NVMe-oF host configuration for RHEL 8.8 with ONTAP

12/19/2023 Contributors

NVMe over Fabrics (NVMe-oF), including NVMe over Fibre Channel (NVMe/FC) and other transports, is supported with Red Hat Enterprise Linux (RHEL) 8.8 with Asymmetric Namespace Access (ANA). In NVMe-oF environments, ANA is the equivalent of ALUA multipathing in iSCSI and FC environments and is implemented with in-kernel NVMe multipath.

The following support is available for the NVMe-oF host configuration for RHEL 8.8 with ONTAP:

Support for NVMe over TCP (NVMe/TCP) in addition to NVMe/FC. The NetApp plug-in in the native nvme-cli package displays ONTAP details for both NVMe/FC and NVMe/TCP namespaces.

For additional details on supported configurations, see the NetApp Interoperability Matrix Tool.

Known limitations

In-kernel NVMe multipath is disabled by default for RHEL 8.8 NVMe-oF hosts. Therefore, you need to enable it manually.
On RHEL 8.8 hosts, NVMe/TCP is a technology preview feature due to open issues.
SAN booting using the NVMe-oF protocol is currently not supported.

Enable in-kernel multipath

You can use the following procedure to enable in-kernal multipath.

Steps

Install RHEL 8.8 on the host server.
After the installation is complete, verify that you are running the specified RHEL 8.8 kernel.
```
# uname -r
```
Example output
```
4.18.0-477.10.1.el8_8.x86_64
```

Install the nvme-cli package:

rpm -qa|grep nvme-cli

Example output

nvme-cli-1.16-7.el8.x86_64

Enable in -kernel NVMe multipath:

# grubby --args=nvme_core.multipath=Y --update-kernel /boot/vmlinuz-4.18.0-477.10.1.el8_8.x86_64

On the host, check the host NQN string at /etc/nvme/hostnqn:

# cat /etc/nvme/hostnqn

Example output

nqn.2014-08.org.nvmexpress:uuid:f6517cae-3133-11e8-bbff-7ed30aef123f

Verify that the hostnqn string matches the hostnqn string for the corresponding subsystem on the ONTAP array:

::> vserver nvme subsystem host show -vserver vs_fcnvme_141

Example output

Vserver     Subsystem       Host NQN
----------- --------------- ----------------------------------------------------------
vs_nvme161 rhel_161_LPe32002    nqn.2014-08.org.nvmexpress:uuid:f6517cae-3133-11e8-bbff-7ed30aef123f

If the host NQN strings do not match, you can use the vserver modify command to update the host NQN string on your corresponding ONTAP NVMe subsystem to match the host NQN string /etc/nvme/hostnqn on the host.

Reboot the host.

If you intend to run both NVMe and SCSI co-existent traffic on the same host, NetApp recommends using the in-kernel NVMe multipath for ONTAP namespaces and dm-multipath for ONTAP LUNs respectively. This means that the ONTAP namespaces should be excluded from dm-multipath to prevent dm-multipath from claiming these namespace devices. This can be done by adding the enable_foreign setting to the /etc/multipath.conf file:

# cat /etc/multipath.conf
defaults {
  enable_foreign  NONE
}

Configure NVMe/FC

You can configure NVMe/FC for Broadcom/Emulex or Marvell/Qlogic adapters.

Broadcom/Emulex

Steps

Verify that you are using the supported adapter model:

# cat /sys/class/scsi_host/host*/modelname

Example output:

LPe32002-M2
LPe32002-M2

# cat /sys/class/scsi_host/host*/modeldesc

Example output:

Emulex LightPulse LPe32002-M2 2-Port 32Gb Fibre Channel Adapter
Emulex LightPulse LPe32002-M2 2-Port 32Gb Fibre Channel Adapter

Verify that you are using the recommended Broadcom lpfc firmware and inbox driver:
```
# cat /sys/class/scsi_host/host*/fwrev
14.0.639.18, sli-4:2:c
14.0.639.18, sli-4:2:c
```
```
# cat /sys/module/lpfc/version
0:14.0.0.18
```
For the most current list of supported adapter driver and firmware versions, see the NetApp Interoperability Matrix Tool.

