A fundamental principle of shared storage is multipath configuration. With VMware vSphere, you should know some considerations about the default configuration, as well as what might need to be changed if you know you'll have a path failure.
The built-in multipathing policies and their behavior are described:
- Fixed: This fixed policy will take the first working path and use it continually until it becomes unavailable. This uses the designated path flag, and the I/O traffic will return to the designated path when it returns as a valid path.
- Round Robin: This policy will rotate through working paths with the I/O traffic. This is not link aggregation to the datastore from the ESXi host.
- Most Recently Used (MRU): This multipathing policy is the default and takes the first path discovered to the datastore and holds that until it is unavailable again.
If any storage systems have installed their own plug-in or multipathing policy, the systems may enumerate in the datastore's path management drop-down.
In most situations, the MRU multipathing policy is fine, but there are reasons (such as maintenance and utilization) to change the multipathing policy. For a fibre channel storage network, if a single switch of a pair would need to go offline for critical maintenance, it may be a good idea to set the path policy to the other switch temporarily by using the Fixed policy and setting a preferred path. If you wanted to utilize all paths of a system, this would be done with the Round Robin multipathing policy. (With Round Robin, all paths are used for the storage I/O.)Within each VMFS datastore, a multipathing policy can be set. Figure A shows a multipathing policy being changed on a datastore. Figure A
Click the image to enlarge.
If you have never changed a policy setting, it may be worth doing. In my practice, I'll put a host in maintenance mode to make a change like that. You technically can do it on the fly, but I prefer not to take a chance with a live workload.
For more information on VMware vSphere multipathing, read the KB article 1011340.
Rick Vanover is a software strategy specialist for Veeam Software, based in Columbus, Ohio. Rick has years of IT experience and focuses on virtualization, Windows-based server administration, and system hardware.