Multi-site block level write consistency
View Patent ↗Maintaining write consistency between at least two mirrored storage devices includes receiving a write request for a particular block, obtaining a reservation for the particular block, waiting for the particular block to become available in response to not being able to successfully obtain the reservation, writing the particular block in response to successfully obtaining the reservation, and writing the particular block at a remote one of the storage devices following writing the particular block at the particular one of the storage devices. If a retryable indicator is received, the system waits a predetermined amount of time and releases the reservation. The system responds with a success indicator when writing the particular block at a remote one of the storage devices succeeds and responds with a failure indicator when writing the particular block at a remote one of the storage devices fails and the retryable indicator is not received.
1. A method of maintaining write consistency between at least two mirrored storage devices, comprising:
receiving a write request for a particular block at a particular one of the storage devices;
attempting to obtain a reservation for the particular block at the particular one of the storage devices;
waiting for the particular block at the particular one of the storage devices to become available in response to not being able to successfully obtain the reservation;
writing the particular block at the particular one of the storage devices in response to successfully obtaining the reservation;
attempt to write the particular block at a remote one of the storage devices following writing the particular block at the particular one of the storage devices;
determining if a retryable indicator is received in response to writing the particular block at a remote one of the storage devices failing;
waiting a predetermined amount of time and releasing the reservation in response to writing the particular block at a remote one of the storage devices failing and receiving the retryable indicator;
responding to the write request with a success indicator and releasing the reservation in response to writing the particular block at a remote one of the storage devices succeeding; and
responding to the write request with a failure indicator and releasing the reservation in response to writing the particular block at a remote one of the storage devices failing and not receiving the retryable indicator.
2. A method, according to claim 1 , wherein the predetermined amount of time is two hundred milliseconds.
3. A method, according to claim 1 , wherein the remote one of the storage device returns the retryable indicator in response to the remote one of the storage device being deemed a primary storage device.
4. A method, according to claim 1 , wherein the at least two mirrored storage devices are coupled to a switch.
5. A method, according to claim 4 , wherein the switch is coupled to a plurality of host devices that access the at least two storage devices.
6. A method, according to claim 4 , wherein the switch causes the at least two mirrored storage devices to appear as a single storage device.
7. A method, according to claim 1 , wherein following waiting for the particular block at the particular one of the storage devices to become available in response to not being able to successfully obtain the reservation, the I/O times out and fails.
8. A method, according to claim 1 , wherein the blocks are a fixed length of 512 bytes.
9. A non-transitory computer-readable medium containing software that maintains write consistency between at least two mirrored storage devices, the software comprising:
executable code that receives a write request for a particular block at a particular one of the storage devices;
executable code that attempts to obtain a reservation for the particular block at the particular one of the storage devices;
executable code that waits for the particular block at the particular one of the storage devices to become available in response to not being able to successfully obtain the reservation;
executable code that writes the particular block at the particular one of the storage devices in response to successfully obtaining the reservation;
executable code that attempts to write the particular block at a remote one of the storage devices following writing the particular block at the particular one of the storage devices;
executable code that determines if a retryable indicator is received in response to writing the particular block at a remote one of the storage devices failing;
executable code that waits a predetermined amount of time and releases the reservation in response to writing the particular block at a remote one of the storage devices failing and receiving the retryable indicator;
executable code that responds to the write request with a success indicator and releases the reservation in response to writing the particular block at a remote one of the storage devices succeeding; and
executable code that responds to the write request with a failure indicator and releases the reservation in response to writing the particular block at a remote one of the storage devices failing and not receiving the retryable indicator.
10. A non-transitory computer-readable medium, according to claim 9 , wherein the predetermined amount of time is two hundred milliseconds.
11. A non-transitory computer-readable medium, according to claim 9 , wherein the remote one of the storage device returns the retryable indicator in response to the remote one of the storage device being deemed a primary storage device.
12. A non-transitory computer-readable medium, according to claim 9 , wherein the at least two mirrored storage devices are coupled to a switch.
13. A non-transitory computer-readable medium, according to claim 12 , wherein the switch is coupled to a plurality of host devices that access the at least two storage devices.
14. A non-transitory computer-readable medium, according to claim 12 , wherein the switch causes the at least two mirrored storage devices to appear as a single storage device.
15. A non-transitory computer-readable medium, according to claim 9 , wherein following waiting for the particular block at the particular one of the storage devices to become available in response to not being able to successfully obtain the reservation, the I/O times out and fails.
16. A non-transitory computer-readable medium, according to claim 9 , wherein the blocks are a fixed length of 512 bytes.