Mini-ITX Build - FreeNAS compatible?

[Constantin]

I was under the impression that their implementation allegedly includes bit rot detection? FWIW, NOT trying to evangelize for Synology, stopped using it for multiple reasons. But it worked as a entry-level system.

 

[danb35]

I was under the impression that their implementation allegedly includes bit rot detection?

Yes, btrfs as a filesystem would do this.

 

[Arwen]

Yes, btrfs as a filesystem would do this.

Except that if BTRFS is used as a plain file system, (no redundancy), it can only detect bit-rot, not fix it. In this scenerio, if BTRFS detects bit-rot, you already lost the data as the RAID-5/6, did not fix it.

On the subject of BTRFS:
They could be using really custom code, (extremely un-likely), that if a BTRFS read detects a problem, then the underlying RAID-5/6 stripe's parity is checked. If bad, attempt recovery. Then if successful, re-try the BTRFS read.

Basically, so complicated that they would be better off using ZFS, (like Netgear does with their ReadyDATA line, not to be confused with the ReadyNAS line).

Last, there are multiple designed in failure conditions for BTRFS. For example, a simple Unix "mv" can cause data loss during power fail. Not possible with ZFS. Or the ever popular, store the checksum with the data, so when it's mis-written to the wrong block, you read good checksums, but bad data.

 

[Constantin]

FWIW, I don't rely on the Z3 pool to survive without backups either, even though the STH calculator tells me that the cumulative total pool failure probability by year 9 is 0.00000%.

Much hilarity may be had from my backups not relying on ZFS or detecting bit rot but such is life. It's an improvement opportunity that I'm well aware of and hoping to fix in the future. In the meantime, my backup system manually compares every file on a monthly basis with the one on the FreeNAS server, a poor mans scrub, I guess.

This is a improvement opportunity for the Mini / XL. Where the Synologies expose a eSATA and USB3 port for external backups, the Mini series does neither.

 

[KrisBee]

Just to add to @Arwen 's comments, this typical discussion shows how Synology's use of Btrfs and marketing about correcting bit-rot caused some confusion when it was introduced in DMS 6.1:

https://www.reddit.com/r/DataHoarder/comments/5yb13m/anyone_using_synologys_dsm_61_able_to_confirm/

 

[Chris Moore]

FWIW, I don't rely on the Z3 pool to survive without backups either, even though the STH calculator tells me that the cumulative total pool failure probability by year 9 is 0.00000%.

As long as you don't neglect the system, you should not have more than one bad drive at a time. We had a system where I work that had three drives needing replacement before the fault was detected but that was because of a laps in coverage caused by an employee moving to another job. The system went almost a year with nobody monitoring it, and the drives did not fail all at once. Also, only one of the drives was totally offline, the other two just had a lot of bad sectors. I was able to bring the system back to health, but the 4TB WD Red Pro drives it has, because they are pretty full, took almost 24 hours to resilver and there were no other drives in the 16 bay enclosure that faulted in any way during the three days of intense use it took to replace the three drives that were problematic. Even in the storage server where I have 60 drives all connected to a single pool, I am using RAID-z2 vdevs. I replace a drive if it even gets a reallocated sector though and I have 15 spare drives for that system sitting in my spares cabinet. Those are the 6TB Red Pro drives and that system takes almost three days to resilver a drive. In the testing I did with RAID-z3, I found that it was very slow when compared to RAID-z2. For my purposes, I would rather put 12 drives in two RAID-z2 vdevs than to put 11 drives in one RAID-z3 vdev, because of the speed, and I have personally tried both configurations.

 

[danb35]

https://www.reddit.com/r/DataHoarder/comments/5yb13m/anyone_using_synologys_dsm_61_able_to_confirm/

Sounds like Synology jumps through a lot of hoops to avoid using ZFS.

 

[Constantin]

I am not looking for peak performance as much as high reliability. With a 8-disk Z3 in the Mini XL (plus a SLOG and the 10GB FreeNAS interface card) I've managed peak throughput of about 300MB+/s using large media files. Smaller stuff runs at around 100MB/s. CPU utilization even with the on-board encryption has yet to crack 40%. That's fast enough for my home use.

Way back when I used ReadyNAS, they had nice feature where a single drive failure could trigger a system shutdown event. That worked really well in the context of home use, i.e. the appliance signalling rather forcefully that something is wrong and needs to be addressed ASAP. No email required, no internet connectivity required, etc. The thing just went safe.

