BUILD Cheapest way to build a FREENAS?

[gpsguy]

Off topic, but ensure that you have 4GB+ RAM for your Sophos box. The newer versions of the firmware will need it. There are a couple of Astaro/Sophos UTM users on this forum. I started on 7.0.


Sent from my phone

 

[alexg]

So, I'm intruiged by this TS140 i3 box, it seems to be an ideal solution that just needs 8 more G of ECC ram... (I would want 12g to host Plex)

Here's my question - Does it use a standard motherboard? I already have a server chassis that I would rather use, it holds up to 19 drives internally (NZXT Whisper) Could I move the TS140 mb, cpu and ram to a different case?

My current WHS has been running reliably on a supermicro atom board, but alas, it doesn't support ECC... (and I totally get why ECC is required. Thanks Cyberjock) so, I guess that MB will be repurposed... (I'd like to build a new home Sophos UTM for home anyway)

I see that Cyberjock has read this thread and didn't smack anyone's heads over the TS140, so I assume it's a good platform for ZFS...

Here is the link to parts list for TS140. Looks like Intel's board, but not sure of the size until I have a chance to open. Be aware, I believe these servers have non-standard power adapter, but you may be able to find converter.

http://download.lenovo.com/parts/ThinkCentre/ts140_partslist_09182013.pdf

 

[marbus90]

the TS140 uses a standard microATX board with propertiary PSU power ports. I believe the Dell T20's implementation is more conform with industry standards since my HP SFF boxes and the Dell T20 have the same pinout.

 

[Trimble Epic]

I believe these servers have non-standard power adapter, but you may be able to find converter.

Could the power supply be transplanted as well, I wonder? I guess a more important question is - could it's non-standard power supply handle more than the number of drives that would fit in its usual case...

Edit: Is this a new low? : http://www.newegg.com/Product/Product.aspx?Item=N82E16859106529 199 after rebate?

 

[alexg]

Could the power supply be transplanted as well, I wonder? I guess a more important question is - could it's non-standard power supply handle more than the number of drives that would fit in its usual case...

Edit: Is this a new low? : http://www.newegg.com/Product/Product.aspx?Item=N82E16859106529 199 after rebate?

I think you need to carefully examine your plan. That power supply in TS140 is only 280W. If you need to put many drives, it may not be enough. Also, look at the cost of just buying new motherboard and CPU instead going with TS140. I'm happy with my setup of 4 drives internal and one over ESATA for replications. There is probably a room for one more inside the case in the floppy disk area. If you don't have any long PCI cards, I can see someday modding the case and placing more drives on the bottom.

 

[Trimble Epic]

I think you need to carefully examine your plan.

I agree, that's what I'm doing. I'm in the planning stage right now. This is just a possible idea. But still, for $200, the TS140 is hard to resist, even if it doesn't become a Freenas server.

 

[alexg]

yep, I have two of them now. My second one is Xeon based and I'm using it as VM lab for testing running XenServer.

 

[R.G.]

As a bit of kibitzing, you're asking the wrong question.

This is an issue I saw memorialized in an ad in a guitar magazine. The picture was a musiciany-guitary guy at a music store saying "I have really big, career-making gig tonight. I want your cheapest guitar."

Data is worth some money to protect. Deciding to protect data worth something to you with the cheapest possible solution is - well, misguided.

The right question is "what is the cheapest way to build a FreeNAS that meets these objectives?" and then lists the objectives. Things like:
- supports X TB of drives in RAIDZ 1/10/2/3/etc.
- has Y onboard SATA ports with/without a bus attached disk adapter
- has/uses 10/100/gb LAN port; or two ports, or intel LAN adapters, etc.
- has PCI/PCIe slots of various capabilities
- costs under $Z regardless of storage size, speed, performance, etc.
Or, better:
- I need to store L TB of data and access it at M bit/sec and have it survive the loss of any N drives in the array; AND I need it to use less than O watts of power.

It's even OK to say you just want to muck about with a very cheap setup of FreeNAS as a more educational alternative to playing video games, and if the data gets lost, well, that's OK too.

The cheapest way to build a FreeNAS is to use old/junk computer parts. Regardless of your experience, you can make this work (I have, it does) and it can be as cheap as free, in spite of having to work through the issues of flakey hardware.

