FReeNAS in SMB?

[ghostlobster]

Hi all,
I've been running FreeNAS 8.3 as my home storage solution, as well as storage for my personal business for a while njow, and apart fomr the occasional hiccup, I like it. However, the load on the server is minimal, as it's just me and my wife here at the house.
I've got a customer who is looking to move away from their aging Windows Server 2003 configuration and they are looking to me for recommendations. We've got about 20 users and about 1.5 TB of data. The data is read from and written to pretty regularly throughout the day so performance bottlenecks cannot be an issue. right now, I've got a Thecus N4100Pro NAS server in place there, acting as a file server for their pictures. I put this in place last year as a proof-of-concept for them regarding NAS solutions over a full blown server. It's worked very very well, and will probably remain in place in its current capacity even after implementing the server replacement.

Anyway, I'm thinking about building them a FreeNAS solution to replace their server now, as all their server does is store/serve files. They will be using it for their day-to-day file storage, QuickBooks file storage, and as a storage location for some shared SQL data (not as a DB server, just as the shared storage location.) In this environment, I'm thinking about putting 3 x 1TB WD Red drives, setup in a RAID Z1 configuration to start with. 8GB of RAM, a Core i5 sitting in a Gigabyte mobo with 6 SATA III ports. Would this config be reliable and robust enough for 10 concurrent users hitting on it throughout the day? Could someone who uses FreeNAS in a business environment post their thoughts, what they would do differently, what works best for them, etc?
Thanks for the help!

 

[ben]

With ZFS, it's always better to size for what you'll EVENTUALLY need, not just what you have now. 2TB useable for 1.5TB current files is fine, but what about three years from now? I recommend 6 1TB (or 2TB!) drives in RAID-Z2, for more protection and more capacity (also, ZFS slows way down starting at about 80% capacity.) Given that size recommendation, I also recommend scaling up the RAM (16GB should be reasonable).

Also, make sure to talk to them about backups. RAID arrays are NOT backups. Snapshots on the same box are NOT backups. Until and unless you've actually confirmed the integrity of a backup (ie restored a database to a version from a remote backup) your backups aren't actually working.

 

[cyberjock]

Ben hit the nail on the head. If you are looking to expand later(which I don't recommend because of what Ben said) then you should oversize your case, hard drive controllers, etc NOW in anticipation of adding more drives. Something like the Norco RPC-4224(that's the one I use) is a good fit and is inexpensive for what you get. Adding drives to an existing zpool requires alot of thought and planning. Read the guide I wrote(link in my sig). That will explain why you can lose all of your data when "expanding" a zpool if you don't know exactly what you are doing. Keep in mind alot of people have made mistakes when trying to add more disks later. We had 2 this year that added the drives wrong and had to destroy the zpool and start over.

 

[ghostlobster]

Thanks for the reply. Yeah, the difference between redundancy and backups has been explained to the customer in earnest and I'll be doing differential RSync backups offsite.

The capacity is a variable for us here. I have no problem tossing more storage space in as we go, and will definitely look into bumping things up to get started. I'm curious about the recommendation for Z2 vs Z1. If you don't mind me asking, why?

Are you using FreeNAS in a business environment? Just curious.

 

[ben]

RAID-Z has the same weakness as the normal single-parity RAID level, RAID 5: When rebuilding, there is NO parity at all. Given that all disks are usually from the same batch and have been under the same wear for the same amount of time, this should make you sweat. Particularly because rebuild times are pretty long these days as disk sizes increase.

Related: MAKE SURE there's a scrub scheduled and performed regularly. That's what actually protects against data corruption. The default (every 35 days) scrub set up in older FreeNAS versions was broken (this is being fixed in 8.3.1).

I work for the developer of FreeNAS, iXsystems. We also a sell a commercial appliance, TrueNAS, which is based on FreeNAS. I've been involved long enough to know the recommendations we make to SMB customers who come to us for FreeNAS and TrueNAS alike. Some of this stuff is relevant even to the smallest home users (like the volume size... it's a HUGE pain to increase volume size after the fact because either you have to rebuild from scratch or you end up wasting more in parity than you needed to while reducing reliability).

 

[ghostlobster]

Thanks a bunch! I totally learned something here today. I was not aware of the issues surrounding increasing the size of a ZFS volume until reading all of this...especially the PPT file in cyberjock's sig. Ugh...this could create a problem for me here at my home with my setup! Right now I've got 4x1TB drives in place, running RAID Z1. I have a 5th drive in there, unused, just in case one fails, I can just do a replace drive from within the GUI before pulling everything apart. And all drives have labels on them with the SN in BIG LETTERS so I can easily identify the bad one if/when it happens.

However, I was working under the impression that I could just slap another 1TB drive in there and add it to the ZFS volume, and presto...more space added to the volume, profit! Oops! Looks to me like when the time comes to increase the storage, I'll be getting 4x2TB or 3TB drives, and swapping out my 1TB drives one at a time.

Would a scheduled scrub once a week be overkill?

 

[ben]

Once a week is more than most people do, but if it's at a time when no access is happening (2am monday morning for example) there's probably no harm in it. DEFINITELY scrub right before one of those planned disk-size-upgrades, since those are effectively deliberate drive failures.

 

[ghostlobster]

I'll schedule it for the 1st and 15th of each month then off-hours. That's easy enough to do.

And the disk upgrades will not be happening for a while, but it's definitely something that will happen within the coming months here at home. With the customer, however, I'm going to oversize the heck out of the implementation. Probably 4x2TB in RAID Z2.

 

[ben]

4*2TB in RAID-Z2 is still only 4TB usable. If the customer's growth rate is fairly low that's probably fine for a while, though.