metadata corresponding to the error encoded slice content data and the bin in the non-volatile memory after flushing the error encoded slice content data to the bin.

9. A dispersed storage unit for use in a dispersed storage network, the dispersed storage unit comprising:

a communications interface;

a volatile memory;

a non-volatile memory; and

a computer processor;

wherein the non-volatile memory includes instructions for causing the computer processor to;

write error encoded slice content data to the volatile memory;

flush the error encoded slice content data to a bin in the non-volatile memory after a determination that the error encoded slice content data was completely written to the volatile memory, wherein the bin is associated with a group of physical memory blocks in the non-volatile memory;

discover a recoverable inconsistency in the bin, where the bin contains one or more of the slice content data that is not referenced by either a metadata catalog or a metadata journal;

initiate a recovery process to recover missing metadata by performing a scan of metadata for a safe point marker, wherein the discover a recoverable inconsistency in the bin occurs from the safe point marker to a write pointer or last append point of the bin;

write a first missing metadata entry including missing metadata corresponding to the error encoded slice content data after the error encoded slice content data has been flushed to the bin, wherein the missing metadata is determined by recognition of bin entries in bin files that are not in the metadata journal including pointers to the bins or not in the metadata catalog, including an encoded slice name and size information; and

write a second missing metadata entry including the missing metadata corresponding to the bin after the error encoded slice content data has been flushed to the bin.

10. The dispersed storage unit of claim 9, wherein the non-volatile memory further comprises instructions for causing the computer processor to flush the error encoded slice content data to the bin in the non-volatile memory using a pattern of append only writes.

11. The dispersed storage unit of claim 9, wherein the non-volatile memory further comprises instructions for causing the computer processor to write the missing metadata corresponding to the error encoded slice content data and instructions for causing the computer processor to write the missing metadata corresponding to the bin, sequentially.

12. The dispersed storage unit of claim 9, wherein the non-volatile memory further comprises instructions for causing the computer processor to write the missing metadata corresponding to the error encoded slice content data and instructions for causing the computer processor to write the missing metadata corresponding to the bin, in parallel.

13. The dispersed storage unit of claim 9, wherein the non-volatile memory further comprises instructions for causing the computer processor to write the error encoded slice content data to the volatile memory using an atomic write structure.

14. The dispersed storage unit of claim 13, wherein the atomic write structure includes one or more of:

defined headers, length indicators, defined footers or data checksums.

15. The dispersed storage unit of claim 9, wherein the non-volatile memory further comprises instructions for causing the computer processor to:

write second error encoded slice content data to the volatile memory;

determine that the second error encoded second slice content data was incompletely written to the volatile memory; and

mark an area of the volatile memory associated with the second error encoded slice content data as invalid.

16. A dispersed storage network comprising:

a plurality of dispersed storage units;

a dispersed storage processing unit

wherein a first dispersed storage unit of the plurality of dispersed storage units includes;

a volatile memory;

a non-volatile memory; and

a computer processor;

wherein the non-volatile memory includes instructions for causing the computer processor to;

write error encoded slice content data to the volatile memory;

flush the error encoded slice content data to a bin in the non-volatile memory after a determination that the error encoded slice content data was completely written to the volatile memory, wherein the bin is associated with a group of physical memory blocks in the non-volatile memory;

discover a recoverable inconsistency in the bin, where the bin contains one or more of the slice content data that is not referenced by either a metadata catalog or a metadata journal;

initiate a recovery process to recover missing metadata by performing a scan of metadata for a safe point marker, wherein the discover a recoverable inconsistency in the bin occurs from the safe point marker to a write pointer or last append point of the bin;

write a first missing metadata entry including missing metadata corresponding to the error encoded slice content data after the error encoded slice content data has been flushed to the bin, wherein the missing metadata is determined by recognition of bin entries in bin files that are not in the metadata journal including pointers to the bins or not in the metadata catalog, including an encoded slice name and size information; and

write a second missing metadata entry including missing metadata corresponding to the bin after the error encoded slice content data has been flushed to the bin.

17. The dispersed storage network of claim 16, wherein the non-volatile memory further comprises instructions for causing the computer processor to flush the error encoded slice content data to the bin in the non-volatile memory using a pattern of append only writes.