ESXi, Dell Poweredge SC1425, RAID Controller add on suggestions

Berkson Wein
Berkson Wein used Ask the Experts™
on
Hi all-
We're (stuck) using two or three Dell Poweredge SC1425 servers as ESXi 4.1 machines.  I'd like to get a hardware raid controller in there that will be recognized by ESX(i) 4.1.  

The add on slot is a 3.3-V, full-length, 64-bit, 133 MHz PCI-X
 
Budget is everything here.  They just don't have $200+ to spend on high quality adaptec.

Anyone have experience with this: http://www.overstock.com/Electronics/SYBA-PCI-X-4-port-SATA-II-RAID-Controller-Card/4381430/product.html  ?  I don't even know if this is hardware or software raid, let alone something that would work with ESX(i).

Looking for suggestions of cards to consider, cost, places to buy used, etc.  

This install is for a charity (fully documented and registered), so if there are sources that might be willing to donate the hardware, that's even better!

Thanks
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Always use the VMware Compatibility guide when trying to procure hardware.

http://www.vmware.com/resources/compatibility/search.php


Berkson WeinTech Freelancer

Author

Commented:
Thanks arunraju for your reply.  I should have put in my original post that I'm aware of the compatibility guide and that I kknow that the SC1425 isn't on it.  Of course there are also tons of posts that talk about perfect installations on the SC1425 without issue, not using the built in SOFTWARE raid of course.
I know we can get ESXi working on the SC1425, but want a very inexpensive compatible SATA raid card that'll work in the PCI-X slot of this machine.  The HCL only lists "certified" and expensive cards.  I'm looking for a low cost acceptable card for them.
Specifically, I want to know if the SYBA PCI-X 4-port SATA II RAID Controller Card that's on overstock is hardware or software raid and if it's hardware, if it'll work with ESXi.  I don't care that it's not on the HCL and that it'll be unsupported by vmware support staff.   My gut says that it's software raid, so useless, but I can't find anything definititve.
 
 
 
 
President
Top Expert 2010
Commented:
Speaking as an architect who has been writing storage & RAID diagnostics, firmware, subsystem code since late 90s, let me give you a qualified, and blunt response.  You risk data by using hardware not on the HCL.  Due to NDAs and such with many of the manufacturers, I won't give specifics on certain chipsets, but I assure you, there are a lot of ways you lose data.   I've seen writes completely disappear from I/O queue, so they never get written to disk. I've seen an unrecoverable read error cause timeout that takes perfectly good disk offline, and initiates a rebuild, and a few minutes later the same problem cascades to a different disk in raidset.   100% data loss due to controllers that expect a disk to time out on reads within X seconds, but the disk takes X + 3 seconds.

The HCL means that hundreds of man-hours (sometimes much longer) has been spent making sure you have 100% data integrity, not only when things are working fine, but during error recovery scenarios.

There is a huge problem out there when people report that something "works", they don't have the experience, knowledge, or equipment to test failure scenarios.  ECC errors, unrecoverable read errors, XOR/PARITY errors can easily cause an entire stripe to be lost forever, or worse, 100% data loss.   Pretty much everything "works" that isn't on the HCL.  But these same things that "work" in an error-free scenario, can and do have catastrophic consequences during failure scenarios.

Don't risk your data.   If you want to save a buck (which we all do), then do not buy new, buy HCL gear used.   It is still a buyers market for used gear, and there are a lot of bargains.   You can pick up a used LSI SAS/SATA PCI-X card on ebay that is on the HCL for a good price.   The SYBA product is a toy, and not appropriate.

Also, do NOT get the consumer/desktop class disk drives.  Some of my best fees come from people who make this mistake.  Enterprise class drives have at least 10X better ECC, but more importantly, they handle error recovery scenarios within the appropriate time frame.  Desktop class drives are fine for some software-based RAID, but in general are unacceptable for most hardware-based RAID controllers.

Berkson WeinTech Freelancer

Author

Commented:
dlethe- thanks for the candid reply.  All stuff I knew, BUT reading it so directly sealed the deal for me.  Three is no budget, but I won't let them skimp either.  Appreciated.
DavidPresident
Top Expert 2010

Commented:
Thx - It really is pretty bad out there with people saying something "works", and they never tested any failure scenarios by even just yanking a disk in the middle of I/Os..  (Let alone, yanking disk when I/O queue is FULL -- rookie mistake by highly experienced storage people, but nothing you think about unless you've done time in a testing lab);  or programmatically generating ECC errors and/or unrecoverable read errors in singles, or groups, or groups with multiple threads.

Let's face it, don't trust anything that isn't in an HCL.  If something is in the HCL, see specifically which disks and firmware they use.  Sometimes firmware is VERY important.  That is why people pay the big bucks for NetApp, EMC, and others.  I've got first-hand experience with the amount of testing they do and have sat in on those meetings (with one of these vendors), talking for hours about just one firmware bug that has potential to cause data loss and seeing a man-year spent on one of them by a team.   While same bug is out there, and affects other controllers, and I know that they just ignore it.

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial