FreeNAS as a backup device?

[ferg22]

Hello all,

I have setup my first ever ZFS NAS, based on a 4x6TB disk RAIDz2 run on a HP. I was assuming it was a nearly perfect storage device. But everything I read about ZFS+FreeNAS stresses that FreeNAS should never be used as a backup device. However that is its primary requirement in my world. It will act as my apple timemachine server.

My question is ... what suggestions or best practice are there for backing up my 12TB NAS? This is a DIY home office type of site, tons of rubbish, family photos and movies, plus the timemachine backups.

I was wondering about a separate single 12TB disk that I would sync the ZFS-NAS to now and again. The idea being that while the ZFS-NAS seems to cope with any of the 6TB disks failing, if the NAS itself failed it appears I would loose every single byte of data. The 12TB backup would then hopefully allow me to recover MOST of the original data.

 

[Nick2253]

But everything I read about ZFS+FreeNAS stresses that FreeNAS should never be used as a backup device

I'm not sure where you've read this, because many, many people use FreeNAS for just that purpose.

However, I'm wondering if there's some confusion in your question. There are some challenges with using any kind of non-Apple NAS as a Time Machine backup target that are specific to Time Machine, so you need to make sure you address those. Here is a thread that deals with this: https://forums.freenas.org/index.ph...-machines-with-osx-server-style-quotas.47173/

For backing up your FreeNAS device itself, I would recommend another FreeNAS setup. At the very least, I would recommend a ZFS based backup target, simply for the ease of doing ZFS send/receive with snapshots. If your backup plan involves setting up a separate non-ZFS backup target, you'll need some other tool, like rsync, to do the backups for you. However, many people here use rsync with great success. If you don't use ZFS and you want additional features, like deduplication or snapshots, you'll probably want to look into a third-party backup tool.

 

[Chris Moore]

What you read was probably something to the effect that an array of disks is not a backup.

 

[Yusuf Limalia]

Hello all,

I have setup my first ever ZFS NAS, based on a 4x6TB disk RAIDz2 run on a HP. I was assuming it was a nearly perfect storage device. But everything I read about ZFS+FreeNAS stresses that FreeNAS should never be used as a backup device. However that is its primary requirement in my world. It will act as my apple timemachine server.

My question is ... what suggestions or best practice are there for backing up my 12TB NAS? This is a DIY home office type of site, tons of rubbish, family photos and movies, plus the timemachine backups.

I was wondering about a separate single 12TB disk that I would sync the ZFS-NAS to now and again. The idea being that while the ZFS-NAS seems to cope with any of the 6TB disks failing, if the NAS itself failed it appears I would loose every single byte of data. The 12TB backup would then hopefully allow me to recover MOST of the original data.

I think there's some overlap between high-availability and backup. There are also many disclaimers that you would find regarding RAID / ZFS. I guess people are trying to cover themselves in the case where you lose all your cat pictures because you've had some drives die on you.

From my perspective RAID / ZFS is more of a high-availability implementation. If a drive(s) dies, my data will still be there and available.
Some best practices for backup would be making sure your data is not just stored in one single place. In the event of a natural disaster / theft / case where all your drives die you would lose everything if you had a single place where your data was. Ideally you want an off-site backup of your data.

Some options for off-site backup may be:
As @Nick2253 mentioned FreeNAS replication / rsync backup.
Replication will cost more, but you've got another set of redundant disks looking after your data.
This option means you have to manage and maintain the hardware etc.

Or if you don't mind, somewhere in the cloud. There's many options here. personally I choose AWS S3, which is native to FreeNAS as well.
AWS Glacier will set you back around $0.004 / GB per month, so 12TB would cost around $50 ($0.004*1024*12) which can be a lot. $600 / year.
AWS makes lots of guaranteed for durability 99.999999999% and they will maintain and manage all of that.

unless you can absolutely guarantee against natural disasters, theft, an angry ex-partner armed with

Code:

sudo rm -Rf /* 

or 3 of your drives dying simultaneously, I would recommend an off-site backup be it self maintained or cloud based.

 

[Chris Moore]

I have had hardware failure of a SAS controller in my NAS and all the data was just fine when I got the hardware working again. Unless you have some truly catastrophic event like a fire or lightning strike, the NAS failure taking your data is unlikely. That said, you should still have a backup, for the possibility that a person makes an accidental deletion, or the other options already mentioned above.
You should not rely on a single storage device. I run a pair of FreeNAS servers at home, and one of those is just a backup of the first one. I also have a second array of drives connected to the main NAS in an external enclosure and it is a second Zpool that I use to backup the main storage pool.
It all depends on how much you value that data.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[wblock]

But everything I read about ZFS+FreeNAS stresses that FreeNAS should never be used as a backup device.

That sounds like a misunderstanding. The important point is that disk redundancy is not a backup, just a protection from hardware failures.

 

[garm]

Ya I’ve seen this elsewhere as well. “RAID” is (by right) considered a poor backup medium due to its increased complexity over tape or plane drives. Thus the notion that ZFS is unsuitable for long term storage..

I my opinion there is no contest when it comes to long term storage, at the moment ZFS is the best (if not only viable) solution for cost effective long term storage in “small scale”. A single node is in itself not backup. Saying you intend to fill up a NAS with 11 TB of data and 1 TB of timemachine means you have 11 TB of storage and 1 TB of backup. The 11 TB of storage needs to be backed up as well, preferably on ZFS.

 

[wblock]

Ya I’ve seen this elsewhere as well. “RAID” is (by right) considered a poor backup medium due to its increased complexity over tape or plane drives. Thus the notion that ZFS is unsuitable for long term storage.

That still misses the difference between storage and backup.

Consider deleting a single file that it turns out later was really needed. The NAS faithfully removes that file. It's gone. With a backup, traditionally removable media, you can go back and recover that file. Backups have a history, more than one level.

With ZFS snapshots, something similar can be done, but suitability as backup depends on the use case.

 

[Chris Moore]

Ya I’ve seen this elsewhere as well. “RAID” is (by right) considered a poor backup medium due to its increased complexity over tape or plane drives. Thus the notion that ZFS is unsuitable for long term storage..

This idea of RAID is not a backup represents a misunderstanding or a partial answer to a larger question. A single RAID array (even if that RAID is created using ZFS) is not a backup, because (as @wblock stated) if you delete a file, by accident or otherwise, that file is gone. The fragile nature of storage, and the need of a backup, is not directly related to RAID. The notion that, "the data is safely stored in the RAID array, so we don't need a backup," is what needs to be examined. You can't just put all your data into ANY storage system and call it secure. I use a philosophy of, "one is none," which means, if it isn't clear, if you only have one instance of the data, you don't have a copy. The suggested best practice in the industry (and many people are aware of this) is to have a working copy, a backup copy and an offsite copy. More copies of the data make it more secure, but it is also more work and more expense. For example, I do not keep my personal data in any offsite backup except for the few documents I put in my Google Drive. I am simply not willing to pay the monthly note on it. The organization I work for owns and operates their own bunker (blast doors and all) for a full offsite backup including a small IBM supercomputer. Ref: http://www.navy.mil/submit/display.asp?story_id=67524
How much you can back things up depends on your budget though.

I was wondering about a separate single 12TB disk that I would sync the ZFS-NAS to now and again.

Instead of spending for a single 12TB drive, you might want to do what I do for my home NAS. I have a primary storage pool and a backup pool Everything that is in the primary pool is copied to the backup pool once a week by a cron job. I also use ZFS snapshots on the primary pool so I can go back to a previous time in history if I accidentally delete a file or something. There are many features of ZFS that make it superior to any hardware RAID technology, so many things that you have learned about hardware RAID may not apply to ZFS at all. Take some time and learn more about it.

 

[garm]

When I say RAID is unsuitable for storing backups, I mean traditional hardware or software arrays where data gets sprinkled over multiple devices. To many components in that chain can screw you over. With ZFS you give the file system direct access to the drives so the concern is not valid, but I see that distinction isn’t clear for many “unenlightened”.

And I never said a single FreeNAS machine was a backup, unless you store your stuff in two different locations minimum, it’s not a backup, it’s just storage.

 

[Nick2253]

When I say RAID is unsuitable for storing backups, I mean traditional hardware or software arrays where data gets sprinkled over multiple devices. To many components in that chain can screw you over. With ZFS you give the file system direct access to the drives so the concern is not valid, but I see that distinction isn’t clear for many “unenlightened”.

And I never said a single FreeNAS machine was a backup, unless you store your stuff in two different locations minimum, it’s not a backup, it’s just storage.

I wondering if there is a language difference here, because to me you are drawing a distinction without a difference. And you're also mangling the important distinction between backup and redundancy.

Your claims about RAID being unsuitable for storing backups is simply a round-about way of making the claim that RAID is unsuitable for storing data. And considering that petabytes of data are stored in RAID arrays all throughout the world, that claim is just wrong.

The real concerns about hardware RAID are concerns about redundancy, which is why things like ZFS, ReFS, and BTRFS have come into being. They address failure modes that could not easily be addressed with existing infrastructure, like copy-on-write and bit-rot. However, the idea that "sprinkling" data over multiple devices is somehow problematic is absurd; that is exactly the point of hardware RAID, ZFS, etc: spreading data across multiple devices to maximize the redundancy in the event of failure.

 

[Chris Moore]

When I say RAID is unsuitable for storing backups, I mean traditional hardware or software arrays where data gets sprinkled over multiple devices. To many components in that chain can screw you over. With ZFS you give the file system direct access to the drives so the concern is not valid, but I see that distinction isn’t clear for many “unenlightened”.

And I never said a single FreeNAS machine was a backup, unless you store your stuff in two different locations minimum, it’s not a backup, it’s just storage.

I was trying to expand and clarify the information you provided, mostly for the benefit of the original poster.
Please do not feel that I am trying to attack you.

 

[garm]

I’m not feeling attacked :) I just battle this discussion so much that I feel this is a good place to try and clear it up for anyone searching for it, maybe I should write a blogpost instead on the virtue of ZFS ^^

When you discuss these topic of backup storage, you often come across the claim that RAID is a tool for high availability and not backup, the risk of corrupting data on traditional arrays are just to high. With ZFS I believe though that their concerns are void. ZFS in my opinion outperforms traditional backup storage media. Drawback is of course that you need to have the drives spinning and that is not always a valid solution in a commercial scenario, but for my home storage it is. For personal storage I have moved away from shelf drives and optical kettle underlays to storing eveything on the machine that is up all the time any way, my home media and backup server. ZFS not only keeps my data safe against plate ware and the odd neotrino, I’m also confident that the data I send to backup is the same data that I wrote.

 

[Chris Moore]

You can spin the drives down in a cold storage server, you just need to be mindful to configure the server in such a way that the drives will only spin up when there is a real need to access the data, to reduce wear from the start/stop activity. A regular storage server should not spin the drives down because the high start/stop activity will probably cause premature drive failure. There are many advantages, among which is the protection from bit-rot, to using ZFS in cold storage and with the high price of the tape robot systems and the capacity of drives going up at the same time the cost of high capacity drives is dropping, it is an attractive option to use large disk arrays for cold storage instead of putting data on tape. Another reason being the speed a cold storage array can be accessed compared with the speed of finding and retrieving data from tape. There is real competition in the industry at certain price / performance levels. It often comes down to how much data and how quickly the backup/retrieval needs to run. Where I work, we have a large tape robot system that takes up the floor-space of about twenty server racks. We use that to backup the SAN, but for some other smaller systems, it is more economical to buy a second server and backup to disk instead of investing in more tape library systems. Our IT Director is close to retirement and he had the final say on a lot of the decisions with the new SAN and there were several things he insisted on doing that were counter to the recommendations of the vendor. I think it might have been better, even for the SAN, to use cold storage on disk instead of that enormously expensive tape system. Our SAN was implemented by Sun/Oracle (they had a team of engineers on-site for a couple weeks) it uses ZFS and there is no question that a ZFS array is up to the task of storing monumental amounts of data and doing it reliably.
Unfortunately, there is a lot of bad, incomplete, or outdated information still filling the internet and there is often little or no distinction made between a system like ZFS and the hardware RAID of the past which could often be less than desirable, but was the only answer at the time.

I was wondering about a separate single 12TB disk that I would sync the ZFS-NAS to now and again. The idea being that while the ZFS-NAS seems to cope with any of the 6TB disks failing, if the NAS itself failed it appears I would loose every single byte of data. The 12TB backup would then hopefully allow me to recover MOST of the original data.

We have not heard back from the OP and I wonder if we have gotten the discussion a little off topic. @ferg22 , have we provided you with any answers to your question? If the hardware you use for your NAS fails, but the disks are not damaged, you can import the Zpool in another FreeNAS system and regain access to your data. A failure of the NAS hardware does not impact your data. However, you should still have more than one copy of the data to protect against an erasure (accidental or intentional) that is not authorized. I am not sure how you are using your system, but if you can give us some more information on your concerns, we might be able to better answer them.

 

[ferg22]

Thanks. I have read and digested most of what was said.

1) The data+backup+offsite strategy. Well I have my live data sprinkled about the place on various systems, with I might add a fair amount of duplication which I was allowing to happen as it is also a form of backup. The NAS was intended to be my official backup device, hidden away, and slightly safer from theft and fire. Indeed I was thinking about spinning disks down when not in use - I guess its the only reason my circa 2004 ICYBOX lasted so long. But given the docs and comments I am now more set on a single 12G external disk, for occasional 'archives' of the NAS.

2) The FreeNAS Guide from Cyberjock stated the issues most clearly:

Code:

	 ZFS has very few “recovery tools” unlike many
	 other file systems. For this reason, backups are
	 very important. If the zpool becomes
	 unmountable and cannot be repaired there are
	 no easy software tools or reasonably priced
	 recovery specialists you can use to recover your
	 data.

But the sentiment appears all over the web. And while the web does focus on the negative or problems with tech, as a newbie I have no way of knowing how likely it is for my zpool to become unmountable or how best to protect against it. I see use of a UPS mentioned in several places, but I really dont want to use a UPS.

3) I have some experience of RAID and its failures. While I do not work in storage I had noticed over ten years ago the issue with a second disk failing in a RAID5 set while its was recovering a failed disk. Bad. I had thought that ZFS had this licked - which I think it has. But I have this new issue of the zpool failing to cope with.

 

[ferg22]

More.... thoughts. The threats we face these days have increased.

• The sheer size of the data we aspire to keep around means that older tech such as Raid1-5 is no good.
• Aggregation means that the net value of the data has increased. Who cares if you lose one photo of the family, but to lose a thousand?
• The length of time we wish to keep DIY domestic stuff around for is probably much longer than most business retention policies. Hence we want a storage solution that is better in some ways than that used by business.
  
• I consider the threat of ransomware as a "when" and not an "if". While I am aware of the ransomware risks - that wont stop me doing something stupid and of course there is always the rest of family that could import something nasty. Hence, for DIY work I want manual and not regular automated backups. My timemachine backups are currently manually started.

 

[danb35]

ZFS snapshots are an effective mitigation strategy against ransomware. I also don't see how that threat leads you to prefer manual vs. automated backups.

 

[wblock]

The sheer size of the data we aspire to keep around means that older tech such as Raid1-5 is no good.

RAID1 is mirroring, which is still very usable. RAIDZ1 is essentially the ZFS version of RAID5, still useful for drives of relatively limited size.

but I really dont want to use a UPS

Why not? Most of these questions just boil down to "how much is the stored data worth?" Or, to put it another way, these are insurance prices. A UPS is very cheap insurance.

 

[Chris Moore]

But the sentiment appears all over the web. And while the web does focus on the negative or problems with tech, as a newbie I have no way of knowing how likely it is for my zpool to become unmountable or how best to protect against it. I see use of a UPS mentioned in several places, but I really dont want to use a UPS.

If you don't use a UPS, you are just asking for your data to be corrupted. The UPS is a requirement of any storage system.

 

[ferg22]

Thanks for the further feedback.

RAID1 is mirroring, which is still very usable. RAIDZ1 is essentially the ZFS version of RAID5, still useful for drives of relatively limited size.

RAID1 on > 1TB drives is IMHO just not useful; rebuild just takes too long, if it works. But I guess I had not really considered 3 or 4 way mirrors. RAID5: I have twice personally seen a second disk in a raid5 set fail while the set was being rebuilt causing the whole set to be lost But for the moment I am happy to go with raidz2 with carefully purchased disks.

ZFS snapshots are an effective mitigation strategy against ransomware. I also don't see how that threat leads you to prefer manual vs. automated backups.

Snapshots are totally new concept for me which have I guess I will come around to. You are right, they should be a useful strategy against ransom-ware. Perhaps I am a control freak, but I wish to know when my Macs are being backed up to timemachine, rather than many partial backups as and when apple and network conditions allow.

UPS feedback. Interesting! Is it being suggested that the lack of a reliable power supply is a way to loose my zpools? In which case would an alternate strategy be to keep the system spun down except when a backup or restore was in progress.
I have some experience with UPSs, IMHO they are very expensive, need constant monitoring and often let me down. The APC stuff is real garbage and at the same time it is the only game in town. (I get less than 10 power cuts a year, but at least one of those will be long enough to cause my APC smart ups 1500 to totally collapse.) But I guess it is all down to the difference between looking at freeNAS as a storage solution or a backup solution. My live storage is elsewhere, once the freeNAS is up I will probably decant pictures, tunes and movies to a time capsule and use its single disk as live storage.