Verify that lpfc_enable_fc4_type is set to 3:

# cat /sys/module/lpfc/parameters/lpfc_enable_fc4_type
3

Verify that the initiator ports are up and running and that you can see the target LIFs:

# cat /sys/class/fc_host/host*/port_name
0x100000109b1c1204
0x100000109b1c1205

# cat /sys/class/fc_host/host*/port_state
Online
Online

# cat /sys/class/scsi_host/host*/nvme_info NVME Initiator Enabled
XRI Dist lpfc0 Total 6144 IO 5894 ELS 250
NVME LPORT lpfc0 WWPN x10000090fae0ec88 WWNN x20000090fae0ec88 DID x0a1300 ONLINE
NVME RPORT       WWPN x2049d039ea36a105 WWNN x2048d039ea36a105 DID x0a0c0a TARGET DISCSRVC ONLINE
NVME RPORT       WWPN x204bd039ea36a105 WWNN x2048d039ea36a105 DID x0a100a TARGET DISCSRVC ONLINE
NVME Statistics
LS: Xmt 0000000134 Cmpl 0000000134 Abort 00000000
LS XMIT: Err 00000000  CMPL: xb 00000000 Err 00000000
Total FCP Cmpl 000000000825e567 Issue 000000000825d7ed OutIO fffffffffffff286
abort 0000027c noxri 00000000 nondlp 00000a02 qdepth 00000000 wqerr 00000000 err 00000000
FCP CMPL: xb 00000782 Err 000130fa

NVME Initiator Enabled
XRI Dist lpfc1 Total 6144 IO 5894 ELS 250
NVME LPORT lpfc1 WWPN x10000090fae0ec89 WWNN x20000090fae0ec89 DID x0a1200 ONLINE
NVME RPORT       WWPN x204ad039ea36a105 WWNN x2048d039ea36a105 DID x0a080a TARGET DISCSRVC ONLINE
NVME RPORT       WWPN x204cd039ea36a105 WWNN x2048d039ea36a105 DID x0a090a TARGET DISCSRVC ONLINE
NVME Statistics
LS: Xmt 0000000134 Cmpl 0000000134 Abort 00000000
LS XMIT: Err 00000000  CMPL: xb 00000000 Err 00000000
Total FCP Cmpl 000000000826ced5 Issue 000000000826c226 OutIO fffffffffffff351
        abort 0000029d noxri 00000000 nondlp 000008df qdepth 00000000 wqerr 00000000 err 00000000
FCP CMPL: xb 00000821 Err 00012fcd

Marvell/QLogic FC Adapter for NVMe/FC

Steps

The native inbox qla2xxx driver included in the RHEL 8.8 GA kernel has the latest upstream fixes essential for ONTAP support. Verify that you are running the supported adapter driver and firmware versions:
```
# cat /sys/class/fc_host/host*/symbolic_name
```
Example output
```
QLE2772 FW:v9.10.11 DVR:v10.02.07.900-k-debug
QLE2772 FW:v9.10.11 DVR:v10.02.07.900-k-debug
```
Verify that ql2xnvmeenable is set. This enables the Marvell adapter to function as an NVMe/FC initiator:
```
# cat /sys/module/qla2xxx/parameters/ql2xnvmeenable
1
```

Enable 1MB I/O (Optional)

ONTAP reports an MDTS (Max Data Transfer Size) of 8 in the Identify Controller data which means the maximum I/O request size can be up to 1MB. However, to issue I/O requests of size 1 MB for a Broadcom NVMe/FC host, you must increase the lpfc value of the lpfc_sg_seg_cnt parameter to 256 from the default value of 64.

Steps

Set the lpfc_sg_seg_cnt parameter to 256.

# cat /etc/modprobe.d/lpfc.conf
options lpfc lpfc_sg_seg_cnt=256

Run a dracut -f command, and reboot the host.

Verify that lpfc_sg_seg_cnt is 256.

# cat /sys/module/lpfc/parameters/lpfc_sg_seg_cnt
256

This is not applicable to Qlogic NVMe/FC hosts.

