Administer and use EVA 4100 without fabric?

[gordonf4hp]

Let's try this again...

I have a EVA 4100 that was decommissioned and replaced with a 4400. The 4100's been reassembled at a remote location, and as best as I can tell is working correctly. I will lose fiber connectivity to it frequently though, Command View 7 will fail to contact it often, and disks presented to my host will drop off the host's view.

The same host and EVA worked correctly when used with a pair of Brocade switches. Now those switches are gone, along with the blade chassis they were part of, and I'm trying to make it work as a stand-alone array and host.

At first I tried connecting one FC card to one controller, and changing the controller host port settings to "single" instead of "fabric." This makes the controllers visible to the host, but the controllers would complain about the other host port being disconnected. After a while, I'd have to disconnect and reconnect the FC cables to make the controllers visible again.

Then I tried connecting both FC cards to a single controller. This made the controller stop complaining, but I still get these frequently in my Application event log:

--HSV Controller Event--

Date/Time Stamp: 3-Mar-2011/07:32:23
StorageCell WWN - Name: 5000-1FE1-5010-9670 EVA_MCBH
Event Type: HSV Controller Event
Event Code and Description: 09cdc305: CAC=c3 - Storage System Management Interface Entity State Change The state of a Storage System Management Interface entity has changed. A Fibre Channel port has transitioned to the FAILED state.

Also, the Command View EVA service will crash, requiring me to restart it to administer the array.

I'm very sure that the array is cabled correctly after reviewing the 4100 documentation, terminators and all, and the controllers themselves don't seem to be complaining about any cabling except that of the host ports.

Fabric switches are expensive and since this will be the only host that will connect to this EVA from now on, I'd like to avoid purchasing them.

The host itself runs Windows Server 2003 R2 x86 SP2, and has Command View 7.0 installed. I have an installer for 7.1, along with various add-ons like multi-path I/O support installed. I think I'm missing some fundamental piece to make this work properly and with both controllers instead of just one.

[Mark...]

Hi,
If I read your description correctly then you have two host connections to ONE ctrl ??
Try connecting one connection to each ctrl FP1, setting direct connect and putting loopbacks on both spare FP2 ports.
Mark...

[gordonf4hp]

"If I read your description correctly then you have two host connections to ONE ctrl ?? Try connecting one connection to each ctrl FP1, setting direct connect and putting loopbacks on both spare FP2 ports."

I originally had one host port connected to each controller but the controllers would complain that one host port was down, and I'd still get connection outages. I tried both FC host ports on one controller out of desperation.

OK I'll put the host's FC cards back to one to each controller's FP1, and try a fiber loopback on each controller's FP2, and see what happens.

[gordonf4hp]

So far, recabling the host ports on the 4100's controllers to my single host, and looping the second host port on each controller, is working out. Command View isn't complaining about failed ports, and it's actually staying up and not crashing.

I'm still geting host port failures recorded on my host's event log. I think these were being reported at the old location this EVA was assembled at though, so I might be seeing an old problem that just hasn't been fixed. I might have one bad controller because swapping the GBICs has not made a difference.

I'll see how this holds up.

[gordonf4hp]

Mark...'s solution of recabling the host ports and using loopbacks for the ununsed host ports is working perfectly. The host is no longer complaining about outages, and the EVA isn't complaining about connection problems.

Making a loopback is easy with an existing FC patch cable; the two connectors come apart with a clamp that holds them in place, and you just connect both white ends to one GBIC and both yellow ends to the other GBIC. The controllers were close enough together that I could do this without splitting my FC cable apart.

I'd just want to add one more thing: For a single host with two FC ports make sure you asssign BOTH FC ports to the host in Command View, to make sure the EVA knows you have a redundant path. I don't work regularly with FC storage so this was something I or another newbie FC admin could miss, and I did miss this the first time.

Am I able to assign three more points to his post? :-)

[Uwe Zessin]

> to make sure the EVA knows you have a redundant path.

It does not really matter, because the EVA does NOT know about redundant paths to hosts - it is just a SCSI target. However, doing so it makes it obvious which WWNs belong to the same host, and makes virtual disk presentation easier and consistent.