About Mirroring and High Availability
A mirror is a virtual volume comprised of two volumes, one from each storage server in a partnership. A mirror provides an aspect of fault tolerance in your storage server configuration. A mirror provides data availability by replicating the data written to a virtual volume on one storage server to another virtual volume on a different storage server.
Mirroring is the process of continuously updating data in real-time to maintain two identical copies of data on separate storage devices. Mirroring guards against data loss. To keep a remote copy of data that is consistent with the primary image, SANmelody software uses “synchronous” mirroring over the SAN, in which the write acknowledgement is not returned to the requesting server until the data has been written to both storage servers. The mirrored virtual volumes stay in “lockstep.” Mirroring is generally viable with separations up to 100 kilometers. Synchronous mirroring typically takes place over dedicated high-throughput routes. The bandwidth (or throughput) of the channels used for I/O writes needs to be sufficient so as not to cause undue latency which would affect the responsiveness of servers.
A standard mirror has a primary and secondary volume; these correspond to the storage servers that originated the volumes. The primary volume is from the primary server, which is based on the storage server you selected when creating the virtual volume. The other storage server by default becomes the secondary. A standard mirror uses a single path. The standard mirror protects against a disk drive failure on the primary. Should a primary drive fail, all virtual volumes associated with that drive that have been mirrored to the partner, are now accessed indirectly through the secondary using the mirrored image. The application server is not aware of this action.
In addition to a standard mirror configuration (in which data in one virtual volume is copied to another virtual volume), mirrors can be configured for high availability in different manners, including:
-
Alternate Path (AP)
-
3rd Party AP/MPIO (3PAP)
-
Cluster Path (CP)
Multi-path mirrors (AP, 3PAP, and CP), on the other hand, provides additional protection against storage server outages (planned or unplanned). The application server multi-pathing driver would recognize that its primary path is unable to complete I/O requests and fails over to the alternate path connected to the secondary. If you have multi-path mirrors, they provide High Availability (HA) to the data another level of data protection. HA mirrors support two different mirrored relationships between virtual volumes.
These configurations allow interruption-free access to data in the event of a disk or storage server failure. All of these product features operate independently of the hosts who generate the data stream and the underlying storage devices where their data is housed. They allow central and secure, remote mirroring services in a uniform manner, despite variations in operating systems and disk arrays.
High availability (HA), a system functioning with greater reliability than its individual components are capable, is most often achieved through fault tolerance. Fault tolerance in disk subsystems, is often achieved by including redundant instances of components whose failure would make the system inoperable, coupled with facilities that allow the redundant components to assume the function of the failed ones. Fault tolerant features include mirroring, RAID-5 volumes, and server clusters.
Alternate Pathing (AP) is an extension of mirrors handled by collaborating storage servers. It allows an application server interruption-free access to the virtual volume in the event of a component-volume failure or even the failure of a storage server. AP is able to redirect I/Os down a secondary path to the virtual volume when it detects a fault anywhere in the primary path to its primary volume.
Multi Path I/O (MPIO) is a high availability software feature that provides path redundancy to multi-path storage devices. More than two paths to virtual volumes are supported. DataCore’s support of MPIO simulates multi-pathing by means of multiple mappings of the same virtual volume. Refer to About Alternate Pathing and MPIO for more information.
With Cluster Path (CP), the virtual volume is presented to two or more application servers; whereas with AP the virtual volume is presented to a single application server (except in the case of Cluster with AP). Additionally, the storage servers allow simultaneous access to the virtual volume via each of the paths. It is the responsibility of the cluster software running on the application servers to coordinate access to the shared virtual volume. The CP option is used in conjunction with cluster-enabled application servers, such as HACMP or Windows Clusters. Refer to About Cluster Path for more information.
If there are multiple paths to a target, to provide mirror path failover, enable the Redundant Mirror Paths feature. Refer to Specifying the Mirror Path.
IMPORTANT: Mirror path configurations, such as using specific mirror paths or redundant mirror paths, must be configured before creating a mirror. After a mirror configuration exists for a virtual volume, mirror path settings cannot be changed. Refer to Specifying the Mirror Path to configure before creating mirrors.
Related Topics:
Mirror Discovery and Recovery Behavior
Create a Virtual Volume Mirror