Best practices
- Choose a hot-plug capable slot and then insert NVMe device. Wait for power indicator being on, if the platform supports driver power control.
- Issue esxcli hardware PCI list and check if target nvme device is listed. Output is similar to following:
0000:86:00.0
Address: 0000:86:00.0
Segment: 0x0000
Bus: 0x86
Slot: 0x00
Function: 0x0
VMkernel Name: vmhba3
Vendor Name: xxxxxxxxx
Device Name: xxxxxxxx PCIe SSD Controller
Configured Owner: VMkernel
Current Owner: VMkernel
Vendor ID:
. . .
- Check readiness of SCSI devices for the NVMe device: run esxcli storage core device list. Output is similar to following:
(t10.NVMe____Dell_Express_Flash_NVMe_PM1725_1.6TB_SFFB25C007192382500
Display Name: Local NVMe Disk (t10.NVMe____Dell_Express_Flash_NVMe_PM1725_1.6TB_SFFB25C007192382500)
Has Settable Display Name: true
Size: 1526185
Device Type: Direct-Access
Multipath Plugin: NMP
Devfs Path: /vmfs/devices/disks/t10.NVMe____Dell_Express_Flash_NVMe_PM1725_1.6TB_SFFB25C007192382500
Vendor: NVMe
Model: Dell Express Fla
...
- If NVMe device contains VMFS datastores, these datastores will be mounted automatically. Run 'esxcli storage filesystem list' to list all datastores claimed.
Removal
- Make sure no IO is running on the NVME device to be removed before starting hot-plug, no matter orderly or surprise.
- On orderly hot-unplug scenario, user needs to run proprietary steps before unplugging. They are:
- In surprise hot-unplug scenario, user stop IOs running on NVMe device and then plug device out immediately not running the above steps.
- Hot-unplug device. Check that the device is removed from the system device listing:
- Issue "esxcli storage core device list", check that the device is not listed.
- Issue "esxcli storage nmp path list", check that there is no path for this device.
- Plug in the device, check that the device becomes available. Steps are:
- If the device was hot-unplugged orderly, it needs to re-enable device being detected during rescan via command "esxcli storage core claiming autoclaim --enabled=True".
- Issue "esxcli hardware pci list|grep -w <vmhba>", check that new adapter is listed.
- Issue "esxcli storage core device list", check that the device is listed.
- Issue "esxcli storage nmp path list", check that the path to device is listed.
- Steps to remount VMFS volumes:
- Mount VMFS volume: esxcli storage filesystem mount -l <datastore>
- List all VMFS volumes: esxcli storage filesystem list
Note: Do not perform hot-plug frequently, as they might lead to potential problems and corner case issues.