Updating the Protocol Endpoint No of Outstanding IOs

How-Tos for VMware Solutions
Audience
Public
Source Type
Documentation

  1. Once the ESXi host is back online check the Device Max Queue Depth setting on the PE to confirm that it is updated. 
    
[root@ESXi-3:~] esxcli storage core device list -p
naa.624a937016d9d3754f8a461800011016
   Display Name: PURE Fibre Channel Disk (naa.624a937016d9d3754f8a461800011016)
...
   Device Max Queue Depth: 128
   No of outstanding IOs with competing worlds: 32
  2. Now the Outstanding IO value can be set on the PE Device, as it's default will still be 32. 
    
[root@ESXi-3:~] esxcli storage core device set -O 128 -d naa.624a937016d9d3754f8a461800011016
[root@ESXi-3:~]

[root@ESXi-3:~] esxcli storage core device list -p
naa.624a937016d9d3754f8a461800011016
   Display Name: PURE Fibre Channel Disk (naa.624a937016d9d3754f8a461800011016)
...
   Device Max Queue Depth: 128
   No of outstanding IOs with competing worlds: 128
  3. Note that even though the ESXi PE has the default number of Outstanding Requests set to 128, when connecting any additional PEs the default will still be set at 32. The device setting will need to be updated manually. We are working with VMware to investigate why the default setting is not getting applied to additional PEs connected.

    With ESXi 6.7 the Cisco fnic driver was replaced with the Cisco nfnic driver for UCS Blades running ESXi. With this update the default queue depth with the nfnic driver is set to 32 and initially there was not a module parameter to adjust the nfnic queue depth. Cisco released an updated version of the nfnic driver (4.0.0.28+) with a configurable module parameter, "lun_queue_depth_per_path". This KB will cover updating the nfnic driver, adjusting the nfnic queue depth and device queue depth.