About sites and remote mirrors

In a Remote Mirror configuration (also known as a campus cluster or stretch cluster) the hosts and storage of a cluster that would usually be located in one place, are instead divided between two or more sites. These sites are typically connected via a redundant high-capacity network that provides access to storage and private link communication between the cluster nodes.

Figure: Example of a two-site remote mirror configuration shows a typical two-site remote mirror configuration.

Figure: Example of a two-site remote mirror configuration

Example of a two-site remote mirror configuration

If a disk group is configured across the storage at the sites, and inter-site communication is disrupted, there is a possibility of a serial split brain condition arising if each site continues to update the local disk group configuration copies.

VxVM provides mechanisms for dealing with the serial split brain condition, monitoring the health of a remote mirror, and testing the robustness of the cluster against various types of failure (also known as fire drill).

For applications and services to function correctly at a site when other sites have become inaccessible, at least one complete plex of each volume must be configured at each site (site-based allocation), and the consistency of the data in the plexes at each site must be ensured (site consistency).

By tagging disks with site names, storage can be allocated from the correct location when creating, resizing or relocating a volume, and when changing a volume's layout.

Figure: Site-consistent volume with two plexes at each of two sites shows an example of a site-consistent volume with two plexes configured at each of two sites.

Figure: Site-consistent volume with two plexes at each of two sites

Site-consistent volume with two plexes at each of two sites

The storage for plexes P1 and P2 is allocated storage that is tagged as belonging to site A, and the storage for plexes P3 and P4 is allocated storage that is tagged as belonging to site B.

Although not shown in this figure, DCO log volumes are also mirrored across the sites, and disk group configuration copies are distributed across the sites.

Site consistency means that the data in the plexes for a volume must be consistent at each site. The site consistency of a volume is ensured by detaching a site when its last complete plex fails at that site. If a site fails, all its plexes are detached and the site is said to be detached. If site consistency is not on, only the plex that fails is detached. The remaining volumes and their plexes on that site are not detached.

To enhance read performance, VxVM will service reads from the plexes at the local site where an application is running if the siteread read policy is set on a volume. Writes are written to plexes at all sites.

Figure: Example of a two-site configuration with remote storage only shows a configuration with remote storage only that is also supported.

Figure: Example of a two-site configuration with remote storage only

Example of a two-site configuration with remote storage only

More Information

Handling conflicting configuration copies