VI3 and EMC CX-300 Storage

[Thomas_Louis]

Hi,

I installed an ESX environment that needed shared storage on a Clarion cx300 storage system.

After the installation, i connected the fibre cable and after the rescan of HBA's and VMFS volumes i saw something that i'm not comfortable with. When i selected 'Storage Adapters' i saw the two HBA's and they already saw a target. Not sure whether it was a disk or not. The LUN ID was something like vmhba0:1:0. Now if i recall correctly, a disk should have a 4 digit address right? Something like vmhba0:1:0:0.

The fact that i'm uncomfortable with this, is that i never presented a lun to the esx environment. The disk/target is identified as 0 bytes.

Can anyone elaborate on this please? I'm not sure if this is a disk or the storage controller. I am familiar with storage, but not with the dell systems.

Thanks

Thomas

[robert_franklin]

It sounds like you have ESX installed on local storage, and then connecting things to the CX300. So, your storage adapters tab in VI3 should list your local SCSI adapter, plus your HBAs.

For example, on one of my blades connected to a CX700, I see an "Ultra320 SCSI" adapter labeled vmhba0, plus my 2 "QLA231x/2340" HBAs labeled vmhba1 & vmhba2. The fibre HBAs also list the world wide names next to them.

Clicking on each HBA should show you below what disks are attached. It is alright that you only see a 3 digit address; my local storage appears as vmhba0:0:0. It should display a capacity in G or T unless it is a CDROM.

With our CX700, the process looks like this:

1) In Navisphere, make sure the server initiator records are registered, and add the LUNs & hosts to a storage group.

2) At this point, ESX should find the LUNs. If they are already formatted for VMFS, ESX will auto-mount them in /vmfs when it finds them. You can initiate a rescan from the command line with the command "vmkfstools -s vmhba1" (or vmhba0, vmhba2, etc.), or rescan from VI3. Note that using the rescan at the Storage Adapters level doesn't always reflect changes for us, but right-clicking on each HBA and selecting rescan does.

3) If the LUN isn't formatted for vmfs, then it won't auto-mount. In your Storage panel you'll have to select Add Storage, and walk through the wizard to select your unformatted LUN and put a file system on it.

I hope this helps--

[Thomas_Louis]

Thank you for your quick reply.

My ESX is indeed installed on local storage. But in 'Storage Adapter' I'm sure i selected the HBA. In the details pane i clearly see a target on the HBA. Something like vmhba1:0:0. I might have given you a wrong example with the vmhba numbering as vmhba0 is indeed local storage.

But this is really a SAN issue. In the meantime i tested with a brand new blade with windows 2003 installed. He also sees some kind of storage. I don't want to take a risk clearing some production disk.

Maybe this sheds some clarity on the situation

Thanks

[robert_franklin]

OK... that sounds to me like you may have a LUN assigned to the server, and the server sees it, but it isn't formatted. If so, you would need to try to Add Storage and see if you see a non-VMFS LUN.

[Thomas_Louis]

The weird thing is that it's a new server and i didn't do any presentation settings on the SAN for the esx server.

Presentation of LUN is active on the san, so i'm not sure if i can safely ignore this volume. Let alone format it as vmfs without losing data.

[badazws6]

You didn't pull hba's out of an old server to put into the new one did you?

[Thomas_Louis]

No, not at all. We installed a new blade enclosure and installed esx on a blade.

We successfully did a san switch fabric extension to the blade switch. I'm talking about brocade switches who are both interlinked (EL200 and a 4/8 switch). Zoning has been set up and rechecked.

[robert_franklin]

Can you confirm in Navisphere that there aren't initiator records registered for your HBAs? You can pull the WWNs for your HBAs from VI, and then check connectivity on your CX300 in Navisphere to see if those names are registered.

[Thomas_Louis]

The hosts haven't been registered yet. To do that you'll need an agent. Don't know if there is one for EMC storage.

The WWN's do appear in the hosts list. But because they aren't registered, no presentation of LUN's can be completed.

[InsaneGeek · Accepted solution]

You don't \*have* to have an agent to register... just that using an agent makes sure that things are configured end to end correctly.

When attatching a clariion to a host, you'll always see something at lun0, it will be very small (I call it a ghost lun, you can't write to it or fdisk format it directly). This allows the host agent to register itself with the clariion (it sends certain commands over the scsi channel to the clariion), once the hba's have been added to a storage group with other luns this "ghost lun" will be replaced with real luns.

Going from memory: On the service console, do a "cat /proc/scsi/scsi" and see if you see type DGZ or DGC, the DGZ is what I'd call a ghost-lun it can't be written to by the host.

[robert_franklin]

Thats good info. I don't know that I've ever seen the type DGZ... but I might not have noticed it on my initial setup with the CX700 since we assigned LUNs right away. That sounds like it would explain exactly what the original poster is seeing.

It's also correct that you don't have to have an agent to register records. You can get the WWNs of your HBAs from the command line or VI client, then go find them in Navisphere. If you right-click on your CX and pick connectivity status, you'll get a list of WWNs of your devices on the fabric. Look for the ones with corresponding WWNs, and you'll be able to register them grouped into a host.

[AMcCreath]

Type DGZ is your Storage Processor, you see this everywhere, from ESX to Windows.

To initiate without the agent for Navi you will need to enter engineering mode and manually add in the HBA WWN, then select the check boxes under the Advanced button in the Connected hosts tab.

That will authorise and allow access to the disk you present to the Storage Group for your ESX farm.

Good luck,

Andy

[whynotq]

let me echo what's been said already.

when you create your zone or attach your server directly to the clariion the initiator ID (wwnn) for each HBA appears in "connectivity status" connected to each visible SP port.

at this point in time the server can see a device on the HBA bus, and in traditional UNIX style, presents a LUN0 (pseudo LUN) to the host so that when the server reboots it doesn't sit trying to scan for a device forever.

as also stated, you don't have to have the Navisphere agent installed to allocate storage. Each adapter can be manually registered against each SP port by highlighting the WWNN and selecting "register". the advantage of the agent is purely from a manageability perspective, it is supported with Flare 19 and above and seems to neccessitate a reboot of the ESX after installation.

[BUGCHK]

Now if i recall correctly, a disk should have a 4 digit address right?

Something like vmhba0:1:0:0.

The first 0[/b] is the adapter number.

The 1[/b] is the SCSI target ID/address. Each controller port that is accessible from a single vmhba[/i] receives an address, starting with 0.

The second 0[/b] is the SCSI LUN address (often called LUN ID while LUN and ID are ambiguous and used in other contexts as well) but that is life...). Each target - e.g. vmhba0:0[/b] or vmhba0:1[/b] has its own LUN address space.

The last digit usually specifies the partition on the adapter:target:LUN. Partition numbers start with 1[/b] but 0[/b] is a special value to mean "whole disk"[/i].

In many situations it is clear that the whole disk is meant, so the partition specifier is left off.

[Thomas_Louis]

InsaneGeek, thanks for your input. Indeed it is the LUNZ device.

This solves my question. Thanks everyone for your replies.

Regards

Thomas