Areca 1880i Noob....help!

P

PhoenixC5

Limp Gawd
Joined
Jun 7, 2004
Messages
335
So I tried posting this in the areca thread, but I think it got lost in the shuffle. So here we go again with a slightly different post:

I ordered the remaining items for my server and they should all be here in the next day or two:
Norco 4224
Areca 1880i
4 x 2TB Hitachi 7K3000 Drives
Cables/Etc.

So...after reading through pages and pages of the areca thread, I think I have the configuration settings close and I now have several questions.

The settings:
128k Stripe
64-bit LBA
Write Cache Enabled
Foreground Initialization (I don't have the BBU, but I will have it hooked to a UPS)
Format using NTFS with a 128k cluster.

Do my settings look reasonable?
What other things should I do to prepare?
Should I update the firmware on the Areca card?
I plan on using one of the X16 slots on the board (leaving another x16 slot open). Are there issues with this?
I plan on eventually adding an HP SAS expander. Will these drives move to the Expander later leaving the array intact?

I will be using the stuff in my server for Hyper V stuff primarily:
Core i7 950
Gigabyte X58-USB3
Windows Server 2008 R2

Any problems I should watch out for in this configuration?

Help an Areca noob! :)
 
Do my settings look reasonable?
- Look standard, nothing wrong with that

What other things should I do to prepare?
- not that i can think of

Should I update the firmware on the Areca card?
- latest is 1.49, but with the main change being added 3TB support, if you are on 1.48 I guess you are fine too.

I plan on using one of the X16 slots on the board (leaving another x16 slot open). Are there issues with this?
- no

I plan on eventually adding an HP SAS expander. Will these drives move to the Expander later leaving the array intact?
- yes
 
Awesome. Thanks!

One more question.

In the Areca thread, I see that Stagger Power On Control should be increased for the 7K2000 drives. I assume this is also true for the 7K3000 drives. Is that a good assumption?
 
PhoenixC5 said:
I will be using the stuff in my server for Hyper V stuff primarily:
Core i7 950
Gigabyte X58-USB3
Windows Server 2008 R2
Click to expand...

If you're going to spend $180 on a motherboard and $300 on a CPU and you're going to be running Server 2008 R2, with a lot of Hyper-V, and above all else a STORAGE SERVER, I wouldn't be buying a Gigabyte motherboard optimized for gaming (three x16 video slots) because I read a lot about issues with certain non-RAID HBA's on other forums - albeit usually about LSI cards - and I would also be buying a Xeon CPU. If your budget were my money there would be no other choice than a Supermicro motherboard, something like an X8SIF, or if you like more slots then an X8SIA-F, and something like a Xeon 3440.

Everything else about your build looks great. You'll want a separate volume for your VM files though. Keep those separated from your big storage volume. And maybe a single fast SSD as your boot volume.

PhoenixC5 said:
Format using NTFS with a 128k cluster.
Click to expand...

No. A larger cluster size would not be a performance multiplier in your scenario. I generally recommend 16k for people with 12-24 x 2TB drives in a single NTFS volume. That's what I have on all my volumes. Look at the following table and ask yourself how big you could see yourself growing any single volume, then use the appropriate value.

ZgsFo.jpg
 
Last edited:
PhoenixC5 said:
In the Areca thread, I see that Stagger Power On Control should be increased for the 7K2000 drives. I assume this is also true for the 7K3000 drives. Is that a good assumption?
Click to expand...

No, that thread was started in the time of Areca 1680 + earlier versions of Hitachi 7K2000 firmware. With an 1880i and 7K2000 drives I'm able to use the default 0.7 and never see false-positive "timeouts" in the event log which used to require a hard reboot. I expect the situation to only have improved on the 7K3000/5K3000 since it's newer controller firmware. You should be fine with the defaults. Here are my power control settings for each of two Areca 1880i's, each with about 44 x 7K2000's attached to them.

5DcIM.jpg
 
Last edited:
One more thing to add! And, IMHO, this is very important!

NEVER mix RAID levels within same RAIDset. Areca allows you to create more than one volume set within the RAIDset and certainly allows you to choose different RAID level for each volume set (one volume set could be configured as RAID10, while the other as RAID5 or RAID6). Most likely it will work fine for many years to come. BUT, in case if you decide to expand the RAIDset by adding more drives to it you may end up having your RAID10, and possibly other volume set(s), screwed up beyond recovery. I'd recommend having all volume sets within the RAIDset configured with the same RAID level, be it only RAID6, RAID5 or whatever else you choose.
 
Good advice. I'd even say keep the raidset to volumeset relationship 1:1. Just because Areca allows you to create multiple volumesets per raidset doesn't mean it's a good idea.
 
Am i the only one that is intrigued by 128kb NTFS cluster size? I thought 64kb was the max

@echo What Jus said
 
I would echo Odditory's suggestion about the motherboard. If you plan on adding an HP SAS expander at a later time it is fairly picky about motherboards. By and large most of the 'server' motherboards from Intel, SuperMicro etc WORK with the HP SAS Expander but 'consumer' motherboards are absolutely hit or miss. It's a pain in the ass to swap out a motherboard (if you can avoid it) so getting a motherboard that is known to work with the HP SAS Expander can really save you a lot of trouble down the road.
 
I've tried the Areca cards in 5 different motherboards so far. 2 were server class and 1 we'll call prosumer and 2 were gaming boards.

I got the Arecas to work in all of them, in fact, sometimes they worked for 2 years before having an issue..but every single non server class board exhibited some issue sooner or later. Either very, very long firmware startup times, or the card failing the pci bus test, etc.

As above guys said, if you care at all about your data, get a server class board. And you cannot go wrong with a Supermicro. I wish I had purchased one originally when I got my first areca 3-4 years ago.

Also as pirivan said, if you add an HP Expander it will *definitely* have issues on consumer class boards. I tried one in all of the boards listed in my sig, and it only worked reliably on the supermicro and the poweredge, that's it.
 
I have both SM and Asus Server boards and I really like my Asus too. They leave old cpu micro code in their BIOSs so you can run ES CPUs. (ie. Cheap Xeons on ebay).
 
Jus said:
One more thing to add! And, IMHO, this is very important!

NEVER mix RAID levels within same RAIDset. Areca allows you to create more than one volume set within the RAIDset and certainly allows you to choose different RAID level for each volume set (one volume set could be configured as RAID10, while the other as RAID5 or RAID6). Most likely it will work fine for many years to come. BUT, in case if you decide to expand the RAIDset by adding more drives to it you may end up having your RAID10, and possibly other volume set(s), screwed up beyond recovery. I'd recommend having all volume sets within the RAIDset configured with the same RAID level, be it only RAID6, RAID5 or whatever else you choose.
Click to expand...

Wow, this is the first time I Have heard this.

I do this because I mainly run raid6 on the vast majority of my storage but for MySQL I do raid10:

Code: 
# cli64 vsf info
 # Name             Raid# Level   Capacity Ch/Id/Lun  State
===============================================================================
 1 OS RAID6           1   Raid6     64.0GB 00/00/00   Normal
 2 DATABASE RAID10    1   Raid1+0   30.0GB 00/00/01   Normal
 3 DATA RAID6         1   Raid6   8891.0GB 00/00/02   Normal
 4 SSD RAID0          2   Raid0   1032.0GB 00/00/04   Normal
===============================================================================
GuiErrMsg<0x00>: Success.

Of course I never use nor plan to use the OCE. I am glad to find out it doesn't work right with different types of raids in the raidset though. That seems like a pretty major bug to me.
 
TBH you would probably be much better off if you would separate those out some. Do you not experience really bad disk contention with all three volumes on the same raidset?
 
houkouonchi said:
Of course I never use nor plan to use the OCE. I am glad to find out it doesn't work right with different types of raids in the raidset though. That seems like a pretty major bug to me.
Click to expand...