As a note, the supporters of FreeNAS seem to have given up on the idea of making it be everyman's NAS and working on the lowest common denominator hardware in favor of making it a semi-industrial product, so it's possible that the newest releases will not work on older hardware, as the older hardware may not meet the minimum requirements of the newer releases. For instance, when I got into this, the boot image on flash was pushed hard to keep it in 2GB. That approach has (reasonably) been abandoned in favor of 4GB and multiple images; this makes sense in a world of $10, 16GB flash drives. But it's easy to get old hardware that won't work with the newest software upgrades of FreeNAS.

I used ASUS motherboards and AMD processors for a couple of years for their claimed support of ECC memory. Worked fine, never had an issue. But I don't really know if I had an ECC error or not. It just worked. I finally got convinced that since Supermicro prices were relatively lower, I could afford a Supermicro. It works. Well, I think it works. I'm having erratic disk errors that are not reflected on the SMART logs and I can't quite trace them down, but this could be other issues. Never had that with ASUS/AMD.

But I digress. I think you're asking the wrong question(s).

 

[Trimble Epic]

Perhaps you're right, from a certain perspective.

I'm still not completely sure FreeNAS it the way I want to go... but I'd love to be able to play with it as a "more educational alternative to playing video games", it's just that after reviewing the info on ECC for FreeNAS, I'm not willing to go without it. So, a TS140 (+4g) would make a good entry point to test it and learn more, and perhaps set up a first* FreeNAS, perhaps leading to a second using bigger and better hardware.

I'm also evaluating Windows Server 2012 R2 (as to if it fits my needs)... It's my understanding that a TS140 could also run that. So again, I'm lead back to that, and I'm asking myself if it's possible to keep my existing server chassis either way. I really like the way it holds drives.

What I'm replacing is a Windows Home Server (v1.. ugh) that has been running reliably for years. My wife doesn't even understand why I want to to mess with what seems to work fine... sigh. One of my concerns that I'm not sure FreeNAS can beat is the backup software that WHS uses. I understand WS2012R2 also has it, so I'm REALLY tempted to go that direction.

I also have a Drobo FS that I just aquired, and it's doing a fine job of juggling disks of random size, which I'm learning is not something ZFS can do (as well as a Drobo... I could swear I had once read that it could, but I recognize that sometimes things aren't what you thought they were... wait, there isn't an easter bunny?)


* actually, it would be my 2nd FreeNAS - I build one many years ago before my WHS build, LONG before ZFS was involved. (I guess that's nas4free now)

 

[R.G.]

My own personal objectives were heavy on preventing data loss. This led me down the path of spending more to get more data security; not tossing money at the problem, just doing the right job. As such, I was able to look at the rough cost of replicating the data versus making it not get lost in the first place.

If you're after testing and learning - and I consider that a MUCH higher calling than an alternative to video games, myself :) - then that frees you to do more things in a less disciplined manner than some others. In fact, it frees you from the issue of ECC entirely, as your objective isn't wrapped up in the never-ever-have-any-data-loss paradigm. And it frees you from some of the issues with number of SATA ports, LAN adapters, and other stuff that get long, involved discussions here.

Frankly, if I were running down your path, I'd go find an ASUS M4A... or M5A... motherboard, an Athlon II and cram that in a surplus case that has enough drive positions to make you feel good. These things are on the cusp of going obsolete, so good used ones would make a great educational setup, and a good place to make the inevitably errors nonfatal. Something like that gets really cheap.

Be very, very careful of project scope creep. There is a danger that the project will succeed just fine, and you'll come to think of it as good enough for "production", outside one or more of the objectives you originally set. That sets you up for a data disaster. As the ancient Greeks used to say - those that the gods would humble, they first make proud. :)

 

[Trimble Epic]

There is a danger that the project will succeed just fine, and you'll come to think of it as good enough for "production", outside one or more of the objectives you originally set. That sets you up for a data disaster.

Well, That's the reason right there that I would demand ECC... From what I've read so far, I think a TS140 with 12g of ECC probably WOULD be sufficient as a production server for my use... and my use is to hold my family's files (school and college papers), photos, video files, a Plex server, and iSpy. And I'm backing everything up offsite.

I'm currently executing a project to back up every video file to blu-ray for offsite storage... because I don't trust the Drobo. (their forum is filled with people who "loose the whole pool". there are no positive experience posts over there.. to find people who are actually SATISFIED with their drobos, you have to go read the comments on Amazon) (oh, and blu-ray because they are immune to bitlocker, et al)

Again, I haven't decided yet. I'm in the analyze phase of my project ;) But the ZFS technology has fascinated me for a while, and when I decided to work on replacing the WHS, I wanted to at least evaluate it.

