C6300 Disk partitioning...

Hello

1.  Can anyone describe how one allocates hard drives in the 24-HDD collection to each compute block?  Are there any limitations to what can be allocated to each block?

Each slot is assigned to a specific node. For example, on the 12x3.5" backplane the first 3 slots on the left are assigned to node 1. There are diagrams in the manuals that list module/node assignment of the slots:

www.dell.com/support/home/product-support/product/poweredge-c6300/

2.  If we create a RAID partition and allocate it to a specific compute block and that compute block subsequently dies, is it a straight-forward thing to import the RAID partition with another compute block?  Are there any issues to consider when doing this?

Drive assignment is not dynamic. You don't allocate disks or virtual disks to a module. You would need to physically move the drives to the slots assigned to another compute module, or you would need to move the other compute module to the slot assigned to the drives you want to use.

Thanks

I was told by a Dell Support person some weeks ago that I could configure the C6300 to connect a single compute node to a maximum of 12 of the 24 2.5-in hard drives if I so desired.

They are correct. I overlooked the expander backplane option. If you order the system with an expander backplane it can be configured to use presets of either 2x12 or 4x6. Those are the only two presets I see available. The other available backplanes are direct backplanes and operate as I described above.

Thanks

Daniel,
           Thanks for the quick response.  

I was told by a Dell Support person some weeks ago that I could configure the C6300 to connect a single compute node to a maximum of 12 of the 24 2.5-in hard drives if I so desired.  It sounds like what you are saying is that equal allocation (6 drives each node) is the only option.  Is that correct?

Daniel,
           I've tried for a few days now to get a Dell part number for that backplane expander which you mentioned.  Is that something that you have available and which you can help us with?

Thanks,

raidnoob

I've tried for a few days now to get a Dell part number for that backplane expander which you mentioned.  Is that something that you have available and which you can help us with?

It is not a part that you add. It is one of the backplane options that the system is ordered with. Changing the backplane after point of sale is not supported.

One of the part numbers for the expander backplane is YDWNX.

Since changing the backplane APOS is not supported I'm not sure if that is the only part you will need or what steps may be involved. If you attempt to change the backplane it will be trial and error.

Thanks

Daniel,
            Thanks once again for the feedback.  We are becoming quite confused by this question, so please indulge me with one more line of inquiry.

When I look at the Dell C6320 specs online (http://i.dell.com/sites/doccontent/shared-content/data-sheets/en/Documents/Dell-PowerEdge-C6320-Spec-Sheet.pdf), I see that in the section titled "Flexibly mix workloads in one chassis", the claim is made that, with the backplane extender installed, one can zone any number of drives with each compute node to achieve total flexibility.  

Based on our conversation to date, does that statement make any sense to you or is it simply marketing hype?  Is it possible that it is another backplane extender that is under discussion in that spec document?

Thanks.

I have very little documentation on the expander backplane, but what I have states that there are only two available splits(4x6,2x12). Most of the documentation refers to the zoning utility for more information, but the zoning utility does not contain any documentation.

The zoning utility allows drive assignment beyond what is listed as available in other documentation. I suspect the utility is used in other platforms, and many of the options in the utility are not supported. Due to the limited information available and conflicting information I will need to submit for more information. This could take some time. I will update you when I have more information, it will likely be next week.

Thanks

Daniel,
           Is there anything to report on the question of the flexibility of the C6300 disk allocation?

Thanks.

Is there anything to report on the question of the flexibility of the C6300 disk allocation?

No, I escalated it last week and received the same information I had already supplied to you. It has gone up another level to try to get the information.

There is ambiguity in the documentation, and I'm trying to get something that is clear on the zoning options for a PERC and an LSI branded controller in the system. I'll update you when I find something, I have no ETA.

Thanks

Daniel,
           It sounds as though zoning is a capability of the controller.  I don't know if this helps, but when Dell quoted the C6300 system with 2 compute nodes to us, this was the entry related to the controller:

SKU                    Description                                                                           Qty

470-ABXE           C6A, LSI 2008 Controller, C6320 1U Expander BP,            2
                            Supports up to 12x2.5 in Hard Drives

Perhaps, the promised flexibility is that you're limited to a maximum of 12 drives (1-12) per node with the expander BP installed and a maximum of 6 drives (1-6) per node without the expander BP installed.  I could see the marketing folk making that sound like much more flexibility than it really is in the technical specs documentation.

