Dell MD3000i and ESXi v5 update 1

[seanhaynes]

Morning people,

I have a Dell MD3600i for primary shared storage that so far works fine with vsphere 5, I also have a MD3000i as secondary shared storage that is giving me problems since the upgrade to ESXi 5.

I have 3 Dell R510 hosts that connect through 2 Cisco 4900M switches.

The issue I have is that there is a constant disconnect and reconnect to the LUNs on the MD3000i - the only way I can get it anywhere near stable is through the Dell software to change the controller 'owner' of the LUNs.

There are 2 LUNs on the MD3000i, both vmfs3, both RAID 5 - the MD3000i itself has dual controllers, each controller has 2 ports, each port on its own iSCSi vlan.

ESXi on each of the hosts can see the LUNs but errors with messages like; LUNx lost due to connectivity errors, attempting to reconnect - it reconnects, then drops again. If I change the ownership of the path to the other controller it works again without error.

Are any other MD3000i owners having these issues?

[MauroBonder]

You can check if both storages are compatible with esxi5.1 here -- www.vmware.com/go/hcl

To iscsi you need configure vmkernel port to acess lun, you already have this configuration ?

Discussion moved from VI: VMware ESXi™ 3.5 to VMware ESXi 5

[seanhaynes]

Thanks for the response, yep all I beleive is configured correctly - it can see the LUNs and connects, then randomly drops out, then reconnects, then drops out etc.

[MauroBonder]

Ok. Try check in HCL if is compatible, and which kind of policy failover is recommended to use with this storage, if wrong configured, can have problem. Both you can see in hcl

[AndreTheGiant]

In the HCL there is also the minimum firmware level. Does it match yours?

[DavidKlee]

Do you have any unusual activities in the network switch logs or the SAN logs? I have noticed this sort of activity on an IBM version of the same device, and it was occuring under both vSphere 4.1 and vSphere 5.0. I have another client with a Dell MD3000i on vSphere 4.1 and we do not notice any of the same errors in the logs there.

Are the two iSCSI VLANs independently routable?

Is round robin configured on each datastore?

Are either of the controllers having issues? Could Dell do a full log review and tell you if anything looks strange?

[seanhaynes]

the only error showing on the SAN MD3000i log is that is says ' iscsi session terminated unexpectedly' - the switches show no errors.the 2 vlans are complletely seperate. Under vsphere 4.1 it worked like charm. I did have RR configured but wnet back to last path used etc.

I've done a log dump on the MD3000i, there are no issues other then a failed battery that was replaced a couple of months back and it is running the latest firmware.

I raised a ticket with VMWare and they say it's a Dell issue as they haven't provided the 'software' support, aka drive issue I suspect.

Unfortunately I can't roll back to 4.1 as I updated the virtual hardware in the VMs so it looks like I'm going to have to invest in another SAN.

Annoying really as one of the selling point's of going virtual is the ability to eak out a longer working life from equipment that otherwise would have been retired; it's starting to be a very expensive exercise. If in the next version of vSphere Dell  / VMWare drop the MD3600i that I think we'll have to move back to one box one server - at least then you know it works.

That said someone did say to me to have a look at the MS offering which I might just do - I can't risk the network on a whim of someone at VMWare or Dell not being bothered to support paying customers.

[cjraiter]

Yikes!  Before you go back to the one server/one box method, I'll see if I can help.  I have the same setup with ESXi5 U1 and MD3000i (and also an MD3200i) that we haven't had any production issues with.  This configuration served us all the way back from ESXi 4.0 U2. 

I have 2 switches dedicated to iSCSI only, no VLANs.  On the MD3000i controller 0, port 0 is subnet A, port 1 is subnet B.  Controller 1, port 0 on subnet A, port 1 subnet B.  From the ESXi Host, there are 2 iSCSI vSwitches, one with a VMK port configured for subnet A and the other configured for subnet B.  For simplicity's sake in this discussion, there is one pNIC per vSwitch so I have 1 path to each physical iSCSI switch.  RR is set up for each datastore, so for the MD3000i I have 4 paths per LUN (2 Active, 2 in Standby).  The only time I ever have to redistribute the disks on the MD3000i is when a switch would reboot, which leads me to...

Check the firmware on your switches.  We had a time a few months ago where the Dell 5424 switches would spontaneously reboot for no obvious reason (no errors were logged).  Updated the firmware to the latest version and no more reboots!

Hopefully this helps you out and maybe there's something in my config that leads to a solution for you.  I don't think Dell or VMware is going to be dropping support for the MD3600 anytime soon.

[Datto]

I don't know if you've seen this thread but it might be of interest to you:

http://communities.vmware.com/message/1830517#1830517

Datto

[seanhaynes]

Thanks for that - essentially the only difference is that I do use vlans to seperate iSCSi traffic - out side of that server side one nic to one vSwitch etc.

I am running the latest firmware on the MD3000i - like I said I also have a 3600i that doesn't suffer any issues.

[seanhaynes]

Yes I had read the blog thanks - it seems definately a bit hit and miss as to wether it works or not.