Maintaining data consistency in mirrored cluster storage systems with write-back cache
View Patent ↗Techniques for maintaining mirrored storage cluster data consistency on systems with write back cache can employ write-intent logging and periodic log clearing. The techniques can track any outstanding I/Os, data in caches, and data that has gone out of sync between mirrored nodes due to link failures. The techniques can ensure that a power failure on any of the storage nodes does not result in inconsistent data among the storage nodes. The techniques can use gate bitmaps to log outstanding I/Os. The gate bitmaps may be periodically cleared. The clearing period may be selected as to allow enough time for all caches to be flushed and for all synchronizing I/Os to complete. Efficient queuing of I/O requests in conjunction with the efficient operation of multiple gate bitmaps can reduce performance impacts of the gating or logging operations.
1. A method for maintaining mirrored data consistency between a primary networked storage node and a secondary networked storage node using write-intent logging, the method comprising:
establishing a delay period;
logging, on the primary networked storage node, an intent to write data, comprising setting a bit in a gate bitmap, the bit corresponding to a storage area within the primary networked storage node where the data is to be written;
writing the data to the primary networked storage node after logging the intent to write data;
relaying a request to the secondary networked storage node to minor the data after logging the intent to write data; and
after the delay period completes, clearing the bit in the gate bitmap to protect an underlying write-back cache, wherein any entries of the write-back cache made before the delay period began are flushed to a disk prior to the delay period completing.
2. The method of claim 1 , wherein logging further comprises storing the gate bitmap to the disk.
3. The method of claim 1 , wherein the delay period is longer than a specified time for the underlying write-back cache to flush.
4. The method of claim 1 , wherein the delay period is longer than a specified time for relaying the request to the secondary networked storage node.
5. The method of claim 1 , further comprising providing a delta tab in the primary networked storage node to temporarily store a difference between data on the primary networked storage node and data on the secondary networked storage node.
6. A computer storage medium having computer-executable instructions stored thereon which, when executed by a computer system, cause the computer system to:
establish a delay period;
receive an input/output (I/O) request;
record the I/O request in a log comprising a gate bitmap, bits of the gate bitmap corresponding to respective storage areas of a disk;
store the log to the disk;
execute the I/O request after the log has been stored to disk;
relay the I/O request to a mirrored computer system for execution;
wait for the delay period to complete; and
clear the log after the wait by clearing the bits of the gate bitmap to protect an underlying write-back cache, wherein any entries of the write-back cache made before the delay period began are flushed prior to the delay period completing.
7. The computer storage medium of claim 6 , wherein the delay period is longer than a specified time for flushing the underlying write-back cache.
8. The computer storage medium of claim 6 , wherein the delay period is longer than a specified time for relaying the I/O request to the mirrored computer system.
9. The computer storage medium of claim 6 , wherein storing the log to the disk comprises flushing a disk cache.
10. The computer storage medium of claim 6 , further comprising providing a delta tab to temporarily accumulate differences between data on the computer system and data on the mirrored computer system.
11. The computer storage medium of claim 6 , wherein relaying the I/O request to the mirrored computer system comprises logging the I/O request in a delta tab until the relayed I/O request is completed.
12. The computer storage medium of claim 10 , wherein clearing the log excludes clearing entries pending in the delta tab.
13. A method for maintaining mirrored data consistency between a primary networked storage node and a secondary networked storage node using write-intent logging, the method comprising:
establishing a delay period;
receiving an input/output (I/O) request at the primary networked storage node, comprising:
evaluating a log entry within a flush gate bitmap associated with an area of a disk targeted by the I/O request;
evaluating a log entry within a committed gate bitmap associated with an area of the disk targeted by the I/O request;
placing the I/O request into a wait queue if the log entry within the flush gate bitmap is not set;
placing the I/O request into a hold queue if the log entry within the flush gate bitmap is set and the log entry within the committed gate bitmap is not set;
executing the I/O request if the log entry within the flush gate bitmap is set and the log entry within the committed gate bitmap is set; and
processing the wait queue and the hold queue;
logging the I/O request to a log at the primary networked storage node;
storing the log to a disk associated with the primary networked storage node;
executing the I/O request at the primary networked storage node;
relaying the I/O request to the secondary networked storage node for execution;
waiting for the established delay period; and
clearing the log after waiting.
14. The method of claim 13 , wherein the log comprises a gate bitmap, bits of the gate bitmap corresponding to respective storage areas of the disk.
15. The method of claim 13 , wherein processing the wait queue and the hold queue comprises:
retrieving the I/O request from the wait queue;
setting the log entry associated with an area of the disk targeted by the I/O request retrieved from the wait queue; and
placing the I/O request retrieved from the wait queue into the hold queue after setting the log entry.
16. The method of claim 13 , wherein processing the wait queue and the hold queue comprises retrieving the I/O request from the hold queue prior to executing the I/O request at the primary networked storage node.