It sounds as though zoning is a capability of the controller.  I don't know if this helps, but when Dell quoted the C6300 system with 2 compute nodes to us, this was the entry related to the controller:

SKU                    Description                                                                           Qty

470-ABXE           C6A, LSI 2008 Controller, C6320 1U Expander BP,            2
                            Supports up to 12x2.5 in Hard Drives

I discussed the case internally today. It has been recognized that there is a lack of documentation on how the zoning feature of the expander backplane functions. The process has begun to have documentation created and published. This will likely take a long time, I have no ETA.

In regard to your questions about functionality, you cannot choose which drives are assigned to a node. You choose the number of nodes/modules in the system and the number of drives assigned to each. It assigns those drives in sequence to the modules. Your idea of assigning drives to a different module in the event of a module failure would only work in limited configurations/scenarios.

I don't have details on any limitations to the number of drives that can be assigned to a single module or what/if any differences exist between a PERC and LSI branded controller. It is assumed that 12 is the maximum number of drives that can be assigned to a module.

Thanks

Daniel,
            Thanks for all the follow-up on this matter.  I appreciate you going the extra mile on our behalf.

Best regards.

Austin Pitt

There is a lot of information that is left out of the documentation for the C6320 and C6300 chassis.

I would also like to know the capabilities of the C6300 chassis.  One sell sheet says, "With the expander backplane option, dynamic drive assignment for 2.5" drives provides even more flexibility by zoning any number of drives to each node, giving nodes additional or fewer storage resources as workloads demand."

This thread discusses 2x12 and 4x6 modes, which sound different from the description of the expander backplane, and would make sense if a chassis was configured with only 2 sleds (also allowing a C0TG2 riser to breakout both PCIe slots).  In such a case, a 2x12 configuration would be desirable so all the drive bays would be used.    But the question is how is this controlled? Is there any documentation that references 2x12 and 4x6?  This does not allow the flexibility described in the expander backplane option which clearly says "any number of drives" can be assigned so I am hoping this is a standard backplane feature.

For others who are looking to answer questions on this topic, be advised that there is a "C6320 HDD Zoning Configuration Utility" available under C6320 downloads.  My guess it that this controls the expander backplane option only, but perhaps it also can change 4x6 to 2x12?   That download info says, "This tool is specific to an 2.5" HDD expansion config" and "This tool changes what drives are seen by what nodes."   It says it applies to "CRD,BKPLN,HD,2,5,EXP"

I found some information on the expander backplane option in the manual, and I have been looking at the C6220 and C6420 manuals to see if they describe how those systems worked for this type of option.   In particular, the C6220 (II?) "expander" backplane may be the same part number (currently my best guess is D97YM on that).

So the way this apparently works is probably similar to the connection to an external drive shelf.  Instead of the rudimentary "1 cable to 1 drive" fanout from the mini-SAS (in the standard C6300 chassis), the HBA or RAID card can control a bunch of slots over just a single mini-SAS cable.  Even a single LSI-2008 apparently can control up to 3000 devices up to 290,000 IO's per second according to a data sheet... and multiply that by 4 of them in a C6300 chassis with uniformly configured 4 sleds. 

SFF Backplane Expander Configuration 

How to remove/install the backplane expander

Cable Routing from Mid-Plane to Backplane Expander Controller Card

The links above show a VERY different backplane for this option along with a connector for an "expander card" controller at the bottom.  The card only has 4 mini-SAS connectors on it, whereas the "standard" backplane has fanout cables from the midplane to each slot in the backplane.  So that whole rats nest of cables behind the backplane can be yanked out.  Very nice.

Aside from the 4 mini-SAS connectors, I think there may be one other connector on the expander controller card.  I am looking at online photos of part number R9J6V and it looks like that connector says "FCB Crtl to Expander Card" on it. The top link above references it as "power control connector" (#1 in the last diagram.) I don't see the cable routing referenced in the manual.

So far, it doesn't seem like it would be that bad to physically change the backplane, especially as the wiring appears to be greatly simplified in the "expander" setup.  But considering the lack of documentation and other information, I would give this about a zero percent chance of being as simple as it seems to pull off, with a very real possibility of hitting a brick wall with no way to find a solution.  It doesn't seem like Dell sold a lot of these units because answers and information can be hard to come by.

In particular, my guess is that the sleds need to be configured properly along with the chassis firmware and the firmware in smaller components like the expander board and even the FCB (not sure if that is "front" or "fan" control board).    I had all kinds of issues with the first C6300 chassis I had (search my early threads here) where the fans would run at full speed and I never was able to get that resolved.  I ended up swapping the chassis for another.