 

[pschatz100]

The cheapest way to build a FreeNAS is to use old/junk computer parts. Regardless of your experience, you can make this work (I have, it does) and it can be as cheap as free, in spite of having to work through the issues of flakey hardware.

As a note, the supporters of FreeNAS seem to have given up on the idea of making it be everyman's NAS and working on the lowest common denominator hardware in favor of making it a semi-industrial product, so it's possible that the newest releases will not work on older hardware, as the older hardware may not meet the minimum requirements of the newer releases. For instance, when I got into this, the boot image on flash was pushed hard to keep it in 2GB. That approach has (reasonably) been abandoned in favor of 4GB and multiple images; this makes sense in a world of $10, 16GB flash drives. But it's easy to get old hardware that won't work with the newest software upgrades of FreeNAS.

The hardware requirements that FreeNAS has are due to 1) FreeBSD support, and 2) the decision to use ZFS filesystem and drop support for UFS. If you want to use "lowest common denominator" hardware, then you should use NAS4Free which can be configured for more modest hardware or use a system that is based on Linux. They are out there... and they cost the same as FreeNAS...

Why should the FreeNAS developers be obligated to make an "everyman's NAS?" It is difficult, if not impossible, to make something that works for every use case and runs on any/all hardware in existence. The developers seem to be pushing forward on certain new technologies, such as ZFS, which they feel will make their product a more robust tool. I agree with this approach, which is why I'm on board. They are also dropping support for certain old technologies which are increasingly difficult to support or don't contribute to their vision of a more reliable and sustainable NAS appliance.

You might not agree with their decision to drop UFS. However, if you understood how ZFS works, then you would know why ECC memory is so important and why 8Gb is the suggested minimum. If you understood how system updates are handled in version 9.3, then you would know why they now recommend larger flash drives for the boot device.

I would say that, if you don't generally agree with the vision, then maybe you should use another system. While I won't claim to agree with every decision they make, I do agree with their vision.

It is ironic that Apple does exactly the same thing with their systems, and nobody gets all upset about it.

 

[R.G.]

Why should the FreeNAS developers be obligated to make an "everyman's NAS?"

They shouldn't have to, and there was no implication in my comment that they should. I was noting a fact.

You might not agree with their decision to drop UFS. However, if you understood how ZFS works, then you would know why ECC memory is so important and why 8Gb is the suggested minimum. If you understood how system updates are handled in version 9.3, then you would know why they now recommend larger flash drives for the boot device.

I think I'm having a bad attitude day. :)
- I have no opinion on the decision about UFS either way. I came to FreeNAS looking for ZFS after Oracle neutered Open Solaris.
- Having been chewed on by some experts, I get all ruffled at statements in the form of "Well, if you were not deficient in not knowing or understanding X, then you would know Y," the implication being that not knowing Y means you're deficient in some way. That's probably not what you meant, and if it's not, I apologize at getting worked up at it. I pretty much fully understand ECC and the various tradeoffs involved, as well as why it's important. Maybe you meant that for another poster.
- I have no opinion about the larger flash image one way or the other. I stated that as a fact. It makes no difference to me at all. I stated that fact as an illustration that there is some older hardware that won't work with the most recent changes to FreeNAS.
- For me personally, I'm all in favor of many of the changes. I was trying to help the original poster think about FreeNAS and how works for him.

I would say that, if you don't generally agree with the vision, then maybe you should use another system. While I won't claim to agree with every decision they make, I do agree with their vision.

Again, I don't know exactly why that strikes me wrong. Maybe I haven't had enough coffee yet. :)
Personally, I don't care in the slightest about the developer's vision. It either works for me or it doesn't, and if it doesn't, I'll go elsewhere in a heartbeat. As noted, I already did that once. I went to Open Solaris trying to get ZFS for some data reliability and self healing, which is what my needs are. When Oracle folded that up, I went looking, found FreeBSD, and then FreeNAS, which cases up things that make it easy for me to do what I want. This is great, and I try to help others with any give back I can do based on my experience. If that becomes too difficult or burdensome, well, OK. There are other things to do to fix my personal data needs.
FreeNAS is good, does what I want, and I like the simplicity of using it. I've used other file management systems and FreeNAS is a very handy tool. But it is a tool, not a life's calling.

It is ironic that Apple does exactly the same thing with their systems, and nobody gets all upset about it.

