Troubleshooting VMware ESXi iSCSI Connections from CLI

    Troubleshooting iSCSI connections… we have all been there, at the data center or your desk trying to solve some dead path issue. I like to drop back to the CLI for this task as it’s much easier to get the information out of VMware for troubleshooting purposes, and the Web clients don’t always have the most current information, without requiring a refresh.

    First, let’s get a physical layout of the land we are dealing with on this host.

    [root@esxi:~] esxcfg-nics -l
    Name    PCI          Driver      Link Speed     Duplex MAC Address       MTU   
    vmnic0  0000:01:00.0 tg3         Up   1000Mbps  Full   b8:2a:72:dc:2d:e8 1500
    vmnic1  0000:01:00.1 tg3         Up   1000Mbps  Full   b8:2a:72:dc:2d:e9 1500
    vmnic2  0000:02:00.0 tg3         Up   1000Mbps  Full   b8:2a:72:dc:2d:ea 9000
    vmnic3  0000:02:00.1 tg3         Up   1000Mbps  Full   b8:2a:72:dc:2d:eb 9000
    vmnic4  0000:05:00.0 tg3         Up   1000Mbps  Full   00:0a:f7:64:54:9c 1500
    vmnic5  0000:05:00.1 tg3         Up   1000Mbps  Full   00:0a:f7:64:54:9d 1500
    vmnic6  0000:05:00.2 tg3         Up   1000Mbps  Full   00:0a:f7:64:54:9e 9000
    vmnic7  0000:05:00.3 tg3         Up   1000Mbps  Full   00:0a:f7:64:54:9f 9000

    Note the physical adapters that are dedicated to iSCSI on this host are vmnic2, vmnic3, vmnic6, and vmnic7. The physical iSCSI adapters are split between the onboard NIC and a card. Notice the iSCSI vmnics are set with an MTU of 9000.

    [root@esxi:~] esxcfg-vmknic -l
    Interface  Label        IP Address        Netmask         Broadcast      MTU
    vmk0       Mgmt Network 1500
    vmk1       iscsi1 9000
    vmk2       iscsi2 9000
    vmk3       iscsi3 9000
    vmk4       iscsi4 9000

    Note: The output from the above command was edited for brevity, removing information that would be meaningless in this context.

    If you are having trouble with a specific VMK adapter / IP Address, you can validate that there is indeed a problem with this adapter using vmkping command. Attempt to ping an address that the problematic VMK should be able to reach using the option -I to indicate which VMK adapter to send packets out. For example:

    [root@esxi:~] vmkping -I vmk3
    PING ( 56 data bytes
    64 bytes from icmp_seq=0 ttl=64 time=0.340 ms
    64 bytes from icmp_seq=1 ttl=64 time=0.138 ms
    64 bytes from icmp_seq=2 ttl=64 time=0.177 ms
    --- ping statistics ---
    3 packets transmitted, 3 packets received, 0% packet loss
    round-trip min/avg/max = 0.138/0.218/0.340 ms

    For VMK adapters that fail to reach the address, I would go back and verify the VMK adapter configuration, port group settings, and physical switch port configuration for as they can all stop the iSCSI from working. Adapters and switch ports should be able to pass frames above 9000 bytes. To test this you can use the vmkping option -s to specify the ping size.

    [root@esxi:~] vmkping -s 9216 -I vmk3

    If you fail to receive a response you might not have the entire iSCSI connection chainset with an MTU of 9000 or greater. Verify the VMK adapter settings, virtual switch this VMK is associated with, also check the physical switch port settings.

    Filed in: ESXi, iSCSI
    Reading Time: 3 minute(s)

    VMware vSphere ESXi iSCSI Walk

    Recently I found myself having to perform a walk of physical connections of some ESXi hosts and two (2) QNAP storage arrays being used for storage of VMware Virtual Machines (VM). The idea was to remove the Network Port Binding from two (2) of the four (4) iSCSI VMK adapters, move the adapters to the destination switch.

    The first step was to remove the two (2) VMK adapters network bindings from the ESXi host, forcing the iSCSI traffic to the remaining two (2) VMK adapters that are still binding.

    ESXi iSCSI Network Port Binding

    You can monitor the Usage of a VMK adapter from the Monitor > Performance > Advanced section in the vCenter Web Client for an ESXi Host.

    ESXi iSCSI Network Port Binding Removal

    Now, you can move the cables attached to the VMK adapters to the destination switch. Once the cables are moved, you can add the VMK adapters back to network binding. The ESXi host will reestablish paths over those adapters to the storage targets now on the destination switch.

    ESXi iSCSI Network Port Binding Add

    Next was to move two (2) of the four (4) connections on the storage array to the destination switch. You will notice the paths transition to the Dead state in the Paths tab of the iSCSI adapter, which the cable is being moved. But, since there are other targets to the array, there should be no problems.

    Now, we bind the network ports and perform a re-scan of the HBAs. Verify-in vCenter that the ports are Active. Repeat these steps to move the other VMK adapters and remaining storage respectively.

    At this point, if you’re having trouble establishing paths, you should check MTU settings on the virtual switch, physical switches, and VMK adapters. Verify VLAN any other Layer 2 settings like Data Center Bridging (DCB).

    Rinse and repeat the steps for the other ESXi VMK adapters and storage targets.

    I don’t know if this is recommended by VMware or not, but it worked in my situation.

    Note: The storage array in my situation didn’t have any type of vertical port failover features.

    Filed in: ESXi, iSCSI
    Reading Time: 2 minute(s)