Connecting W2K3 Servers to MSA3000

[Rob Leadbeater]

Hi,

It would also be useful to see the output of the "switchShow" command as that will allow us to see what devices are connected to the SAN switches.

The zoning configurations you've posted shows that there are two SAN switches. The zoning needs to be right on both of them...

So for both switches, if you could post the output of "switchShow" and "cfgShow".

Cheers,

Rob

[WCzar]

Removed the 20: entry. Checked the WWID for the HBA in HPSANSRV - it is the same as displayed by Insight Mgr. Only trying to use switch12 as the W2K server (HPSANSRV) has only a single port HBA. Did move the cable between switch11 to switch12 for the shows only. Verifying the EVA controller addresses now. Attached are the cfgshow and switchshow for both switch 11 & 12 (1 document).

[Rob Leadbeater]

Hi,

The zoning all looks good to me, which puts things back onto the Command View EVA...

Try removing the host entry for your server, reboot the SAN appliance, and then recreate the host entry. The WWN of the HBA should then hopefully appear in the drop down list for you to add...

Cheers,

Rob

[WCzar]

In CommandView we deleted the vdisk and host for hpsansrv. Shutdown and restarted the appliance. Re-created the vdisk and host (selected hpsansrv from the pulldown list). Presented the disk. Still cannot see the san. Tried rebooting hpsansrv. Is there a mount command we need to issue before we do the re-scan?

[Rob Leadbeater]

Hi,

Where are you looking for the SAN drive on the Windows box ?

It should appear in the Disk Management tool, as a new disk...

Cheers,

Rob

[WCzar]

We have been looking in Windows Disk Management tool all along. The SAN disk doesn't show after we do a re-scan or reboot of the W2K3 server. Running DiskProbe from Microsoft displays only drive 00 (local disk). Is there not a mount command that needs to be done in order for the re-scan to see the assigned SAN disk? Seem to recall (years ago) this had to be done when connecting Windows servers to EMC. Would it be the same procedure to connect to StorageWorks?

[Rob Leadbeater]

I've never had to use a mount command, the disk usually just appears...

Can you see any reference to the EVA in Device Manager ?

Are the correct drives loaded for the HBA ?

When you presented the Vdisk to the host, in Command View EVA, what LUN number did you use ? If you let the system auto choose it, try unpresenting, and specifically Assign a LUN number...

Cheers,

Rob

[Steven Clementi]

There is no mount command you need to do in Windows.

You claim to have deleted the host in command view and Re-created i.t. Was the WWID of the server in the drop down box when you re-created the host in Command View?

Can you go to the "Ports" tab of the hosts properties, screenshot it and post it?

(sigh... wish I had remote access.) ;o)


Steven

[Uwe Zessin]

Windows 2003 Enterprise edition does not automatically mount 'foreign' disks and assigns drive letters. You can change that with behaviour MOUNTVOL or through DISKPART.

[Uwe Zessin]

On second thought, this only applies if the disk is partitioned and has a filesystem on it.

A new virtual disk just created on the EVA does not, so my previous comment does not apply.

I recommend you stop with the virtual disks for now and do some connectivity troubleshooting. Go into the Windows device manager and check if you see some "EVA" or "HSV" or some unknown controller devices.
Or go into the registry editor open [HKEY_LOCAL_MACHINE\HARDWARE\DEVICEMAP\Scsi] , look for the SCSI ports represented by the Emulex ports and the traverse the bus/target/lun hierarchy. You _should_ see the EVA controller devices at LUN address0, which are always present - independed from host entries/presentations:

C:\>reg query "HKLM\HARDWARE\DEVICEMAP\Scsi\Scsi Port 5\Scsi Bus 1\Target Id 0" /s

HKEY_LOCAL_MACHINE\HARDWARE\DEVICEMAP\Scsi\Scsi Port 5\Scsi Bus 1\Target Id 0\Logical Unit Id 0
Identifier REG_SZ HP HSV200 5000
Type REG_SZ ArrayPeripheral
InquiryData REG_BINARY 0C000512F70000324850202020202020485356323030202
0202020202020202035303030
SerialNumber REG_SZ PB5Axxxxxxxxxx
DeviceIdentifierPage REG_BINARY 0C8300140103000850001FE1500xxxxx0114000400000003


If these devices do not show up, the problem is most likely an incorrect zoning the Fibre Channel switch(es). Don't waste ANY time on virtual disks unless you do see these device!

[WCzar]

The WWID of the server (hpsansrv) was in the pull-down after we rebooted the appliance - we selected the WWID, we did not type it in. Screen print for the host's (hpsansrv) ports tab exceeds 1 MB, but the FC port address is correct (1000-0000-c93f-1eb9). Remote access, hmmmmm. This thread has got to be one of the longest ever.