I personally believe that this is a quirk of demographics. The IBM PC "harvested" the vast majority of the DIY types in the computer market, and then blurred out into the other various operating systems as PCs grew up into the larger computing world. Apple essentially abandoned the DIY types with the Apple II and had a near-death experience, were saved by the ipod, and came back with the closed-system Macs. The Apple concentration on making the user not need to know or care about the workings behind the user interface led to both its adoption by the fraction of the market that did not WANT to know of care about what was behind the screen, but also to Apple's fencing out the users to keep them from hurting themselves or the system. I think it's a natural outgrowth of the markets they appealed to. IBM PC evolution selected for the folks who want to go right down to the bare metal, Apple evolution selected for the people who don't want to know. And now that the expectations have been set, the people who adopt the Apple philosophy don't have any expectation that they should have an open system.

In my opinion. :)

 

[alexg]

I thought this thread was about cheapest way to build FreeNAS :)

 

[R.G.]

It is. Or should be. :)

A clear statement of objectives really, really helps in deciding what "cheapest" means in context. Continuous, always-on, no-data-loss as a requirement can push the minimum expense up. Using the system for a learning/teaching tool can make it quite inexpensive.

 

[Robert Trevellyan]

I'm going to throw some information into this thread, not intended as recommendations for others to follow, just FYI on how I've achieved my own low-cost FreeNAS setup.

My first experiment with FreeNAS was back around late version 7.x or early 8.x, stuffing as many IDE drives as I could into an old Dell Dimension 4400. I made it work, but I had no real use for it, so I discarded it. It was all parts bin except for a 4GB IDE DOM, which cost me about $40.

My next go around was with 9.2.x and a Dell Vostro 420 that I received as a hand-me-down. With 7 on-board SATA ports I immediately thought of FreeNAS. I bought 8GB of RAM and a HighPoint Rocket 640L and realized I could actually make good use of a 9.x series FreeNAS box, primarily as an onsite backup destination for several Macs, but also some Plex and BTSync action. I stuffed it with used hard drives, but the lack of ECC began to nag at me, plus the fact that the box was maxed out at 8GB. Per the often asked question by resident experts here, if I didn't have ECC, why was I even using FreeNAS? I learned that it was really handy having an eSATA port on a FreeNAS box - it really simplifies disk replacement and pool replication when tinkering with different vdev layouts, especially if you already have something like this. The 8GB of RAM cost me about $80 if I remember correctly, and the HighPoint 640L was about $30.

Having learned a lot along the way and decided to continue using FreeNAS, but knowing that my requirements do not include high performance or massive storage, I recently acquired a used Dell PowerEdge T110 for $275. The key to this purchasing decision was that it was already maxed out with 16GB of ECC RAM, where buying a T20 or TS140 or TS440, which I think are all attractive options, would require spending another $200 or so to get to 16GB. I transferred the HighPoint 640L and installed 6 used 500GB hard drives, facilitated by this. Total cost $297, less what I got back on eBay for two hard drives and a RAID card that came with the T110.

No doubt I'll run into limitations eventually with the T110, not least being maxed out at 16GB of RAM, but I expect it to last quite a while based current usage.

EDIT: almost forgot, this machine also runs an Ubuntu VM for local web development via the VirtualBox plugin.

 

[Wolfeman0101]

Sorry to resurrect an old thread but I figured it was better than starting a new one. I have a 6 drive FreeNAS box right now but it's been randomly shutting off and I'm not sure what the issue is. It was kinda a Frankenstein box so rather than dump money into it I'd rather just take the 6 drives and put it in a new system. I guess my only requirement is the 6 drives but smaller is better.

 

[Robert Trevellyan]

I guess my only requirement is the 6 drives but smaller is better.

The T20, T110 and TS140 entry level servers only support 4x 3.5" hard drives out of the box. The T20 has a neat trick with 4x 3.5" and 2x 2.5" if you don't need an optical drive. Otherwise, you have to add brackets and an HBA to install 6 drives, and then cooling may be sub-optimal. However, the TS440 can be configured with 8 hot-swap 3.5" bays.

 

[di0de]

Is the TS140 still a good buy at $350? i3, 16GB ECC RAM, no drives. Money is tight but my old set up is on its last legs. Will the i3 bottleneck me? I use NFS for a couple Kodi (OE) Pi's, Plex and Supersonic for streaming to a few devices off network. I would like to add a few more plugins as well.

 

[gpsguy]

If you live in the US, you can get a PowerEdge 20 for $139 (until 12/24) for $139. See - http://bensbargains.com/bargain/del...dual-core-server-139-at-dell-business-112509/

16GB of ECC RAM from www.crucial.com costs ~ $138. @Robert Trevellyan is using a PowerEdge 20.