Storage system and management method for storage system
A storage system includes a storage apparatus having a plurality of physical drives and a controller. In a case where a physical failure occurs in a physical drive, the controller additionally installs a drive provided from a cloud, and maintains a redundant array of inexpensive disks (RAID) configuration before the occurrence of the physical failure in the physical drive using the physical drives excluding the physical drive where the physical failure has occurred and the additionally installed drive provided from the cloud.
1 . A storage system for restoring full redundancy in a redundant array of independent disks (RAID) while avoiding permanent allocation of on-premises spare disks, the storage system comprising:
a plurality of on-premises memory apparatuses interconnected by a local storage fabric;
a hardware RAID controller including at least one processor, a non-transitory memory that stores firmware instructions, and a wide-area-network (WAN) interface; and
a cloud-access module authenticated to provision, over the WAN, an ephemeral block-level memory area hosted in a public-cloud availability zone physically remote from the on-premises memory apparatuses;
wherein the firmware instructions, when executed by the processor, cause the hardware RAID controller to:
monitor the plurality of on-premises memory apparatuses for a physical-failure event;
detect that a particular memory apparatus among the on-premises memory apparatuses has experienced the physical-failure event;
provision, in response to the physical-failure event and only after the event, a remotely hosted memory area in the ephemeral block-level memory area with a capacity at least equal to a capacity of the particular memory apparatus that failed;
incorporate the remotely hosted memory area into the RAID together with the on-premises memory apparatuses other than the particular memory apparatus such that a RAID level that existed immediately before the physical-failure event is re-established;
service host read and write operations during re-establishment of the RAID by committing data destined for the remotely hosted memory area to a non-volatile write-back cache and flushing the data asynchronously over the WAN; and
after the particular memory apparatus is replaced and the RAID is re-mirrored locally, copy data or parity stored in the remotely hosted memory area back into a replacement memory apparatus and detach the remotely hosted memory area.
2 . The storage system of claim 1 , wherein, before the physical-failure event, the RAID is configured by the plurality of on-premises memory apparatuses together with the remotely hosted memory area, and only the remotely hosted memory area includes a parity-storage region that stores parity calculated from data written to the RAID.
3 . The storage system of claim 2 , wherein the firmware instructions further cause the hardware RAID controller, when the physical-failure event occurs in one of the on-premises memory apparatuses, to substitute the remotely hosted memory area for the particular memory apparatus and to reconstruct, in the remotely hosted memory area, data formerly stored in the particular memory apparatus based on
data stored in the remaining on-premises memory apparatuses and in the remotely hosted memory area.
4 . The storage system of claim 1 , wherein,
before the physical-failure event, the RAID is configured exclusively by the plurality of on-premises memory apparatuses, and the plurality of on-premises memory apparatuses includes a parity storage memory apparatus that stores parity calculated from data written to the RAID.
5 . The storage system according to claim 4 , wherein
the firmware instructions further cause the hardware RAID controller, when the particular memory apparatus experiences the physical-failure event, to substitute the remotely hosted memory area for the particular memory apparatus and to reconstruct, in the remotely hosted memory area, the parity
formerly stored in the particular parity stored in the memory apparatus based on data stored in the remaining on-premises memory apparatuses.
6 . The storage system of claim 1 , wherein,
after the physical-failure event has been remedied by installing a replacement on-premises memory apparatus, the firmware instructions cause the hardware RAID controller to detach and delete the remotely hosted memory area.
7 . The storage system of claim 1 , wherein,
while the RAID includes the remotely hosted memory area, the firmware instructions cause the hardware RAID controller to calculate data to be written to the provisioned remotely hosted memory area using a write-back cache method based on data slated for storage in the on-premises memory apparatuses and in the remotely hosted memory area.
8 . A computer-implemented method for restoring full redundancy in a redundant array of independent disks (RAID) of a storage system while avoiding permanent allocation of on-premises spare disks, the method comprising:
monitoring a plurality of on-premises memory apparatuses for a physical-failure event;
detecting that a particular one of the plurality of memory apparatuses has experienced the physical-failure event;
provisioning, in response to and only after the physical-failure event, an ephemeral block-level memory area over a wide-area network (WAN), the memory area being hosted in a public-cloud availability zone physically remote from the on-premises memory apparatuses and having a capacity at least equal to that of the particular memory apparatus;
incorporating the provisioned block-level memory area into the RAID together with the on-premises memory apparatuses
other than the particular memory apparatus, thereby re-establishing the RAID level that existed immediately before the physical-failure event;
servicing host read and write operations during re-establishment of the RAID by committing data destined for the provisioned block-level memory area to a non-volatile write-back cache and flushing the data asynchronously over the WAN;
after the particular memory apparatus is replaced and the RAID is re-mirrored locally, copying data or parity stored in the provisioned block-level memory area back into the replacement memory apparatus; and
detaching the provisioned block-level memory area after the copying is complete.
9 . The storage system of claim 4 , wherein, when the physical-failure event occurs in a data-storage memory apparatus other than the parity-storage memory apparatus, the firmware instructions further cause the hardware RAID controller to (i) substitute the remotely hosted memory area additionally installed, and for the particular memory apparatus, (ii) copy the parity stored in the parity-storage memory apparatus into the remotely hosted memory area, and (iii) regenerate, in the parity-storage memory apparatus, data formerly stored in the particular memory apparatus based on data stored in the remaining on-premises memory apparatuses and in the remotely hosted memory area.