A quick update for the latest VSI release 10.1. Most of the new features are for PowerStore but one of my requests for PowerMax made it into the release. I asked development to add the ability to set host best practices for NVMeoF devices. This is similar to the best practices for NMP, except that NMP does not support NVMeoF devices rather HPP, or the High Performance Plug-In is used. With NMP, we recommend using Round Robin and switching paths every one I/O or iops=1. With HPP, we have the ability to use latency instead, which allows VMware to choose the path based on which one is performing the fastest. Normally, we recommend this for SRDF/Metro devices since there might be a significant difference in latency between paths, but since VSI does not know whether replication is in use, and since latency has been demonstrated to perform as well as iops=1 for non-SRDF/Metro devices, I recommended to the developers we set latency for all NVMeoF devices.
There are two best practices we can set, one is on existing NVMeoF devices, and the other is to create a rule so that all future NVMeoF devices get the same setting. These are mirrors of the NMP settings. Since VSI only support NVMe/TCP and not FC-NVMe, these settings only apply to the former.
Although PowerFlex also supports NVMeoF, both the LUN setting and rule will only impact PowerMax devices.
For the LUN setting, navigate to the Dell VSI menu and select Apply Host Best Practices.
Select the PowerMax array from the drop-down list to reveal the two sections for setting best practices:
In the bottom section, expand the LUN Settings and check the box for Set HPP Path Selection Scheme to LB-Latency. Then hit SAVE. This will change the selection for all existing NVMe/TCP devices.
Note the column Host Reboot Required is set to Yes. This actually is not a requirement, since this will be done on-the-fly. Here is what my pre-device pathing appears like, with LB-RR set.
After saving the LUN setting, here is the result. Note that after changing the LUN setting it will be necessary to run a refresh in the client to see the change, however the CLI will show it immediately (second image):
The rule is created in much the same way. When you select the PowerMax from the drop-down, the rule section is at the top. Check the box for the HPP rule and select SAVE. Note the column Host Reboot Required is also set to Yes like the LUN, but as long as the rule is loaded (i.e., there is a “runtime” class as well as “file”), new devices will receive the LB-Latency setting.
Rules are not available through the GUI, so use the CLI to check on the creation. There will be two rules, one for each array model:
In addition to the host best practice addition, there were a couple bugs fixed for PowerMax in this release. They are:
- During NFS datastore creation, only a single NAS server is displayed.
- Provisioning PowerMax RDM/Datastore with MaxTotalCapacity User limit does not work.