[Steven Clementi]

"This thread has got to be one of the longest ever."

Naa.. it's only half as long I think right now.


So, the EVA can see the host, or at least seen the host at some point (confirmed by the fact that you can see the wwid in the drop down).

Can you zip the screenshot? or convert it to a jpg?


In the Windows Device Manager under SCSI and RAID devices, yo ushould see 2 entries for the EVA. Can you confirm? (2 entries since you said the server only has 1 hba)



Steven

[WCzar]

Attached is a screen print for the registry hives mentioned earlier. The identifier parameter shows HSV101. Is it possible the LUN assignment on the Host (hpsansrv) within CommandView is not correct or did we miss some config setting when we installed the Emulex 9802. We are able to use the Emulex HBAnywhere on the W2K3 server (hpsansrv) and echo test back to the 50: addresses though. There are two 2/16 switches. Each has a connection to both hsv controllers. Do we need the mpio or the multipath pieces (they are installed).

[WCzar]

There are entries in device manager for Emulex and EVA - there are no "unknown" entries.

Attached is a .bmp for hpsansrv port tab.

[WCzar]

Attached is a .bmp for hpsansrv presentation tab.

The LUN is = 1 - not sure if this agrees with how the Emulex is configured.

[Rob Leadbeater]

Hi,

The host presentation looks OK to me...

A couple of things from the registry output and screen grabs however.

This bit:
"Identifier"="HP HSV101 3010"

would appear to suggest VCS 3.010 rather than 4.007 which you mentioned earlier. I'm guessing that's a red herring, but it's worth sanity checking, as the difference between v3 and v4 is the move from Active/Passive to Active/Active controllers, which may have an impact when you've only got one connection to the server.

The other thing I noticed was the warning triangle in Command View for the Hardware. Can you drill down there and see what's wrong... If it happens to be something to do with the cache batteries that may well explain everything... Bad batteries = No LUNs.

Cheers,

Rob

[WCzar]

From CommandView:
Initialized System Properties shows;
Type: HSV100
Version: 4007
Software: SR0045runp-4007

Checked the hardware. There is 1 failed drive, which we knew about before the cabinet was moved. Sometime after we started to configure for our test server (hpsansrv) a failed battery appeared on one of the controllers. Checking this afternoon we see two failed cache batteries - 1 on each controller (A & B). Each controller has two batteries - there is 1 good battery remaining on each.

[Uwe Zessin]

> This bit:
> "Identifier"="HP HSV101 3010"
>
> would appear to suggest VCS 3.010 rather than 4.007

No, you will see this on other EVAs as well - WCzar's box apparently has some history ;-)

The EVA keeps the SCSI inquiry string (at least for Windows systems) even after an update of the controller firmware to prevent Windows 'hotplug' events from happening.

[Rob Leadbeater]

Hi,

> there is 1 good battery remaining on each

OK that shouldn't affect things, but I'd be more convinced if all the batteries were good. EVAs seem to do odd things when batteries go bad...

Another thought, have you tried restarting the EVA itself ?

Cheers,

Rob

[WCzar]

Completely powered down and restarted everything yesterday.
Attached is a screen print from the emulex hbanywhere utility installed on the w2k3 server that can't see the san from Disk Management. It shows the 50: addresses. We can echo back to those addresses successfully from the server (LUN 0). That should equate to having a connection from purely a Emulex hba standpoint. So if that works, then back to the original issue - why can't Disk Management in Windows see it after doing a re-scan. What should the LUN assignment on the host (hpsansrv) in CommandView be if the physical box for hpsansrv shows LUN 0?

[Uwe Zessin]

The existence of LUN 0 (the green 'card' icon suggests it is a controller device, not a presented disk) shows that we have basic fabric connectivity. That's good so far.

But you should see a LUN 1 (and a disk icon left to it). As long as that is not the case - you won't see a disk in Windows disk management.

This really sounds like the EVA is not presenting the disk due to a problem with the cache batteries - can you put an intact pair into one controller? The output from HBAnywhere shows that the adapter has access to both controllers.
If Rob's guess is correct you should now see a disk icon and "LUN 1" on one of the paths.

[WCzar]

The problem has been resolved! We replaced all 4 cache batteries and now see the SAN from the Windows test server. We created a new zone and were able to connect a VMWare host. Everything seems to be working as expected. Many thanks to all of you who responded with such great support and most of all your patience. Thanks again.

[WCzar]

When we started to work on resurrecting the EVA3000 all the cache batteries were OK. However, shortly thereafter one of the 146 GB drives failed and then the cache batteries, one by one. Not sure when the first battery failure occurred, but the W2K3 test server was only able to see the SAN after we replaced all 4 failed batteries (rather expensive). Since the hardware is active\active we thought there would still be some redundancy - not sure if that was ever the case.