In a campus cluster or remote mirror configuration, the hosts and storage of a cluster are allocated between two or more sites. These sites are typically connected through a redundant high-capacity network or Fibre Channel that provides access to the storage and communication between the cluster nodes.
If a disk group is configured with storage at multiple sites and if inter-site communication is disrupted, then a serious split-brain condition may occur. This happens when each site continues to update the local disk group configuration copies without being aware of site disruption. For service(s) to come up on a site while other sites are down, a complete (at least one complete plex for each volume) copy of data is needed. Currently, there is no mechanism to ensure that all volumes have a complete data plex at each site. Data plex keeps getting changed when a volume is resized, a disk is relocated, or when a new volume is added.
Site-aware allocation feature enables applications and services to function properly at a site when other sites become inaccessible. It means that even during site disruption at least one complete plex of a volume is available at each site. Such type of allocation is known as site-based allocation. Users can specify sites when creating volumes or mirrors, and site boundary limits are maintained for operations like volume grow, subdisk move, and disk relocation. Site boundary limit is crossed when a plex is not totally within a site and allocation of the plex crosses the available site boundary.
Site-aware allocation facilitates following types of site-based allocation:
The following table describes the terms that are used in the context of site-aware allocation.
Table: Site-aware allocation and related terminology
|
Terminology
|
Description |
|
Site
|
Logical representation of a set of hosts and set of arrays or enclosures. |
|
Site Separated
|
Storage for a volume can be taken from a site (s) that are specified during volume creation. Storage from multiple sites is supported for such type of allocation.
Storage for a volume is allocated so that each plex of the volume resides completely on the same site, i.e., if a Site Separated volume has two plexes on two sites and , each plex resides completely on a separate site.
Volume Resize, relocation, relayout, or any such operation keeps each plex on its own site.
Multiple plexes can reside on the same site
|
|
Site Confined
|
Storage for a volume can be taken from only a site that is specified while creating a volume. Multiple sites cannot be allocated for such type of volume allocation. The volume resides entirely on the same site.
Resize, relocation, relayout, or any such operation only uses storage from the same site.
|
|
Siteless
|
Refers to a volume that is not tagged with any site information or site properties. By default, all volumes are Siteless.
Note: |
On upgrading to SFW 7.3 from any previous release versions (which did not have the Siteless option), all volume types are "Siteless" by default. You can manually change the property of volumes after upgrading to either Site Confined or Site Separated provided that conditions like "volume need to be entirely on the same site" or "each plex of the volume resides entirely on a site" are met. |
|
|
Site boundary
|
Site boundary limit is said to be crossed when a plex is not totally within a site and allocation of the plex crosses the available site boundary.
Automatic operations like hot relocation does not adhere to site boundary restrictions. Storage that is configured with such auto operations become Siteless once site boundary limit is crossed.
When a storage becomes siteless, user is notified and Event Viewer logs displays logs to verify the same.
|
