Well, shit. I've had one of these as a stand-in for hopefully eventually getting a Synology NAS, and now I'm paranoid about continuing to use my WD MyCloud. The thing is, I do believe I have a lot of reasons to believe I can trust Synology more, I don't even want to trust anyone. Not Intel or AMD, not WD or Synology. Computers are quickly becoming a source of implicit distrust for me.
> Well, shit. I've had one of these as a stand-in for hopefully eventually getting a Synology NAS, and now I'm paranoid about continuing to use my WD MyCloud.
The hardware is still fine. You can put Debian on it!
There is a very active forum of people replacing the WD firmware with Debian on various models (EX2 Ultra, EX2100, EX4100):
https://forum.doozan.com/list.php?2
Sadly homebrew Debian variants aren't very safe either. Usually they lack timely OS updates, ship pregenerated ssh host keys in the images, stuff like that. Even Raspbian is pretty bad at this though better than many smaller ones.
I guess people here now that. The thing is, these little consumer devices can be had dead-cheap. I bought one for my parents and it cost €30 more than the hard-disk inside would have costed retail. For €30 more I got a nice little case, an embedded device, that is good enough for the task and now I will install OMV on it.
Sure, at home I have a real server with ECC RAM, running on a Solaris derivate with additional VMs... but hey...! ;-)
Aahhg. This get repeated so often. FreeNAS and ZFS are not any worse and probebly better then everything else at dealing non-ECC ram. Its recommended because its generally a smart idea.
ZFS is worse because ZFS in general only has one failure recovery method: restore from tape.
Most other mainstream filesystems have tools to restore and attempt at repair the filesystem meaning that you can often restore the data.
ZFS also utilizes the memory more which means that it is more likely to be affected by memory issues.
Is it likely? No. But since us mortals seldom have tape backups and usually can't afford to backup everything on their NAS I'd say that ECC is recommended for ZFS, more so than most other filesystems. Just as a consequence of how it is used and the lack of recovery tools, not because of the filesystem itself.
If you're upgrading the overall specs, it's going to be more expensive, yep.
If you're trying to provide equivalent functionality, you can keep the same equivalent specs (or more likely, get a spec upgrade anyway) and deliver the same quality end result (IE, no ECC RAM) at a lower cost. You're unlikely to suffer a performance or integrity penalty compared to the equivalent off-the-shelf NAS solution.
If you want to substantially improve on the off-the-shelf NAS solution, ECC RAM is recommended, and will at that point, probably become more expensive.
Then again, if you want 10+ drives, you'll probably still be cheaper doing it yourself.
You need to upgrade the specs because ZFS is much more demanding than a normal linux fs that I assume most COTS solutions use (the absolute minimum requirements for freenas is 8 GB of ram, I would guess most of the shelf solutions have no more than 2 GB). The CPUs used in COTS solutions also wouldn't cut it in a freenas system.
The big death blow however is the inability to grow a vdev which means you have to pay the for redundancy each time you upgrade. Which isn't cost effective, so the alternative is to buy all drives up front which robs you of taking advantage of falling prices (as you probably don't need all that space on day one but gradually increase the storage used). This forces you to pick smaller drives that are more cost effective today (compared to buying a few larger drives that will become cheaper in the future), which means that you will have to replace the drives sooner when you run out of room in your chassi (or get tired of noise / power consumption).
I love freenas, in my opinion it is the best solution for the home NAS. I use it myself, but I do consider it to be quite costly (whether it is worth it is up to you). The end results all things considering is easily more than double that of a COTS solution that you can easily grow or a more regular linux+mdadm setup. And you also need to buy more of that up front rather than spreading it out over the years.
But it does depend on what your needs are and how much data you use.
You may not be able to grow an existing vdev, but you can add another vdev to an existing zpool. And although having all vdevs be identical gives more consistent performance, it’s also possible to mix different vdevs in the same zpool. Multiple vdevs also result in better performance, as each raidz/raidz2/raidz3 vdev only has the equivalent IOPS of a single drive, but without the write hole drawbacks of more traditional RAID5. Recovery time with multiple vdevs would also be much better than a single large vdev, which is becoming more and more of a concern as drive capacities continually increase while IOPS remain more or less the same.
Yes, all very true, but all of that is utterly irrelevant for the home NAS in 95% of cases since performance is not that critical and it comes at an enormous cost of wasted space, also physical space is a huge deal since adding two redundancy drives for each raidz2 quickly adds up. As well as noise etc.
And if we want to talk optimized setups it really isn't recommended to add new vdevs to a already populated pool since the usage will be quite unbalanced and affect performance as well.
Synology has its own backdoor as well [1]. Luckily the backdoor is only active when the OS crashed and the device is in recovery mode.
They also have some very odd practices, like hard-coding specific users and the shell to be used inside OpenSSH binary [2]. This was 2012; haven't checked if the code improved since then.
I'm aware of FreeNAS. I used to run a custom NAS setup. What I've learned is that its not worth my time to be making NAS setups. It can be a surprising ordeal from both a hardware and software PoV. If I were to build my own NAS, the first item on my wish list would be open source hardware with no firmware blobs and no hidden processors that I didn't ask for. Where do you go to get that? AMD has PSP, Intel has ME, ARM has TrustZone (yes I'm aware these are not identical,) and all of them feel untrustworthy to me now.
If RISC-V is fast enough to power a NAS and open source SATA controllers exist, I would absolutely prefer a setup involving that.
If I'm not going to get that... I'd probably prefer a setup from Synology. They'll do a good enough job, probably better than I could, even if the price is a premium.
I have won and its pretty awesome, running for years, have replaced the disks multible times by now.
Its very upgradable as well, you can put in more ram, caches and so on.
Open Hardware is pretty much a no-go at the moment, but iXSystems does support a lot of Open Source.
I would recommend this setup above any of these Synology things, when I compare my problems with my friends who have Synology, I usually come out pretty happy with my choices.
I understand what you’re trying to say, but what's the alternative? Stop using computers? Once Meltdown and Spectre are out of the picture, there will be another crisis. That’ll happen forever, I think.
Honestly, the older I get, the more sick of IT I become, but what can we do about this?
Im in the market for a NAS right now, so this is perfect timing. I’m more willing to trust FreeNAS than MyCloud but as soon as I write that, I remember lots of Open Source is crap too.
You get FreeNas on good hardware by people who know what they are doing. I have been running this for years and I have no complaints exept that the 8 slot variant did not exists when I bought mine.
I'm having modest success using an eSATA enclosure on Windows with Storage Spaces. On the Linux side, BTRFS may be suitable but harder to manage (This is part of what Synology/NetGear are providing management around.)
Unfortunately, I've learned this the hard way. Twice.
I'm sure it's more stable now than it was when I used it last, but it's frustrating when people say "no, this time it's really stable!" and then you lose data due to a bug. This same thing has never happened to me with ZFS on Linux, which is a shame because the licensing issues make it painful for me to use. I like up to date kernels.
I feel like btrfs should be phased out. It was clearly developed in a fashion that did nothing to prevent preventable bugs. This may be the same frustration that's leading RedHat to work on their own competing filesystem...
Any suggestions for a dynamically grow-able/shrinkable system of mixed drives for linux? When I last looked (and why I went with Storage Spaces) was that ZFS requires pools to be fixed at initialization, and I wasn't sure what other OSS solutions were mature at the time.
I have no suggestions honestly, I have looked into it pretty extensively and would like to go ZFS. But it will cost a decent amount for the setup I am looking for. It seems to be "all or nothing"