Configure NVMe/TCP

NVMe/TCP does not have auto-connect functionality. Therefore, if a path goes down and is not reinstated within the default time out period of 10 minutes, NVMe/TCP cannot automatically reconnect. To prevent a time out, you should set the retry period for failover events to at least 30 minutes.

Steps

Verify that the initiator port can fetch the discovery log page data across the supported NVMe/TCP LIFs:

nvme discover -t tcp -w host-traddr -a traddr

Example output:

# nvme discover -t tcp -w 192.168.111.79 -a 192.168.111.14

Discovery Log Number of Records 8, Generation counter 10
=====Discovery Log Entry 0======
trtype:  tcp
adrfam:  ipv4
subtype: unrecognized
treq:    not specified
portid:  0
trsvcid: 8009
subnqn:  nqn.1992-08.com.netapp:sn.154a5833c78c11ecb069d039ea359e4b:discovery
traddr:  192.168.211.15
sectype: none
=====Discovery Log Entry 1======
trtype:  tcp
adrfam:  ipv4
subtype: unrecognized
treq:    not specified
portid:  1
trsvcid: 8009
subnqn:  nqn.1992-08.com.netapp:sn.154a5833c78c11ecb069d039ea359e4b:discovery
traddr:  192.168.111.15
sectype: none
=====Discovery Log Entry 2======
trtype:  tcp
adrfam:  ipv4
subtype: unrecognized
treq:    not specified
portid:  2
trsvcid: 8009
subnqn:  nqn.1992-08.com.netapp:sn.154a5833c78c11ecb069d039ea359e4b:discovery
traddr:  192.168.211.14
sectype: none
..........

Verify that the other NVMe/TCP initiator-target LIF combinations can successfully fetch discovery log page data:

nvme discover -t tcp -w host-traddr -a traddr

Example output:

# nvme	discovery	-t   tcp    -w	192.168.111.79   -a	192.168.111.14
# nvme	discovery	-t   tcp    -w	192.168.111.79   -a	192.168.111.15
# nvme	discovery	-t   tcp    -w	192.168.211.79   -a	192.168.211.14
# nvme	discovery	-t   tcp    -w	192.168.211.79   -a	192.168.211.15

Run the nvme connect-all command across all the supported NVMe/TCP initiator-target LIFs across the nodes, and set the controller loss timeout period for at least 30 minutes or 1800 seconds:

nvme connect-all -t tcp -w host-traddr -a traddr -l 1800

Example output:

# nvme	connect-all	-t	tcp	-w	192.168.111.79	-a	192.168.111.14	-l	1800
# nvme	connect-all	-t	tcp	-w	192.168.111.79	-a	192.168.111.15	-l	1800
# nvme	connect-all	-t	tcp	-w	192.168.211.79	-a	192.168.211.14	-l	1800
# nvme	connect-all	-t	tcp	-w	192.168.211.79	-a	192.168.211.15	-l	1800

Validate NVMe-oF

You can use the following procedure to validate NVME-oF.

Steps

Verify that the in-kernel NVMe multipath is enabled:
```
# cat /sys/module/nvme_core/parameters/multipath
Y
```
Verify that the appropriate NVMe-oF settings (such as, model set to NetApp ONTAP Controller and load balancing iopolicy set to round-robin) for the respective ONTAP namespaces correctly reflect on the host:
```
# cat /sys/class/nvme-subsystem/nvme-subsys*/model
NetApp ONTAP Controller
NetApp ONTAP Controller
```
```
# cat /sys/class/nvme-subsystem/nvme-subsys*/iopolicy
round-robin
round-robin
```

Verify that the namespaces are created and correctly discovered on the host:

# nvme list

Example output:

Node         SN                   Model
---------------------------------------------------------
/dev/nvme3n1 81Gx7NSiKSQeAAAAAAAB	NetApp ONTAP Controller


Namespace Usage    Format             FW             Rev
-----------------------------------------------------------
1                 21.47 GB / 21.47 GB	4 KiB + 0 B   FFFFFFFF

Verify that the controller state of each path is live and has the correct ANA status:

NVMe/FC

# nvme list-subsys /dev/nvme3n1

Example output:

nvme-subsys3 - NQN=nqn.1992-08.com.netapp:sn.ab4fa6a5ba8b11ecbe3dd039ea359e4b:subsystem.rhel_161_Lpe32002
\
 +- nvme0 fc traddr=nn-0x2048d039ea36a105:pn-0x204cd039ea36a105 host_traddr=nn-0x20000090fae0ec89:pn-0x10000090fae0ec89 live non-optimized
 +- nvme1 fc traddr=nn-0x2048d039ea36a105:pn-0x204ad039ea36a105 host_traddr=nn-0x20000090fae0ec89:pn-0x10000090fae0ec89 live optimized
 +- nvme2 fc traddr=nn-0x2048d039ea36a105:pn-0x204bd039ea36a105 host_traddr=nn-0x20000090fae0ec88:pn-0x10000090fae0ec88 live non-optimized
 +- nvme4 fc traddr=nn-0x2048d039ea36a105:pn-0x2049d039ea36a105 host_traddr=nn-0x20000090fae0ec88:pn-0x10000090fae0ec88 live optimized

NVMe/TCP

# nvme list-subsys /dev/nvme0n1

Example output:

nvme-subsys0 - NQN=nqn.1992-08.com.netapp:sn.154a5833c78c11ecb069d039ea359e4b:subsystem.rhel_tcp_165
\
 +- nvme0 tcp traddr=192.168.111.15 trsvcid=4420 host_traddr=192.168.111.79 live non-optimized
 +- nvme1 tcp traddr=192.168.111.14 trsvcid=4420 host_traddr=192.168.111.79 live optimized
 +- nvme2 tcp traddr=192.168.211.15 trsvcid=4420 host_traddr=192.168.211.79 live non-optimized

Verify that the NetApp plug-in displays the correct values for each ONTAP namespace device:

Column

# nvme netapp ontapdevices -o column

Example output:

Device        Vserver   Namespace Path
----------------------- ------------------------------
/dev/nvme0n1 vs_tcp           /vol/vol1/ns1



NSID       UUID                                   Size
------------------------------------------------------------
1          338d73ce-b5a8-4847-9cc9-b127c75d8855	21.47GB

JSON

# nvme netapp ontapdevices -o json

Example output

{
  "ONTAPdevices" : [
    {
      "Device" : "/dev/nvme0n1",
      "Vserver" : "vs_tcp79",
      "Namespace_Path" : "/vol/vol1/ns1",
      "NSID" : 1,
      "UUID" : "338d73ce-b5a8-4847-9cc9-b127c75d8855",
      "Size" : "21.47GB",
      "LBA_Data_Size" : 4096,
      "Namespace_Size" : 5242880
    },
]

}

Known issues

The NVMe-oF host configuration for RHEL 8.8 with ONTAP release has the following known issues:

NetApp Bug ID	Title	Description	Bugzilla ID
1479047	RHEL 8.8 NVMe-oF hosts create duplicate persistent discovery controllers	On NVMe over Fabrics (NVMe-oF) hosts, you can use the "nvme discover -p" command to create Persistent Discovery Controllers (PDCs). When this command is used, only one PDC should be created per initiator-target combination. However, if you are running Red Hat Enterprise Linux (RHEL) 8.8 on an NVMe-oF host, a duplicate PDC is created each time "nvme discover -p" is executed. This leads to unnecessary usage of resources on both the host and the target.	2087000

NetApp Bug ID

Title

Description

Bugzilla ID

1479047

RHEL 8.8 NVMe-oF hosts create duplicate persistent discovery controllers

On NVMe over Fabrics (NVMe-oF) hosts, you can use the "nvme discover -p" command to create Persistent Discovery Controllers (PDCs). When this command is used, only one PDC should be created per initiator-target combination. However, if you are running Red Hat Enterprise Linux (RHEL) 8.8 on an NVMe-oF host, a duplicate PDC is created each time "nvme discover -p" is executed. This leads to unnecessary usage of resources on both the host and the target.

2087000

NVMe-oF host configuration for RHEL 8.8 with ONTAP

Creating your file...

Known limitations

Enable in-kernel multipath

Configure NVMe/FC

Enable 1MB I/O (Optional)

Configure NVMe/TCP

Validate NVMe-oF

Known issues