Modifying the overall network link MTU for the storage network

For clusters without RDMA enabled, when the storage network bandwidth is 25 Gbps or higher, you need to modify the MTU of the overall network link on the physical switch, ESXi host, and SCVM ports to enhance the overall bandwidth performance. Otherwise, you can skip the following steps.

Procedure

1. Log in to the physical switch management page and change the MTU of the physical ports corresponding to the storage network to the maximum MTU supported by the switch. For the specific value, refer to the manual of the corresponding switch.

2. Enable the SSH service on the ESXi host, connect to the ESXi host through the SSH client, and follow the steps below to configure.

   1. Run the following command to change the vswitch mtu value of the standard switch for the storage network to 9000 and check if the modification has taken effect. vSwitch1 indicates the name of the storage network standard switch.

      [root@localhost:~] esxcfg-vswitch -m 9000 vSwitch1
      [root@localhost:~] esxcfg-vswitch -l

   2. Run the following command to change the vmkernel mtu value of the storage network to 9000 and check if the modification has taken effect.

      [root@localhost:~] esxcfg-vmknic -m 9000 -p VMkernel
      [root@localhost:~] esxcfg-vmknic -l

3. Refer to step 2 to complete the configuration on all ESXi hosts and use the ping command between any two ESXi hosts to verify connectivity, which can indicate if the above configuration is successful.

   Note:

   When using the ping command, if the ping packet size is the default value (56 Bytes), you will not be able to determine whether the network is functioning normally with an MTU of 9000. In this case, run the ping -s 8972 -d $IP command to set the ping packet size to 8972 Bytes.

4. Log in to the SCVM of the ESXi host and follow the steps below to configure.

   1. Edit the storage network NIC configuration file /etc/sysconfig/network-scripts/ifcfg-ensxxx, where ensxxx represents the specific NIC name. Add the following content to the file and save.

      MTU="9000"

   2. Run the systemctl restart network.service command to restart the network.

   3. Run the following command to check if the configuration has taken effect, where ensxxx represents the specific NIC name.

      [root@localhost network-scripts]# ifconfig ensxxx
      ensxxx: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 9000
              inet 10.10.1.253  netmask 255.255.0.0  broadcast 10.10.255.255
              inet6 fe80::1b09:b2d2:ffd6:3595  prefixlen 64  scopeid 0x20<link>
              ether 52:54:00:e7:27:25  txqueuelen 1000  (Ethernet)
              RX packets 417875601  bytes 71409590956 (66.5 GiB)
              RX errors 0  dropped 390657  overruns 0  frame 0
              TX packets 156741156  bytes 62038586752 (57.7 GiB)
              TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

5. Refer to step 4 to complete the configuration for the SCVMs on all ESXi hosts and use the ping command between any two SCVMs to verify connectivity, which can indicate if the above configuration is successful.

   Note:

   When using the ping command, if the ping packet size is the default value (56 Bytes), you will not be able to determine whether the network is functioning normally with an MTU of 9000. In this case, run the ping -M do -s 8972 $IP command to set the ping packet size to 8972 Bytes.