KM_MPL1
Contributor
Contributor

ESXi not attaching all iSCSI volumes

Ive run into an issue that I cant quite seem to understand. 

I have 3 hypervisors that are configured to connect to two different iSCSI storage devices (10.255.1/2.6 and 10.255.1/2.8) using multipathing. The hosts and the storage devices are connected to an isolated switch. The switch has two vlans, 20 and 21, for the above subnets. The storage qnaps have interfaces assigned with appropriate addresses for their vlans, as listed above. The problem I am encountering is two of the three hosts have been able to connect via dynamic targets, to the second storage device, 10.255.1/2.8, but cannot seem to connect to the second, 10.255.1/2.6. The third host cannot connect to either. Ive tried  pinging from the hosts to both storage devices, pings reply. Hosts can ping each other. SVI interfaces on the switch are configured and pingable from the host as well. So all hosts can ping each others vmkernels, all hosts can ping the gateway (the switch vlan SVI's, and all hosts can ping both adapters on both storage devices. MTU for everything is set to 9000. What am I missing? 

 

0 Kudos
3 Replies
scott28tt
VMware Employee
VMware Employee

@KM_MPL1 

Moderator: Moved to vSphere Discussions, doesn't seem to related to vSphere Hypervisor (free ESXi)

 

0 Kudos
NathanosBlightc
Commander
Commander

Can you explain why you dedicated two separate VLAN IDs (20,21) for the iSCSI traffic? Is there any special consideration for such as this config? or you can ignore them and make it simpler via putting the QNAP and all hosts inside a single VLAN!

First of all try this: If you remove the initiators of 1st & 2nd hosts, and just map the 3rd one, then is the problem still persist?

Can you check your solution with MTU size 1500, not Jumbo Frame?

Please mark my comment as the Correct Answer if this solution resolved your problem
Tags (1)
0 Kudos
a_p_
Leadership
Leadership

Some thoughts:

  • when using 2 subnets/VLANs, create separate vSwitches (one for each subnet) instead of one vSwitch
  • make sure that both, the port group as well as the vSwitch have the proper MTU configured
  • make sure that your physical switch, and storage system also support Jumbo frames
  • test connections using vmkping -d -I vmk# -s 8972 storage-IP to ensure the Jumbo frames work without being defragmented
  • with multiple subnets you must not use explicit port binding, so remove the vmk's from the port binding section in the Software iSCSI adapter's settings. Explicit port binding requires that all vmk adapters are able to access each target IP.
  • depending on the storage configuration, you may need to define the initiators (ESXi hosts) by using either their IP addresses, or iSCSI IQN, so that you can present the LUNs to them.

André

0 Kudos