 

[danb35]

Way back when I used ReadyNAS, they had nice feature where a single drive failure could trigger a system shutdown event.

Should be easy enough to do in a script if desired.

 

[Constantin]

Sounds like Synology jumps through a lot of hoops to avoid using ZFS.

I imagine it has a lot to do with what they know and have experience with. Synologies have a very small hardware footprint for the types of arrays they service... even with the optional upgrade, my DS1512+/DS1812 only had 8GB of RAM, IIRC? In other words, it's a very resource-constrained environment.

 

[Constantin]

Should be easy enough to do in a script if desired.

True enough... but with ReadyNAS, this was a GUI feature that could be enabled.

Also, on the next boot, the ReadyNAS would flag the issue in the web browser, disable the auto-shutdown once you acknowledged it, allowed you to rebuild the array, and so on. It's not as simple as shutting down the array with a cron job, you need to add some smarts on the next bootup as well.

I don't doubt that this can be done by the demi-gods here. But that's way, way, WAY above my ability.

 

[Chris Moore]

Also, on the next boot, the ReadyNAS would flag the issue in the web browser, disable the auto-shutdown once you acknowledged it, allowed you to rebuild the array, and so on. It's not as simple as shutting down the array with a cron job, you need to add some smarts on the next bootup as well.

I don't think anyone will be implementing that feature, not me for sure. In my world, up time rules all. I don't ever want a system to go down, certainly not for a failed drive, that is one of the reasons for RAID to begin with, high availability.

 

[Constantin]

Totally understand your POV and my needs as a home user are totally different from those in a production environment where businesses grind to a halt when the server goes down. Similarly, a Synology, ReadyNAS, etc. are built with very different uses in mind than high-IOP/throughput/etc. production.

 

[Chris Moore]

Hi everyone, first time posting here.

I want to build my first NAS, and I'd like to have your suggestions about it and if it would run FreeNas.

The specs:
Motherboard: Gigabyte GA-J3455N-D3H (integrated Intel Celeron J3455).
RAM: 8GB DDR-3 SO-DIMM (planning to buy another 8GB module in the future).
PSU: FSP Group FSP250-50LC (80 Pluz Bronze Certified).
Case: ABLECOM CS-M50 Mini-ITX.
Discs: For now, I want to install 2x WD Red 8TB HDD (WD80EFZX)

It will serve mainly to store my big media collection (about 1TB of photos and RAW files and about 2TB of videos and RAW).
and as a Plex Server (1 or 2 1080p streams). I need now about 7TB of storage (was thinking about a RAID1). Let me know if more information is needed.

Could it run FreeNAS? What would I need to do if I need to expand my storage in the future? Hopefully, I can have some good advice for
a first build and something else you think could be useful to know.

Thanks in advance!

I think that we kind of hijacked your thread. Sorry. Did you get the answers you were seeking?

 

[Ericloewe]

I imagine it has a lot to do with what they know and have experience with. Synologies have a very small hardware footprint for the types of arrays they service... even with the optional upgrade, my DS1512+/DS1812 only had 8GB of RAM, IIRC? In other words, it's a very resource-constrained environment.

Even so, a little more RAM and some optimization to the OS go a long way. I mean, FreeNAS is a bit of a hog, with most stuff being in Python and it's serving two whole GUIs, plus two APIs... And with compressed ARC, the smallish ARC they could afford would go that much of a longer way.

Besides, they could jack up the price, double the RAM and nearly nobody in their userbase would particularly care.

 

[Constantin]

That may all be true, I have no intuition re what is and is not possible re: optimizing ZFS for a smaller memory / processor footprint. The processors that Synology use in their stations are tiny. I imagine it helps a lot that they specify hardware and software, allowing them to integrate the two in ways that the FreeNAS team can only dream of. However, you also need much higher manufacturing volumes to make up for the fixed costs up front developing it all. Having a low price point to entice customers from other solutions like the even more basic competitors offered up by hard drive manufacturers is certainly one reason I can think of why Synology, QNAP, ReadyNAS, etc. keep their prices/costs low.

It is also this legacy of using low-cost CPUs and hardware solutions that likely have convinced these companies to maintain as much of the original codebase as possible. A smaller tree of stuff is easier to prune than jumping whole ship to a new platform with only limited adoption at first, etc. A sprawling codebase has killed more than one company.

