UCS iSCSI Boot for VMware - 2 Questions

randerson · ‎01-10-2013

UCS : 5108 with M2 and M3 blades

STG : NetApp (iscsi)

VM : Vmware 5.0

I've gotten iscsi boot working with VMware 5.0 and have run into a couple of issues. Hopefully you guys can help me out.

1) I've got 2 iscsi vNICs (for redundancy purposes) setup in my UCS template/profile. The problem I have now is that when I got to set the initiator address the ONLY options I have are DHCP and Pool .. I do not have "Static" as an option, which I assume is related to my firmware version (Chassis, IOMs and M2 blades are running 2.0 (2q) while the M3 blades have 2.0(4b)). The problem this presents is that my iscsi vnics are on different networks for redundancy, however, they both select from the same ip address pool which contains both networks - thus, the first iscsi vnic generally picks an ip address from the wrong network since the addresses are all lumped into the same pool.

The two solutions I see for this are either being able to create distinct pools that can be explicitly referenced within the iscsi boot parameters per iscsi vnic (which I don't think is possible) OR being able to set a static IP on the iscsi vnic, which I cannot right now for some reason (maybe due to firmware version, I suppose)

2) When doing iscsi boot, vmware automatically creates a vswitch associated with the iscsi vnic. I'm wondering now if it's best practice to create an additional iscsi vnic in the UCS profile which is dedicated to iscsi boot OR if it's okay to use that same vNIC/vswitch for your vmware datastores as well. If it's okay to do the latter, then there is another issue because I've got a dedicated LUN for the vmware boot drive which has it's own inititator group and initiators. On a Netapp, it's not possible to add an initiator to two different initiator groups, so now I have an issue of not being able to give access to the vmware datastores.

I'm wondering if people have just created one big initiator group that they assign to both the iscsi boot lun as well as the vmfs datastores - I suppose that's the most logical way to do it, but I'm mainly concerned if it's okay to share that iscsi vswitch with the vmfs datastores or if the iscsi boot vswitch should be left dedicated to the iscsi boot or if it can be shared.

Sorry about the massive amount of details, just didn't want to leave anything out.

TIA!

Robert Burns · ‎01-11-2013

Greetings Ross,

I'll address each point.

1. You shouldn't have any issue assigning a static iscsi IP. I checked two systems, one at 2.0(1s) and my 2.1 dev systems and both still have DHCP/Static/Pool options. Using static addresses if your iscsi networks are in different subnets is the way to go. Even in the latest verison you can only create a single iscsi IP Pool - even though you can create multiple IP blocks within the pool. We can't set which IPs from the pools will be allocated.

Can you send me a screen shot of your iscsi boot parameters.

2. I'll have to check whether the iscsi boot NIC can be used to access your VMFS datastores also. I don't "beleive" it can. I'm pretty sure you'll need to create a new vmknic specific for iscsi datastore access. I'll look into this and update the thread.

If anyone else has done this, feel free to add your comments.

Regards,

Robert

randerson · ‎01-11-2013

Thanks for the reply Robert

1) I originally had 2.0(1s) and now have 2.0(2q) - both have the same issue for me

2) I'm able to functionally "use" the iscsibootvswitch for iscsi datastore traffic, however, I just don't know if there is a downside to doing so or if it's best practice to leave that auto-created iscsiboovswitch alone, solely for iscsi boot traffic. Both NICs will be on the same network so I just don't know what vmware recommends and I can't find it in any documentation.

According to the Cisco FlexPod Setup doc, it appears that they use the iscsbootvswitch for normal iSCSI Datastore traffic (there are two iscsi vswitches, one for each storage network); however, only the network that is used to actually boot to (generally, its the "primary" iscsi vnic) is designated as the iscsibootvswitch.

You can see this on the screenshot below the section labeled "Setup iSCSI Boot Ports on Virtual Switches"

http://www.cisco.com/en/US/docs/unified_computing/ucs/UCS_CVDs/cisco_ucs_vmware_ethernet_flexpod.html

Thanks again

randerson · ‎01-15-2013

I figured out the UCS issue - you can only use a Pool within a template. I didn't realize you had to upbind the profile(s) in order to set a Static IP address for the initiator. I suppose that makes sense, given that you each profile will have a different value for IQN, Initiator IP and most likely LUN ID. Once I unbound the profile from the template, I was able to select "Static" for the initiator address.

One unfortunate side effect of using a template with iscsi vNICs is that you must have an initiator pool in order to create the initial template and profiles, however, once you change the profile to use static IP it appears that the profile still remains mapped to the pool ip address - you'd think there would be logic that unmaps that profile, but maybe that is in the next revision ..

As for the iscsibootvswitch, I have configured VMware networking assuming that the iscsi boot vswitch is also used for VMFS/Datastore traffic. This coincides with the Cisco FlexPod setup doc, so I feel okay about that.

Thanks for the input Robert.

vBahubali · ‎06-30-2018

hello

As i can see this is quite a old thread however i am facing same situation but cant find design document which give idea if we can use iSCSI boot vswitch for both boot LUN and datastore LUN. What is downside of it?