I think something got lost in translation there. I don't think Jus was implying that there's an inherent bug with mixing raid levels of multiple volumesets within a raidset. I think he was saying it from the standpoint of a best practice, that its wiser to "keep it simple" and don't do it, for the sake of things like recoverability in the future. At least that's what I got out of it because that's how I see it.

Example, let's say a raidset dies or gets corrupted for whatever reason, and for whatever reason you didn't have a backup, or let's say the raid card dies and you can't get a replacement fast enough. trying to recover data from a bunch of disks with a mixture of blocks of different volumesets of different raid levels, well lets just say doing nuclear fusion might be easier at that point. if you stick to one volumeset per raidset, or at least one raid level per raidset, then you know the block pattern of the data on each disk and have more options in a recovery scenario.
 
Last edited:
Thanks for all of the suggestions. I have considered using a Super Micro motherboard. But, at the moment I have all of the parts put together and ready to go, so I might consider switching to the X8ST3-F (basically the X58 version of the X8SIA-F)
http://www.supermicro.com/products/motherboard/Xeon3000/X58/X8ST3-F.cfm

The only reason I did not go with this board initially is that I had heard it was quite picky about memory. I already have 24GB of RAM and really would prefer not spending the money to change that all out. When I decide to upgrade the NIC I may swap out to the Supermicro and see if it will work with my memory. If so, I can just sell off my Gigabyte board and be happy.

As to the comments that the HP SAS Expander has compatibility issues with motherboards....I have to ask how? As I understand it, the card just using the PCI express bus to supply power and that's all. Is there a power issue?

I plan to keep most of my heavy file sharing things on a separate array of Seagate 2TB drives on the built-in ICH10R. It may suck at write performance, but for basic file sharing for media, it reads at 500MB/s....IE Plenty!

Thanks again for all the great input and suggestions!
 
Also for your file share, if the array does 100mb/s it will be plenty because you will hard pressed to push it any faster over the network.
 
PhoenixC5 said:
As to the comments that the HP SAS Expander has compatibility issues with motherboards....I have to ask how? As I understand it, the card just using the PCI express bus to supply power and that's all. Is there a power issue?
Click to expand...

On the DFI Lanparty NF4 board, if the areca and the expander were filling both PCIe slots, then the board would power up in a strange way. It would have to scan the PCI slots multiple times in order to boot, often with both cards in the slots the Areca would fail the PCI bus test and not boot at all.

On the Tyan K8 board I had an Areca in there for 2 years, and one day out of the blue I had issues with the card not passing the PCI bus test 1 out of every 5-6 reboots.

In others, the PWRGD signal is wrong and won't turn the HP card on properly without taping over the card pins, and it's not the best idea to leave that floating is my understanding.

Also I had issues with forcing the consumer boards to boot off of a PCI video card reliably when both PCIe slots were taken.

I think the bottom line is that they are really designed to have a video card in those slots, and while you can get other things to work, I would never trust my data to that scenario. It's bad enough to have to use consumer level drives and non-ECC ram and non xeon processors.

I put that OCZ Gold in my Supermicro and it seems ok so far.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227365
 
It is always tempting to build a rig out of cheap components. People tend to think - Oh, if I could get same or better performance out of my Core i7 CPU, why would I need to buy a server grade motherboard and slower CPU(s). Why not use what I already have and add a RAID card into the mix? This way I could get the best out of both worlds - insanely fast computer along with a huge capacity storage. Wow!

Wrong!!!
One has to realize that there is a reason for existence of more expensive hardware. Besides stability and reliability, server grade hardware delivers better data integrity, where use of ECC Reg memory is a must. The old saying applies here too - you get what you pay for.
 
Last edited:
For now I will have to work with the hardware I have. If I experience issues, I will likely upgrade when the budget allows. Most likely I would go with the X8STE and the W3550.

Also, I do have the OS loaded on an SSD. Assuming my two UPS packages that are apparently lost in Dallas somewhere (like 5 miles from me), I'll get started on swapping the case out and then trying out the new 1880i.

Thanks again for all of the input!
 
You must log in or register to reply here.
Back
Top