That is not to say that QNAP, et al aren't trying to enter the high-end markets, the QNAP TVS-882T-i5-16G-US is one example. While it doesn't feature a storage solution as reliable as ZFS, it does feature thunderbolt 2 connections (though IP-over-Thunderbolt, not a true DAS experience), two 10GBe network interfaces, 8 hot swap bays for 3.5" drives, 2 hot swap 2.5" bays, and internal M2 storage as well. It ships with 16GB RAM by default, 32GB is possible, an i5 processor, etc. All that with a idiosyncratic but complete GUI that pretty much never needs a user to drop into the command line. 3 HDMI outputs (including one 2.0) allow outputs to various TVs.

To me, this is one of the defining differences re: QNAP and others.. they are marketing their higher-end solutions to folk who want to enjoy their media content, etc. with less regard for data integrity. FreeNAS seems to approach the market with a primary focus on the data integrity and infrastructure bit, with add-on software offerings being fiddly to install for a neophyte and sparse by comparison. I still chose FreeNAS because the ZFS file system has yet to be supplanted re: reliability and long-term stability and those things are way more important to me than being able to play some torrent content directly on an attached TV.

 

[Arwen]

...
This is a improvement opportunity for the Mini / XL. Where the Synologies expose a eSATA and USB3 port for external backups, the Mini series does neither.

I have my FreeNAS Mini expose 2 eSATA ports, (Intel 3Gbps), via the PCIe back panel. That's how I connect my backup disk. Works great, (if you don't need a PCIe card.)

My eventual goal, after the 3 year warrent expires, is to make a hole in the hack panel for 2 / 4 eSATA ports. Thus freeing up the PCIe card slot for 10Gbps Ethernet.

 

[Arwen]

Just to add to @Arwen 's comments, this typical discussion shows how Synology's use of Btrfs and marketing about correcting bit-rot caused some confusion when it was introduced in DMS 6.1:

https://www.reddit.com/r/DataHoarder/comments/5yb13m/anyone_using_synologys_dsm_61_able_to_confirm/

Wow, did they go to a lot of trouble. Every kernel and BTRFS update, they have to re-apply thier custom patch. Having to force MD RAID to perform a consistancy check on a single stripe of RAID-5/6 data is pretty out there.

Plus, if the consistancy check did not find anything wrong, what then?
Allow BTRFS to claim the read error?

 

[Constantin]

My eventual goal, after the 3 year warranty expires, is to make a hole in the hack panel for 2 / 4 eSATA ports. Thus freeing up the PCIe card slot for 10Gbps Ethernet.

Part of my conversion to a RPC-431 will include adding a similar ESATA port to one of the many empty PCIe slots in the back. I figure the board has 12 SATA ports, of which 9 will service my Z3 3.5" spinners, one for the SATADOM, one for the SLOG, leaving one spare. At $3.95 from Amazon, adding a PCIe-panel-based ESATA port is pretty much a no-brainer. Eventually, I hope to add an JBOD 5-drive external pool for backups and snapshots.

Seems to me that Synology is cautiously trying on the BTRFS trousers. Enable the parts that work, gradually adopt the other parts as they reach better levels of stability. So far, I haven't heard of massive issues but then again I haven't been following the Synology forums in a while. I doubt that the FreeNAS devs operate that differently, i.e. gradually adopt new features offered by FreeBSD, SAMBA, and other efforts as they mature and reach production-quality.

 

[Arwen]

...
At $3.95 from Amazon, adding a PCIe-panel-based ESATA port is pretty much a no-brainer.
...

If you have a choice, use an Intel SATA port for the eSATA PCIe panel adapter. Just a gut feeling that Intel would be a better choice than anything else, except LSI.

...
Eventually, I hope to add an JBOD 5-drive external pool for backups and snapshots.
...

Keep in mind that SATA port multipliers are really not a good idea. So you should use either a SAS expander, on a SAS host port. Or use 5 x eSATA cables to an enclosure that supports 5 eSATA inputs. Some external enclosures are SAS with built in SAS expanders. More expensive, but way more reliable.

If the external enclosure has only 1 eSATA port and more than 1 disk slot, it's using a SATA port multiplier. (Or some crappy and cheap RAID controller.)