3-2-1 Back-Up Strategy; Replication on Same NAS?

[mcmehrtens]

I built a NAS last winter (see spoiler below) for storing terabytes of photos that were wandering around in cloud storage from my late father. It also doubles as my HomeLab server. The important fact for my questions is that it has six 8 TB 3.5" HDDs. I'm trying to figure out my HomeLab/data 3-2-1 back-up strategy and I had a couple of questions.

1. The 2 in 3-2-1 stands for "two different media". I often see this referred to as two different machines or a second set of HDDs, but doesn't two different media imply two different "storage" media, e.g., HDD and SDD or HDD and Blu-ray Discs?
2. I have a data that is stored exclusively on my NAS (one copy). If I use a remote NAS as a backup or a cloud backup, that will (maybe) satisfy the second "media" and will definitely satisfy the remote backup (second copy), but what's the best (and most cost-effective) way to get this third copy of the data?
   1. I've read of folks rotating high-capacity hard drives, but there's an unreliable human element in this method.
   2. I could build a second NAS at home, but that's definitely not cost-effective.
   3. Would splitting my NAS into two storage pools and replicating to the second storage pool be a reasonable way to get a third copy of the data or does it existing on the same device defeat the point? If this is a viable option, how would you recommend I set up the storage pools (two raidz1 pools? a raidz2 pool for the working data and just a normal raid pool for the other two drives?)?

Like most people on this forum, I care a lot about keeping my data safe and intact, but this 3-2-1 backup strategy is really testing me. I know at the end of the day it doesn't matter exactly how I implement this—my homelab won't be audited any time soon—as long as it's good enough for my use case (and my peace of mind), but I'm not going to spend the time setting this up if I'm not going to do it right.

Here's my NAS:

Show : My TrueNAS system

PCPartPicker Part List

CPU Cooler: Noctua NH-U9S chromax.black 46.4 CFM CPU Cooler (Purchased For $74.85)
Storage: Kingston NV1 250 GB M.2-2280 PCIe 3.0 X4 NVME Solid State Drive (Purchased For $26.74)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Case: Fractal Design Node 304 Mini ITX Tower Case (Purchased For $122.52)
Power Supply: SeaSonic FOCUS Plus Platinum 550 W 80+ Platinum Certified Fully Modular ATX Power Supply (Purchased For $157.81)
Wired Network Adapter: TP-Link TX201 2.5 Gb/s Ethernet PCIe x1 Network Adapter (Purchased For $32.09)
Case Fan: Noctua A14 PWM chromax.black.swap 82.52 CFM 140 mm Fan (Purchased For $28.76)
Case Fan: Noctua A9x14 HS-PWM chromax.Black.swap 33.84 CFM 92 mm Fan (Purchased For $23.49)
Case Fan: Noctua A9x14 HS-PWM chromax.Black.swap 33.84 CFM 92 mm Fan (Purchased For $23.49)
Custom: Noctua NA-SEC1 Accessory 4-pin Extension Cables (Purchased For $9.58)
Custom: Cable Matters VGA to HDMI Converter (VGA to HDMI Adapter) with Audio Support (Purchased For $20.32)
Custom: Intel Xeon E-2324G Rocket Lake 3.1 GHz 8MB L3 Cache LGA 1200 65W BX80708E2324G Server Processor (Purchased For $300.66)
Custom: Cable Matters Internal Mini SAS HD to SATA Cable 3.3 Feet, 1m (SFF-8643 to SATA Forward Breakout) (Purchased For $21.39)
Custom: SUPERMICRO MBD-X12STL-IF-O Mini ITX Server Motherboard LGA 1200 Intel C252 (Purchased For $298.53)
Custom: Micron 32GB DDR4-3200 ECC UDIMM 2Rx8 CL22 (Purchased For $83.75)
Custom: Micron 32GB DDR4-3200 ECC UDIMM 2Rx8 CL22 (Purchased For $83.75)
Total: $2302.77
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2023-07-21 15:54 EDT-0400

Thanks in advance,
Matthew Mehrtens

 

[samarium]

1 computer, 2 different media pools, 1 power surge or 1 fire, both media pools crisped? Entirely possible.

 

[mcmehrtens]

1 computer, 2 different media pools, 1 power surge or 1 fire, both media pools crisped? Entirely possible.

Yeah, I understand that. This data that is currently exclusively stored on the NAS will also be backed up remotely to Backblaze B2 or an offsite NAS—but that’s still only 2 copies of the data, hence where having two data pools on the primary NAS come in.

Is having two pools on the same NAS and replicating one to the other a recommended way to accomplish the 3 copies of data specified by the 3-2-1 backup ideology? And what of my other question regarding the two different forms of media?

Thanks!

 

[samarium]

Not good enough for me, however you get to make your own choice. For me 321 is just some clues to do a better job. If you feel that 2 copies on the same machine is sufficient diversification of storage locally, then you are done with that aspect. You know the risks, and have made your judgement call, and likely will never need to use the backups.

 

[mcmehrtens]

Not good enough for me, however you get to make your own choice. For me 321 is just some clues to do a better job. If you feel that 2 copies on the same machine is sufficient diversification of storage locally, then you are done with that aspect. You know the risks, and have made your judgement call, and likely will never need to use the backups.

I just want to make it clear that in addition to the 2 copies on the same machine, there would be a third copy offsite on a remote NAS or on the cloud. I'm just wondering if 2 copies on the same machine is a recommended way to get to your third copy of the data or if it it's effectively pointless.

Judging from your opposition to the idea, would you be more likely to recommend doing backups to high capacity hard drives that I rotate on a weekly basis?

 

[mcmehrtens]

1. The 2 in 3-2-1 stands for "two different media". I often see this referred to as two different machines or a second set of HDDs, but doesn't two different media imply two different "storage" media, e.g., HDD and SDD or HDD and Blu-ray Discs?

Also, can anyone answer my first question? What is SOP for the "two different media" in a homelab environment?

I should also clarify that I could easily get three copies of my data with 1.) copy on my NAS, 2.) copy on the remote NAS, and 3.) copy on the cloud, but I want to avoid paying for the cloud if I can. I expect to be storing data for decades and those cloud costs will balloon over time. So, I'm looking for a convenient way to get a third copy of the data that's only stored on my NAS--without resorting to the cloud.

 

[samarium]

Homelab has infinite variety, as such I don't think there is a SOP. My homelab, my procedures. You are taking a very prescriptive look at 321 expecting it to solve all your problems it seems, while I look at as suggestive. I also think you are over complicating two different media, it doesn't say two different media technologies. You can read it this way, and it will provide diverification, but I don't think it materially changes to protection unless you are after long term archival storage, and then you should address this directly as an extra and different backup, and it is more complicated than normal backups. You also want to minimize manual steps if possible. Physical offsite media swap would be difficult to automate though.

I want multiple copies of my data, preferably away from a single point of failure in/of the homelab, and also offsite to guard against fire/theft/disaster.

I also want to avoid cloud cost, and prefer hosting a small low power backup device at friends/relatives, or even trade off offsite backup with a friend, rather than pay offsite hosting hosts. I also compromise on what I backup to offsite, reducing it to a minimum of current data. Bulk data I can just ask them to store a HDD/SSD, easy to get 1TB for a SSD, maybe one of several a broken off mirrors of ZFS pool that I can alternate bringing back and syncing up. Minimizing external storage often means minimizing the data that I want backed up externally. If the xxD was appropriately connected remotely, I could even sync it up remotely using nbd or something similar.

You want to offsite but you won't want to pay for it. I understand, however you don't get the protection from fire/theft/disaster unless you have offsite. Same home lab, means same power circuit, unless you detach a copy into an offline case. I prefer different machines so a destructive issue on one machine won't affect the second copy. Still have the same onsite fire/theft/disaster risk of taking out the entire homelab. If you are replicating data zfs to zfs there is the risk of replicating corruption, so you could use maybe rysnc/rsnapshot for the second leg, but corrupt files will still get rsync'd but will be protected by the older rsnapshots, and this may also be the same with zfs snapshots and corruption, may not too, corruption means unpredictable. You still may get pool metadata corruption which kills the entire pool, unlikely but possible. See recent issue where deleteing a replictaed encrypted incremental snapshot resulted in system crash. Also ensure you run regular zpool scrubs, zfs can't protect you from bit rot unless you allow it to regularly check the data checksums and self heal if possible, or alert you to the need for data recovery.

There was a recent resource posted about using tape backup, a little expensive on capital, but low ongoing if you don't have mechanical failure, and easy to ask someone to store a tape or 5, and they likely don't have the tech to even read the tape. Compared to ongoing cloud costs it is a good deal as long as you have an offsite storage location, could even be a bank safe deposit box, cheap if you already have one, or a storage facility of some kind as long as the costs work for you.

Ultimately you have to decide if what you perceive as risks have appropriate mitigation. Write the risks down, write down the mitigations, write down the which risks you are willing to accept. Write down the costs too, storage and restoration. Some data you can just record that you have it, and maybe how to get it back, and accept that it isn't important enough to backup. Partition your data into different protection classes, and arrange the backups approriately. If you can't figure a cheap offsite storage, then either accept no offsite storage or accept paying the costs, just don't delude yourself that multipe onsite copies are a replacement for offsite. Offsite protects against different things. Double check against 321, does it make you ask more questions about your backup plan? If so iterate.

 

[garm]

I'm just wondering if 2 copies on the same machine is a recommended way to get to your third copy of the data or if it it's effectively pointless.

I wouldn’t say it’s pointless (saved my bacon a while back) but it lacks some key points. The reason behind 3-2-1 is to ensure that the data has no single point of failure. Putting 2 copies of your data on the server makes them share several single point of failures. But so does having them at the same location as well (house hit by asteroid).

It’s a game of statistics and acceptable risk. When I nuked my pool I hade previously accidentally put a backup of my Nextcloud data vault on the application dataset (it was a bad time) and that rescued some immensely important content that was lacking proper backups.

My strategy for backups depends on the value of the data. For the high value stuff (edited photos and videos, documents, etc) I use a “cloud” provider to mirror the content of the dataset. I also burn snapshots to MDisc blueray and put in a fire resistant case (not properly implemented yet).

For critical data that can not be lost for the next few generations all I do is to increase the offsite backups and MDisc copies to two and keep the second MDisc in a different geographic location.

“Bulk” data (raw photos and videos, transient documents, and other easily replaced content is generally not backed up of simply added to the first cloud backup as space allows. Nothing else is done to it.

That’s my two cents, from someone who nuked his pool last year and had his first data loss in over a decade thanks to ZFS

 

[Davvo]

My personal interpretation of the 3-2-1 rule is as follows: 3 places that store data in total, 2 in different systems (on site), and 1 off site.

In my case, I have the NAS, an external HDD, and an external HDD stored in another place.

I backup only the most important data in the off-site drive because it's not pratical to do otherwise... I update it every six months or so.

This is unaccepltable as enterprise level strategy, but for me is enough.

 

[tip0666]

I built a NAS last winter (see spoiler below) for storing terabytes of photos that were wandering around in cloud storage from my late father. It also doubles as my HomeLab server. The important fact for my questions is that it has six 8 TB 3.5" HDDs. I'm trying to figure out my HomeLab/data 3-2-1 back-up strategy and I had a couple of questions.

1. The 2 in 3-2-1 stands for "two different media". I often see this referred to as two different machines or a second set of HDDs, but doesn't two different media imply two different "storage" media, e.g., HDD and SDD or HDD and Blu-ray Discs?
2. I have a data that is stored exclusively on my NAS (one copy). If I use a remote NAS as a backup or a cloud backup, that will (maybe) satisfy the second "media" and will definitely satisfy the remote backup (second copy), but what's the best (and most cost-effective) way to get this third copy of the data?
   1. I've read of folks rotating high-capacity hard drives, but there's an unreliable human element in this method.
   2. I could build a second NAS at home, but that's definitely not cost-effective.
   3. Would splitting my NAS into two storage pools and replicating to the second storage pool be a reasonable way to get a third copy of the data or does it existing on the same device defeat the point? If this is a viable option, how would you recommend I set up the storage pools (two raidz1 pools? a raidz2 pool for the working data and just a normal raid pool for the other two drives?)?

