Load balancing on a storage server is a matter of balancing the loads to the system drives (SDs) on the storage systems to which the storage server is connected. A logical unit (LU), known to the server as an SD, is a piece of disk space managed by the block storage, spread across several physical disks.
The server routes FC traffic to individual SDs over a single FC path, distributing the load across two FC switches and, when possible, across dual active/active or multi-port RAID controllers.
Following the failure of a preferred path, disk I/O is redistributed among other (non-optimal) paths. When the server detects reactivation of the preferred FC path, it once again redistributes disk I/O to use the preferred FC path.
Default load balancing (load balancing automatically performed by the storage server) is performed based on the following criteria:
- “Load” is defined as the number of open SDs, regardless of the level of I/O on each SD. An SD is open if it is in a span that has a file system that is mounted or is being formatted, checked or fixed. SDs count towards load at the target if they are open on at least one cluster node; the number of nodes (normally all nodes in a cluster, after boot) is not considered.
- Balancing load on RAID controller target ports takes precedence over balancing load on server FC host ports.
- Balancing load among a system’s RAID controllers takes precedence over balancing among ports on those controllers.
- In a cluster, choice of RAID controller target port is coordinated between cluster nodes, so that I/O requests for a given SD do not simultaneously go to multiple target ports on the same RAID controller.
You can manually configure load distribution from the CLI (overriding the default load balancing performed by the server), using the sdpath command. When manually configuring load balancing using the using the sdpath command:
- You can configure a preferred server host port and/or a RAID controller target port for an SD. If both are set, the RAID controller target port preference takes precedence over the server host port preference. When a specified port preference cannot be satisfied, port selection falls back to automatic selection.
- For the SDs visible on the same target port of a RAID controller, you should either set a preferred RAID controller target port for all SDs or for none of the SDs. Setting the preferred RAID controller target port for only some of the SDs visible on any given RAID controller target port may create a situation where load distribution is suboptimal.
Manually setting a preferred path is not necessary or recommended.
The sdpath command can also be used to query the current FC path being used to communicate with each SD. For more information on the sdpath command, enter man sdpath command.
To see information about the preferred path, navigate to , then select the SD and click details to display the System Drive Details page. If available, the FC Path section provides information about the path, port, and controller.