Another new FreeNAS'r.. hello!

[Otroi]

My history that I always wanted to build a file server type of machine so I could consolidate my data into one location. I had everything strewn across multiple external hard drives with power cords and usb cords everywhere, it was quite the mess. I never had any issues with any drives failing (yet, I suppose) but I really wanted to eliminate the mess that these drives created. I had started building a system some years back but coincidentally just at that time my main pc's cpu or mobo failed and I ended up turning the hardware I had bought for the server into my new main pc and the project was dropped.

Fast forward to today.. I had already built another new pc and in the process had some of that old hardware laying around.. I figured I'd reuse the hardware to get my server built thereby minimizing the expense. Had a mobo, cpu, and ram and only needed a case and new power supply, and the drives. Everything's working well but I'm already looking at some changes.

My current setup is on an Asus P7P55D-E LX mobo, with a dead LAN port which prompted me to buy a cheap Rosewill (realtek) pci LAN card, no onboard video and my old card was dead so grabbed a cheap Radeon 5450 so I could see what I was doing, 8GB RAM, and a Core i3-540 chip, running on a Seasonic S12II 430w power supply. Then I started up with 4x 3TB HGST Deskstar drives and FreeNAS on a USB stick I had.

Having not done much reading about FreeNAS to begin with, I was under the impression that (like UnRAID, which I had planned on using years ago) if I wanted to expand my storage later I could just add some more drives and be done with it. I've recently found that's not the case and since I just bought two more 4TB HGST drives that were on sale, I'm evaluating my options.

After viewing @cyberjock 's presentation I saw that I can add these drives to a new vdev and add that vdev to the existing zpool (I hope I remembered the correct terminology for this context) but that seems kind of "messy". I don't want to introduce management problems by having to add on to existing data structures. If I have to, I'd rather rebuild the zpool with all the drives and start over again.

(Side thought: This seems like a silly limitation to me, you can't know what your future needs might be, so if you ever want to expand your storage you'd have to rebuild the entire zpool every time? That's not cool...)

I'm fortunate enough to have the data that's on the server now still on those external drives, not much "new" has been added. Only folder structures have been changed. So it's not the end of the world if I have to redo this from the ground up.

But now I'm also considering changing out the mobo/cpu/ram so that I have a proper ZFS server. I've read the arguments for and against ECC RAM (not so much against, rather just an alternative opinion to it's importance) but I feel that if I'm going to put all my eggs into this one basket -- rather than the multiple external drives --- I might as well do it properly for the long term safety of my data. I should have done it in the first place but figured my existing hardware would be "good enough".

I'm considering a sub-$200 Supermicro x10 board, 16GB ECC, and a $70-ish Pentium. Around $400 to redo my hardware, again. I'd then have to learn about server mobos, whatever the IPMI is, what I would do with dual LAN ports, etc. All while still trying to learn and maximize my FreeNAS experience.


TL;DR: I'm new to FreeNAS and don't know what I'm doing ;) Hello!

I'll have to see in which direction this all takes me :)

 

[gpsguy]

Welcome to the forums!

 

[Robert Trevellyan]

Welcome!

if you ever want to expand your storage you'd have to rebuild the entire zpool every time?

Not correct. There are two ways to expand your storage:

1. Add another vdev.
2. Replace all the drives in an existing vdev with larger drives.

Well, three, if you count destroying and rebuilding.

 

[cyberjock]

(Side thought: This seems like a silly limitation to me, you can't know what your future needs might be, so if you ever want to expand your storage you'd have to rebuild the entire zpool every time? That's not cool...)

The limitation, while being silly for people like you and me, is kind of pointless. Let me explain.

Sun Microsystems invented ZFS for large scale file servers and to be future proof with respect to scalability in the upwards direction. Scaling down was not ever a consideration because Sun didn't want to sell small servers. They wanted to sell the biggest damn file servers that you could possible build, which brings in the most money. These would be high end (read: expensive, like 'cost of your house' expensive) file servers running this new "custom file system" that would be all server grade components, all under support contract (read: LOTS of $$$), and all using hardware that Sun would happen to also be selling. The intention was not to have ZFS run on random hardware from your basement, nor was it expect to be able to scale down to the levels that many people want (there's little profit for companies for small scale servers). ZFS does not and will never scale down as small as many people want to go. ZFS was designed to scale up to mind boggling sizes though.

So yes, the ZFS limitations can suck. But "them's the breaks" when you decide to take a product engineered for massive scale and try to downsize it. It only goes so far, and some of the limitations are going to just suck. I wish we could add or remove single disks from a zpool at will, but that's not possible. It is possible that the future may hold that (there's plans to allow you to do this with striped, mirrored, and raidz1, but apparently raidz2 and raidz3 are extremely difficult so it may not ever happen). This means that even if that happens and theres no raidz2 or raidz3 support people like you and me that care about data integrity still should be avoiding raidz1 like the plague, and mirroring is expensive on a GB/$ basis, so the feature really doesn't mean anything until there's options for raidz2 and raidz3.

I myself am looking to add more storage, and the thought of having to buy 11 more disks sucks. But "them's the breaks" when you try to downscale ZFS. ;)

Just use this as a comparison: If you wanted 2TB of storage on ZFS, how much money would you have to spend for "recommended components". A FreeNAS Mini with just 2x2TB drives in a mirror would run you about $1200 or so. That's about as lean as you can reasonably get if you want "small scale" and low power. But if you want a synlogy running their RAID1 option, you could be up and running for probably $400 or so. ZFS just cannot scale down like that, which is why the FreeNAS Mini, while a really cool product, is hard to compete with when compared to Synology equivalents. Mind you Synology doesn't do ZFS, and they don't offer the same featureset that FreeNAS does. But those features may not be worth the extra $800+. For most people that look at FreeNAS, they want ZFS, and they're willing to spend the money to have the checksumming and redundancy that ZFS offers.

 

[Robert Trevellyan]

A FreeNAS Mini with just 2x2TB drives in a mirror would run you about $1200 or so. That's about as lean as you can reasonably get if you want "small scale" and low power.

What about a T20 or a TS140? The Pentium T20 that's been available for $179 is currently out of stock with Dell, but available elsewhere for less than $300. An i3 TS140 can be had for about the same price. Throw in an 8GB stick of RAM and two 2TB drives and you're up and running for less than $600. Even the Xeon T20 is only $489.