Like most people on this forum, I care a lot about keeping my data safe and intact, but this 3-2-1 backup strategy is really testing me. I know at the end of the day it doesn't matter exactly how I implement this—my homelab won't be audited any time soon—as long as it's good enough for my use case (and my peace of mind), but I'm not going to spend the time setting this up if I'm not going to do it right.

Here's my NAS:

Show : My TrueNAS system

PCPartPicker Part List

CPU Cooler: Noctua NH-U9S chromax.black 46.4 CFM CPU Cooler (Purchased For $74.85)
Storage: Kingston NV1 250 GB M.2-2280 PCIe 3.0 X4 NVME Solid State Drive (Purchased For $26.74)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Storage: Seagate EXOS Enterprise 8 TB 3.5" 7200 RPM Internal Hard Drive (Purchased For $165.84)
Case: Fractal Design Node 304 Mini ITX Tower Case (Purchased For $122.52)
Power Supply: SeaSonic FOCUS Plus Platinum 550 W 80+ Platinum Certified Fully Modular ATX Power Supply (Purchased For $157.81)
Wired Network Adapter: TP-Link TX201 2.5 Gb/s Ethernet PCIe x1 Network Adapter (Purchased For $32.09)
Case Fan: Noctua A14 PWM chromax.black.swap 82.52 CFM 140 mm Fan (Purchased For $28.76)
Case Fan: Noctua A9x14 HS-PWM chromax.Black.swap 33.84 CFM 92 mm Fan (Purchased For $23.49)
Case Fan: Noctua A9x14 HS-PWM chromax.Black.swap 33.84 CFM 92 mm Fan (Purchased For $23.49)
Custom: Noctua NA-SEC1 Accessory 4-pin Extension Cables (Purchased For $9.58)
Custom: Cable Matters VGA to HDMI Converter (VGA to HDMI Adapter) with Audio Support (Purchased For $20.32)
Custom: Intel Xeon E-2324G Rocket Lake 3.1 GHz 8MB L3 Cache LGA 1200 65W BX80708E2324G Server Processor (Purchased For $300.66)
Custom: Cable Matters Internal Mini SAS HD to SATA Cable 3.3 Feet, 1m (SFF-8643 to SATA Forward Breakout) (Purchased For $21.39)
Custom: SUPERMICRO MBD-X12STL-IF-O Mini ITX Server Motherboard LGA 1200 Intel C252 (Purchased For $298.53)
Custom: Micron 32GB DDR4-3200 ECC UDIMM 2Rx8 CL22 (Purchased For $83.75)
Custom: Micron 32GB DDR4-3200 ECC UDIMM 2Rx8 CL22 (Purchased For $83.75)
Total: $2302.77
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2023-07-21 15:54 EDT-0400

Thanks in advance,
Matthew Mehrtens

3 boxes=same data(same data/exact duplicate)change ip and operational (automatically/ email, /preferably)…!!!!! Working and operational…. different physical location…. Now implement more redundancy….!!!!! x3.. now back to reality.. same concept/your budget…

Yeah, I understand that. This data that is currently exclusively stored on the NAS will also be backed up remotely to Backblaze B2 or an offsite NAS—but that’s still only 2 copies of the data, hence where having two data pools on the primary NAS come in.

Is having two pools on the same NAS and replicating one to the other a recommended way to accomplish the 3 copies of data specified by the 3-2-1 backup ideology? And what of my other question regarding the two different forms of media?

Thanks!

 

[tip0666]

My personal interpretation of the 3-2-1 rule is as follows: 3 places that store data in total, 2 in different systems (on site), and 1 off site.

In my case, I have the NAS, an external HDD, and an external HDD stored in another place.

I backup only the most important data in the off-site drive because it's not pratical to do otherwise... I update it every six months or so.

This is unaccepltable as enterprise level strategy, but for me is enough.

Budget friendly…I keep all 3 boxes @home. 1xmaster cold (spinning rust)legacy(20y/o,moBo,ecc) archive only, I’ll spin up @random…
1xplex,frequent data(pics/docs/cloud),,, 1xbenchmark//xlightdeployment/trial(charts especially)max/powerfailure/boot/benchbehavior/continue,activity///unexpected,behavior/xreset,