One brand new PE2850(A01) with Perc4/DC(latest BIOS)
One brand new PV220 with Dual EMM
Desired Configuration:
We would like to connect the PE2850 to the PV220 using two SCSI cables so that the failure of a single EMM in the PV does not interupt operation. We have set the PV220 in Joint-Bus mode and connected both SCSI cables, one cable between Channel 0 on the Perc4/DC and the primary EMM, one cable between Channel 1 on the Perc4/DC and the backup EMM.
Problem:
In this configuration, the server takes approximately 20 minutes to progress throught the Perc4/DC BIOS initialisation before deciding that the adapter is not present. If only one SCSI cable is connected the system behaves normally (Boots, initialises, recognises all drives..etc).
Is your internal cage connected to the PERC internally?
If not, I'd personally use split-bus to have 2 SCSI channels for 2x7 disks and give you better performance. Just make sure you create mirror-pairs before creating RAID 0+1. Basically, if one of the EMM goes offline, you'll have the mirrored one on the other channel.
I'm not sure what you mean by 'connected to the PERC internally'. I wasn't aware that was possible.
Our server, the DELL PE2850, has a Perc4/DC peripheral card installed in a PCI expansion slot. The Perc4/DC has two external high density SCSI connectors, one for each of the cards two SCSI channels. We have connected each of these two external connectors to an EMM module on the PV220 using external SCSI cables.
The PV220 contains:
- 2 * 72Gb disks (RAID-1)
- 5 * 146Gb disks (RAID5)
- 1 * 146Gb disk hotspare.
We simply wish to obtain redundancy in the case of an EMM module failure. I realise we could get better performance with a mirrored, split bus configurastion, but this would limit expansion of storage volume in the future.
Is there anyone out there who has implemented this configuration?
The reason I raise that question is because of two things: single SCSI cable connect works perfectly; and you've never mentioned about PERC4/Di.
It is possible for the PE2850 to have an internal connection with the PERC4/DC, (as show here <
ADMIN NOTE: Broken link has been removed from this post by Dell> ) and of course PERC4/DC card comes with external connections, too.
Then again, I've blindly assumed there is only one kind of Perc4/DC card out there, as I have not really been utilising with Dell kit for very long --- I am more familiar with HP Proliants and StorageWorks... Anyway, if a SCSI channel has been used internally, you can't use the same channel externally.
You were saying you understood the benefit of split bus but worrying about the problem of expandability --- Frankly, I don't think that is a worry at all. From the experience of the HP kit, I normally fully populate the disk array with best cost-ratio hard disks, but when I'm almost running out of disk space, I'd simply swap the old disks with bigger disks (of course, one at a time, so that the server can recover data onto new disks). Once it is all done, you can just expand your RAID set!!
Of course, I've never tried that on a Dell kit --- But as an Enterprise class storage solution, it will have to be able to do just that, don't you agree?
The problem with your proposed configuration is that both controllers (initiators) are at ID 7, so you get a SCSI bus address conflict.
The setup you suggested is not supported. It would require both controllers to have active paths to the drives, normally a clustering configuration, but Clustering requires communication between two servers as well.
You can get some performance benefit by using both channels in split bus mode, but this doesn't give the dual channel redundancy you are asking for.
guskwong
10 Posts
211
0
Posted October 14th, 2004 11:00
Is your internal cage connected to the PERC internally?
If not, I'd personally use split-bus to have 2 SCSI channels for 2x7 disks and give you better performance. Just make sure you create mirror-pairs before creating RAID 0+1. Basically, if one of the EMM goes offline, you'll have the mirrored one on the other channel.