ESXi with software iSCSI

Question

Has anyone had any luck using the swiSCSI driver on ESXi? Following the instructions from VMWare.com I get to the point where I have the iSCSI HBA showing up but no LUNs/targets are showing up.

The iSCSI target is running on Solaris 10 update 5 and works with other initiators fine. The ESXi initiator (from the logs) sees the targets but just logs in and out of them every 2 - 5 seconds. We're using unauthenticated discovery, and over and over in /var/log/messages I see:

iSCSI: bus 0 target 0 trying to establish session 0xb203f90 to portal 0, address 10.1.100.9 port 3260 group 1
iSCSI: bus 0 target 0 establish session 0xb203f90 #4848 to port 0, address 10.1.100.9 port 3260 group 1, alias data/ESXi
iSCSI: session 0xb203f90 dropping after receiving unexpected opcode 0x60
iSCSI: session 0xb203f90 to data/ESXi dropped
iSCSI: session 0xb203f90 to data/ESXi waiting 2 seconds before next login attempt

The only other thing that seems out of wack is that my 'Recent Tasks' pane keeps filling with 'Browse Diagnostic Manager' events and /var/log/vmware/hostd.log is filled with messages like this up to two times per second:

[2008-09-19 16:05:57.901 'TaskManager' 196621 info] Task Created: haTask-ha-host-vim.DiagnosticManager.browser-776
[2008-09-19 16:05:57.094 'TaskManager' 196621 info] Task Completed: haTask-ha-host-vim.DiagnosticManager.browser-766

Any help would be appreciated.

Answer 1

I see several references to the Opcode 0x60 error on VMware Communities. All who have seen it are using ...you guessed it, a Solaris target. The threads I've read suggest this is a bug in the iSCSI target driver. See these:

http://communities.vmware.com/message/951142;jsessionid=F31FE5B31800074F0C8997FFFEE30AC6 http://www.opensolaris.org/jive/thread.jspa?messageID=167293

I don't know the Solaris version numbers nor the relationship of Solaris code to OpenSolaris, but I would make sure you are running the latest code on both the Solaris and ESX sides. Are you on 3.0.x or 3.5?

Answer 2

I found there was an issue with the our Sun gear (X4150 talking to ST2510) and VMware ESXi whereby the connection would work if the host's name was short enough.

We had two servers; one's full hostname was 31 characters and one was 35 (from memory) and the shorter host was the only one that worked. It appears that there's a limitation with the length of the IQN.

Our solution: we registered a shorter domain for use with this equipment so we could keep naming our servers after Transformers :)

Answer 3

"The only other thing that seems out of wack is that my 'Recent Tasks' pane keeps filling with 'Browse Diagnostic Manager' events and /var/log/vmware/hostd.log is filled with messages like this up to two times per second:

[2008-09-19 16:05:57.901 'TaskManager' 196621 info] Task Created: haTask-ha-host-vim.DiagnosticManager.browser-776 [2008-09-19 16:05:57.094 'TaskManager' 196621 info] Task Completed: haTask-ha-host-vim.DiagnosticManager.browser-766"

How did you resolve this issue? We have exactly the same error filling up the logs.

Answer 4

At first I tried to restart the service manager in the service concole, but this didn't help.

Then I have rescanned my iscsi hba, went to performance and resource allocation and it stopped... very strange thing, makes me a bit scared!

Answer 5

and it was back again..

I think it might be because I pressed "show next 2048 lines", because the cancel could only be clicked. And I clicked that one earlier today.. so it enabled itself again when logged on using vic I think..

No I tried my own advice, rescanning the iscsi hba, changing views in resource allocation, the virtual machines tab in performance, but that didn't help initially..

I had to press the cancel button in the logview and then switch views to other pages to make it stop again!

Hope this helps

Answer 6

Haveing the same problem. Just finished installing ESX 3.5 Update 3 after downloading it this morning. The server is a Dell R900 running currently on the bench before it replaces a Dell 2950 w/ESX 3.5 Update 1. Now, I'm not sure how fast I want to run and install this if there is going to be an error eating cycles. Anyone find any good fix yet?