Replacing Midplane, controllers of MD3000i

The intermittent issue can be network related and not specific to the Host/Enclosure. Have you checked the network integrity? Do you have management and iSCSI traffic on the same switch? It is advidsable to work with Dell Tech support to resolve issue prior to replacing Hardware.

It sounds like you have a case open with Dell support. What did they tell you when you asked them?

I would think it can be done without losing any data, but raid controller firmware may play a role as the 07-series and 06-series firmware don't mix and you cannot downgrade from 07 to 06, so if you're on 07 right now, and get controllers with 06 series firmware, the results may not be 'optimal'.

I agree with MustKnow that you need to verify the network settings and the network environment before replacing the controllers and midplane. Especially for something like intermittent connections.

That aside, I would also highly recommend that you capture a support bundle through the "Gather Support Information" link under the support tab of the MD Storage Manager. If you are experiencing connectivity issues and are unable to do this, you could also attempt to save the same through the CLI ( <ADMIN NOTE: Broken link has been removed from this post by Dell> for more information)

-Mohan

Hi there,

Thanks for your replies, I have opened a case with Dell, and we are looking into the network switches as a first step.  The reason I wanted to find out about replacing midplanes and controllers is this machine is setup in a VMWare environment with our critical infrastructure on it.  Another user on the vmware forums appeared to have the exact issues I am having, having recently moved to Jumbo Frames, whereby the LUNs intermittedly drop out for no apprant reason.  Their solution was to replace the midplane and both controllers.

Since this has our critical infrastructure, I wanted to find out what the process is in the worst case scenario so I can prepare for that, while going through the process of troubleshooting each component of our iSCSI array.  Hopefully I won't have to go as far as needing to replace the midplane and controllers for this MD3000i.  Mind you, the person I've been talking too has stated that he thinks the reason why he had to replace so many components was partly his fault, because he believes he had a faulty controller.  They sent him out a replacement controller, he popped it in, and didn't give it enough time for it to sync with the inplace controller and started reconfiguring it. 

Thanks again for your replies, and hopefully it is the switches at fault in this instance.

Gene

Which version of ESX/ESXi are you using?

ESX/ESXi 3.x doesn't support jumbo frames on iSCSI... oddly enough it does support jumbo frames on all other communication options (NFS, VMotion link, and even VM traffic (source: VMware KB)).

If you're using ESX 4.x (a.k.a. vSphere), did you use this guide to the letter to set up your systems?

Hi there,

We are using ESX4 and I didn't use the guide, but it has the same exact setup that I have in our system (i.e. two subnets, Jumbo frames).  Interestingly, and as noted in the comments in that thread, we're experiencing performance problems with Round Robin and Jumbo Frames.  So we changed to MRU and Jumbo Frames.  I'm thinking its due to the Broadcom interfaces we're using.  I did use that guide to the letter when it came to setting up Jumbo Frames.

With regard to the problem at hand, I suspect that it is related to our switches, and that we enabled Jumbo Frames and Flow Control on the switches at the same time.  We aren't using very high end switches (Dlink DGS-1224T switches), and I'm thinking that we're getting problems with the switch buffers running out of memory, causing the link and packets to be dropped.  This is a theory, and I've disabled Flow Control for the time being, and we'll monitor over the weekend.

We're also considering dumping the whole Jumbo Frames and moving back to standard frames, since we've had nothing but problems with Jumbo Frames, and we've gained zero performance with them.  Not that we were experiencing performance problems beforehand anyway, and beforehand everything ran smoothly for over a year. 

In the end we'll see how it goes over the next few days.  Thanks again for your help.

Gene

Gene,

I checked the support Matrix  from Dell MD3000i for your configuration and the switch you mentioned is not listed.

Thanks for the reply,

According to the support matrix here: http://support.dell.com/support/edocs/systems/md3000i/en/SUPPORTMATRIX/Sup_Matx.pdf

The matrix states that any industry standard 100/1000 managed or unmanaged Ethernet switch is supported with the MD3000i.  I personally would have hoped so, because iSCSI just requires ethernet, so if the MD3000i is industry compliant then it'll be able to use anything ethernet.

In saying that, we've actually reverted back to standard MTU frames in hope that it will solve our problems.  We're in a monitoring phase at the moment, and we'll see how it goes.  Regardless of whether this works or not, we're budgeting for Cisco 2960's to replace these switches, in order to give us better insight, and help us more easily track down issues like this in the future.  The question is whether we buy this within the next week (if reverting the changes back doesn't fix the issue at hand), or next year (if it does).  Ideally it would be next year, because accountants don't like suprise purchases. 

Thanks for your input!

Gene