Trouble with QLA4010 iSCSI on ESX 3.0.1

[boyedav]

We're running VMWare ESX 3.0.1 (fully patched up to 3/29 patches) with (2) QLogic QLA 4010 iSCSI HBAs per server. Our iSCSI targets are Sanrad V-Switch 2000's.

I can't seem to get the HBAs to work. I've given them each IP addresses, pointed them to targets for dynamic discovery, and set CHAP parameters. I've ran several vmkiscsi-tool -D and vmkiscsi-tool -X, and even rebooted the entire server.

If I plug my laptop into the network ports used by the iSCSI HBAs and give it that same IP as the HBAs would have, I have no trouble pinging the default gateway and the iSCSI gateways. I can even load Microsoft's iSCSI initiator on the laptop, and it dynamically discovers the iSCSI targets just fine. The software-based iSCSI initiator built in to ESX can also discover and talk to the iSCSI targets just fine (I otherwise have the built-in VMWare iSCSI initiator turned off).

I've also gone into the HBA config in VMWare and added the targets statically, but still no luck.

iSCSI works just fine everywhere but with the QLA4010 on VMWare. I've even removed and recreated the target because I heard that was an issue with iSCSI and the Sanrad V-Switch.

Any suggestions are appreciated. I have an open support request with VMWare, but I've not yet heard back from them. Thanks!

[Paul_Lalonde]

What are you seeing in the /var/log/vmkernel log file?

Paul

[boyedav]

Nothing I can see that's out of ordinary, but here it is:

http://web.bvu.edu/staff/david/vmkernel.txt

my iSCSI HBAs are vmhba1 and vmhba2

[christianZ]

You know that the qla4010 is no more supported with 3.0.1 !!

[pdrace]

They've never been supported for production though I've read threads by users who are using them for production.

I currently have two 3.01 test servers running with 4010s.

I've actually had less problems with them than the 4050s with 3.01.

[boyedav]

So what's the story with the QLA4050?

We went with the QLA4010 because, at the time, 3.0 had just come out and our VAR indicated production support for the QLA4010 would be in 3.0.1. Is the QLA4050 production-ready, and is there reason to believe support for it in VMWare won't be sustained? Does VMWare use the QLA4050 for both iSCSI and TCPIP offload?

What about the iSCSI software initiator? I understand the obvious advantage of using a hardware HBA, but anything else noteworthy about using the software initiator?

[pdrace]

The QLA4050 is qualified for production on 3.01 and I believe that others are using it successfully.

I've had some isues that aren't resolved due to the fact that qualification of the storage device I'm using with any ISCSI hba is pending on the part of the storage device manufacturer (NetApp) and the fact that support is pending for the version of the OS on my storage device is also pending with VMWare.

I'm using the SW initiatior and it's working though I'm still seeing occasional issues.

[pdrace]

We're running VMWare ESX 3.0.1 (fully patched up to

3/29 patches) with (2) QLogic QLA 4010 iSCSI HBAs per

server. Our iSCSI targets are Sanrad V-Switch

2000's.

What is the actual target ISCSI server?

It looks like the SanRad is just a switch not a storage device.

[christianZ]

Maybe you can contact sanrad (is a storage head manufacturer) - what they suppose ?

[boyedav]

The storage device is a Gateway 842 SAS SAN, which is just a re-branded Xyratex F5402E. The SANRAD V-Switch is an iSCSI gateway. My VMWare boxes with the QLA4010 are supposed to be talking to the V-Switch, which exposes the LUNs on the Gateway SAN.

The V-Switch and 842 work as expected from all other iSCSI clients. Even the software initiator in VMWare works. The only thing not working are the QLA4010s. I can even plug a laptop with an iSCSI software initiator into the network port used by the QLA4010 and connect just fine.

I've opened and closed a support ticket with VMWare. The QLA4010 isn't supported, end of story.