Alex,

You need to verify that the SDC (the ESXi host vmknic, really) can talk to the MDM's SVM on the data network IPs.

First, make sure you don't have duplicate IPs anywhere.

SDC Module Parameter Retrieval

If the MDM does not recognize the SDC or there is a reason to think that SDC does not communicate with the MDM, the MDM IPs and the GUID can be retrieved by running:

esxcli system module parameters get -m scini
ESXi 6: esxcli system module parameters list -m scini

ESX SDC Connectivity Check

This article assumes two data networks, configured with 192.168.152.x and 192.168.154.x subnets.

Find out which vmknic is on which subnet, as well as MTU:

[root@server1-101:~] esxcfg-vmknic -l
Interface  Port Group/DVPort/Opaque Network   IP Family  IP Address           Netmask         Broadcast         MAC Address        MTU     TSO MSS   Enabled Type         NetStack
vmk0       sys-mgmt-krnl                      IPv4       192.168.105.101      255.255.255.0   192.168.105.255   00:50:56:ff:ee:58  9000    65535     true    STATIC       defaultTcpipStack
vmk3       sys-vmotion                        IPv4       192.168.106.101      255.255.255.0   192.168.106.255   00:50:56:ff:ee:fd  9000    65535     true    STATIC       defaultTcpipStack
vmk1       sio-data1-krnl                     IPv4       192.168.152.121      255.255.255.0   192.168.152.255   00:50:56:ff:ee:56  9000    65535     true    STATIC       defaultTcpipStack
vmk2       sio-data2-krnl                     IPv4       192.168.154.141      255.255.255.0   192.168.154.255   00:50:56:ff:ee:dc  9000    65535     true    STATIC       defaultTcpipStack

Ping the Primary MDM's SVM via each vmknic:

If jumbo frames are configured and this test fails, network connectivity is not 100%:

vmkping -d -s 8972 -I vmk1 192.168.152.101
vmkping -d -s 8972 -I vmk2 192.168.154.101

Try without jumbo frames:

vmkping -I vmk1 192.168.152.101
vmkping -I vmk2 192.168.154.101

Ping the Secondary MDM:

Jumbo:

vmkping -d -s 8972 -I vmk1 192.168.152.102
vmkping -d -s 8972 -I vmk2 192.168.154.102

Normal:

vmkping -I vmk1 192.168.152.102
vmkping -I vmk2 192.168.154.102

Ping the subnet gateways:

Jumbo:

vmkping -d -s 8972 -I vmk1 192.168.152.1
vmkping -d -s 8972 -I vmk2 192.168.154.1

Normal:

vmkping -I vmk1 192.168.152.1
vmkping -I vmk2 192.168.154.1

Next steps, get:

The output of ifconfig run on the acting primary MDM's SVM

The output of "esxcli system module parameters get (list, if ESXi 6) -m scini"

The results of the vmkping -d -s 8972 -I vmkx MDM_SVM_IP commands against the primary MDM SVM IPs per above.

And if all configurations are ok, try search for "scini" in /var/run/log/vmkernel.* logs.

Find the last entry that includes the MDM IPs. If they are not the same as "esxcli system module parameters get -m scini" shows, reboot the ESXi to make sure the current config is effective.

Thanks very much for your help.

I'm 3 ESX for test the scaleio's solution :

ESX1, ESX2 and ESX3 :

esxcfg-vmknic -l in ESX1

In ESX2

In ESX3

My vmk0 are Management Networks.

Since ESX 1 :

vmkping -I vmk1 192.1.0.2 (192.1.0.2 is the vmk1 of ESX2) it 's OK I ping

But I Write this vmkping -d -s 8972 -I vmk1 192.1.0.2

"3 packets transmitted, 0 packets received, 100% packet loss"

I'm Verify the config of my physical switch and for me not have error.

My example config for a port on the physical switch :

Thank you for your help

Alex,

Per the steps in my response above, you need to verify that jumbo frames are configured correctly on this network.

Your switch output indicates that it is configured, so you should be able to jumbo ping the network gateway.

Test pinging this with:

Ping the subnet gateway:

Jumbo:

vmkping -d -s 8972 -I vmk1 192.1.0.x

Normal:

vmkping -I vmk1 192.1.0.x

I notice that you have the first ESX node's vmk1 IP address set to .1, which would likely also be the network gateway's IP.

Please review your network configuration to ensure there are no duplicate addresses, and that pinging with jumbo frame size will also succeed to the network gateway.

Also please confirm jumbo ping from each ESX node to the others. 1 -> 2 and 3, 2 -> 1 and 3, 3 ->1 and 2.

Thank you for you reply.

I can not vmkping my gateway

For exemple the config for one esx (esx1) :

My gateway is 10.24.113.254 for the vmk1

I ping 192.1.0.2 or .3 (esx2 and 3)

But It's impossible to ping the gateway

Thank you

That's fine.

I see you said earlier:

Since ESX 1 :

vmkping -I vmk1 192.1.0.2 (192.1.0.2 is the vmk1 of ESX2) it 's OK I ping

But I Write this vmkping -d -s 8972 -I vmk1 192.1.0.2

"3 packets transmitted, 0 packets received, 100% packet loss"

This indicates to me that you've got something in the network settings is set to MTU of 1500, not 9000 or 9216 like you show the switch is set to.

Check each vSwitch_sio properties for MTU and make sure they're set to >9000.  Retest the ping with MTU 8972.

Check vmknic MTU as well. Retest with MTU 8972.

Also, try pinging the Primary MDM SVM's data network IP.  I see that the VMs are named with 10.24.113.x subnet.

Make sure the SIO data IP on the SVMs are on the 192.1.0.x subnet or they will not communicate.

No

I ping with the mtu 8972

For information my config Data Network and Scale IO VM

What 's can I ping please ?

Alex,

I see the issue.

Your data network subnet in the plugin is 10.24.120.x, but you've set the vmknic for the data network to the 192.1.0.x subnet.

The SDC (vmknic, 192.1.0.x) can't reach the MDM Data IP (10.24.120.x, per above).

Either redeploy with the data network using the same subnet as the vmknics currently have, or change your vmknic IP addresses to be on the same subnet as is configured in the last screenshot image above.

Either way, make sure you can ping the MDM SVM's data network IP from the vmknic with jumbo frames.

Thank you for all I modifie the configuration :

and the data IP :

192.1.0.120/24 and 192.1.2.121/24 and 192.1.0.122/24 and 192.1.0.123/24 and 192.1.0.123/24

The errors are my data IP and VMkernel were not on the same network.

Thanks you