Options

Binding the ESX Software iSCSI Initiator to Two (or more) VMkernel Ports

benbuiltpcbenbuiltpc Member Posts: 80 ■■□□□□□□□□
We're troubleshooting an ESX 4.1 stability issue at work. We've had our hosts completely lock up, at seemingly random times, apparently due to a multipathing issue with iSCSI. We're running a Dell MD3200i SAN. While researching/troubleshooting this issue we've come across the concept of binding vmk's to the iSCSI software adapter:

Binding the ESX Software iSCSI Initiator to Two (or more) VMkernel Ports | A Sys Admin Blog

I understand how to do this, but I don't understand why. I've been doing some tests with and without the bindings, and I seem to get proper load balancing either way. So I'm not convinced that not specifying the bindings will lead to stability issues.

Comments

  • Options
    dave330idave330i Member Posts: 2,091 ■■■■■■■■■■
  • Options
    meadITmeadIT Member Posts: 581 ■■■■□□□□□□
    If you don't bind the NICs, what can happen is multiple vmkernels can use the same NIC. For example, vSwitch1 has vmk0 and vmk1, and uplinks pNIC1 and pNIC2. With the default load balancing, it's possible for both vmk0 and vmk1 to use pNIC1. If you leave the default NIC Teaming, it can confuse some storage arrays because it sees the traffic from the same IPs coming from the different MAC addresses. This can also affect arp tables in the switches they are connected to. By binding the vmkernels, it ensures consistency in where the data originates. You can almost think of it as turning each pNIC into a hba.

    Which version of firmware are you running on the MD3200i SAN? Is it up to date? If I recall correctly, they fixed some multipathing issues in one of their releases here in the last 6 months or so.
    CERTS: VCDX #110 / VCAP-DCA #500 (v5 & 4) / VCAP-DCD #10(v5 & 4) / VCP 5 & 4 / EMCISA / MCSE 2003 / MCTS: Vista / CCNA / CCENT / Security+ / Network+ / Project+ / CIW Database Design Specialist, Professional, Associate
  • Options
    benbuiltpcbenbuiltpc Member Posts: 80 ■■□□□□□□□□
    Ok, I'm getting a better picture now. What if we only have 1 vmk port and 1 nic per vswitch? Could there be MAC address corruption in that case?

    Thanks for the reminder on the firmware, I see that the latest one (released last week) does address iSCSI stability!
  • Options
    jibbajabbajibbajabba Member Posts: 4,317 ■■■■■■■■□□
    This is what we are doing ...

    One vSwitch > One vmkernel > One NIC

    And then we configure the multipathing that way .. There are no MAC address issues .. you just have to make sure those two uplinks are in a portchannel on the switch (physical switch that is).
    My own knowledge base made public: http://open902.com :p
  • Options
    dave330idave330i Member Posts: 2,091 ■■■■■■■■■■
    jibbajabba wrote: »
    This is what we are doing ...

    One vSwitch > One vmkernel > One NIC

    And then we configure the multipathing that way .. There are no MAC address issues .. you just have to make sure those two uplinks are in a portchannel on the switch (physical switch that is).

    Interesting. Any particular reason it was built that way?
    2018 Certification Goals: Maybe VMware Sales Cert
    "Simplify, then add lightness" -Colin Chapman
  • Options
    blargoeblargoe Member Posts: 4,174 ■■■■■■■■■□
    ^Yeah. Binding the vmk to the iSCSI software adapter is pretty trivial.
    IT guy since 12/00

    Recent: 11/2019 - RHCSA (RHEL 7); 2/2019 - Updated VCP to 6.5 (just a few days before VMware discontinued the re-cert policy...)
    Working on: RHCE/Ansible
    Future: Probably continued Red Hat Immersion, Possibly VCAP Design, or maybe a completely different path. Depends on job demands...
  • Options
    meadITmeadIT Member Posts: 581 ■■■■□□□□□□
    blargoe wrote: »
    ^Yeah. Binding the vmk to the iSCSI software adapter is pretty trivial.

    Even more so in vSphere 5. It's all in the GUI.
    CERTS: VCDX #110 / VCAP-DCA #500 (v5 & 4) / VCAP-DCD #10(v5 & 4) / VCP 5 & 4 / EMCISA / MCSE 2003 / MCTS: Vista / CCNA / CCENT / Security+ / Network+ / Project+ / CIW Database Design Specialist, Professional, Associate
Sign In or Register to comment.