Make Your Own Backup System – Part 1: Strategy Before Scripts

Another thing you can do is just run a container or a specific backup user. Something like with a systemd-nspawn can give you a pretty lightweight chroot "jail" and you can ensure that anyone inside that jail can't do any rm commands.

  pacman -S  arch-install-scripts            # Need this package (for debian you need debootstrap)
  pacstrap -c /mnt/backups/TestSpawn base    # Makes chroot
  systemd-nspawn -D /mnt/backups/TestSpawn   # Logs in 
  passwd                                     # Set the root password. Do whatever else you need then exit
  sudo ln -s /mnt/backups/TestSpawn /var/lib/machines/TestSpawn
  sudo machinectl start TestSpawn            # Congrats, you can now control with machinectl

Extra nice, you can use importctl to then replicate.

I fall into the "pull" camp so this is less of a worry. The server to be backed-up should have no permissions to the backup server. If an attacker can root your live server (with more code/services to exploit), they do not automatically also gain access to the backup system.

This is also why I use rclone copy instead of rclone sync for my backups, using API keys without permission to delete objects.

I do both. It requires two backup locations, but I want that anyway. My backup sources push to an intermediate location and the primary backups pull from there. The intermediate location is smaller so can hold less, but does still keep snapshots.

This means that neither my backup sources nor the main backup sinks need to authenticate with each other, in fact I make sure that they can't, they can only authenticate with the intermediate and it can't authenticate with them⁰. If any one or two of the three parts is compromised there is a chance that the third will be safe. Backing up the credentials for all this is handled separately to make sure I'm not storing the keys to the entire kingdom on any internet connectable hosts. The few bits of data that I have that are truly massively important are backed up with extra measures (including an actual offline backup) on top.

With this separation, verifying backups requires extra steps too. The main backups occasionally verify checksums of the data that hold, and send a copy of the hashes for the latest backup back to the intermediate host(s) where that can read back to compare to hashes generated¹ at the sources² in order to detect certain families of corruption issues.

--------

[0] I think of the arrangement as a soft-offline backup, because like an offline backup nothing on the sources can (directly) corrupt the backup snapshots at the other end.

[1] These are generated at backup time, to reduce false alerts from files modified soon after they are read for sending to the backups.

[2] The hashes are sent to the intermediate, so the comparison could be done there and in fact I should probably do that as it'll make sending alerts when something seems wrong more reliable, but that isn't how I initially set things up and I've not done any major renovations in ages.

> My preferred solution is to let client only write new backups, never delete.

I wish for syncoid to add this feature. I want it to only copy snapshots to the backup server. The server then deletes old snapshots. At the moment it requires delete permissions.

this is side effect of soc2 auditor approved disaster recovery policies.

company where i worked, had something similar. i spent a couple of months going through all teams, figuring out how disaster recovery policies are implemented (all of them were approved soc auditors).

outcome of my analysis was that in case of major disasters it will be easier to shut down company and go home than trying to recover to working state within reasonable amount of time.

If it doesn't affect your bottom line enough to do it right, then I guess it's ok?

It's also endlessly surprising how people over-think the process and requirements.

Possibly for legal purposes? Litigation holds are a PITA and generators of additional liability exposure, and backups can come back to bite you.

Wait, the prod db, like the whole thing? Losing 4 days of data? How does that work. Aren't customers upset? Not doubting your account, but maybe you missed something, because for a $1 billion company, that's likely going to have huge consequences.

Ha. That quote made me chuckle; it reminded me of a performance by the band Alice in Chains, where a similar quote appeared.

Re: BCDR solutions, they also sell trust among B2B companies. Collectively, these solutions protect billions, if not trillions of dollars worth of data, and no CTO in their right mind would ever allow an open-source approach to backup and recovery. This is primarily also due to the fact that backups need to be highly available. Scrolling through a snapshot list is one of the most tedious tasks I've had to do as a sysadmin. Although most of these solutions are bloated and violate userspace like nobody's business, it is ultimately the company's reputation that allows them to sell products. Although I respect Proxmox's attempt at cornering the Broadcom fallout, I could go at length about why it may not be able to permeate the B2B market, but it boils down to a simple formula (not educational, but rather from years of field experience):

> A company's IT spend grows linearly with valuation up to a threshold, then increases exponentially between a certain range, grows polynomially as the company invests in vendor-neutral and anti-lock-in strategies, though this growth may taper as thoughtful, cost-optimized spending measures are introduced.

- Ransomware Protection: Immutability and WORM (Write Once Read Many) backups are critical components of snapshot-based backup strategies. In my experience, legal issues have arisen from non-compliance in government IT systems. While "ransomware" is often used as a buzzword by BCDR vendors to drive sales, true immutability depends on the resiliency and availability of the data across multiple locations. This is where the 3-2-1 backup strategy truly proves its value.

Would like to hear your thoughts on more backup principles!

> having backups in multiple places (3-2-1 backup!) is important

Yeah and for the vast majority of individual cybernauts, that "1" is almost unachievable without paying for a backup service. And at that point, why are you doing any of it yourself instead of just running their rolling backup + snapshot app?

There isn't a person in the world who lives in a different city from me (that "1" isn't protection when there's a tornado or flood or wildfire) that I'd ask to run a computer 24/7 and do maintenance on it when it breaks down.

3-2-1 analogy is old. We have infinite flexibility on where we can put data unlike before cloud servers existed.

I'd at least have file system snapshots locally for easy recovery in case of manual mistakes, have it copied at a remote location using implementation A and let it snapshot there too, copy same amount on another location using implementation B and let it snapshot there too, so not only you'd have durability, implementation bugs on a backup process can also be mitigated.

zfs is a godsend for this and I use Borg as secondary implementation, which seems enough for almost any disasters.

> You should never trust an "rsync copy" or even crash consistent backups.

This leads you to the secret forbidden knowledge that you only need to back up your database(s) and file/object storage. Everything else can be, or has to be depending on how strong that 'never' is, recreated from your provisioning tools. All those Veeam VM backups some IT folks hoard like dragons are worthless.

Also if you have a NAS, don’t use the same hard drive type for both.

My personal external backup is two external drives in RAID1 (RAID0 wtfff?). One already failed, of course the Seagate one. It failed silently, too - a few sectors just do not respond to read commands and this was discovered when in-place encrypting the array. (I normally would avoid Seagate consumer drives if it wasn't for brand diversity. Now I have two WD drives purchased years apart.)

It's a home backup so not exactly relevant to most of what you said - just wanted to underscore the point about storage media sucking. Ideally I'd periodically scrub each drives independently (can probably be done by forcing a degraded array mode, but careful not to mess up the metadata!) against checksums made by backup software. This particular failure mode could also be caught by dd'ing to /dev/null.

> Ransomware is really clever these days and if you PUSH your backups, it can also encrypt or delete all your backups, because it has access to everything

That depends on how you have access to your backup servers configured. I'm comfortable with append-only backup enforcement for push backups[0] with Borg and Restic via SSH, although I do use offline backup drive rotation as a last line of defense for my local backup set. YMMV.

0 - https://marcusb.org/posts/2024/07/ransomware-resistant-backu...

Since you mentioned restic, is there something wrong with using restic append-only with occasional on-server pruning instead of pulling? I thought this was the recommended way of avoiding ransomware problems using restic.

> tested to be restorable as fast as possible

That depends on your goal, right? If it took me six months to recover my family photo backups, that'd be fine by me.

> So you could either use readonly media (like Blurays) or PULL is mandatory.

Or like someone already commented you can use a server that allows push but doesn't allow to mess with older files. You can for example restrict ssh to only the scp command and the ssh server can moreover offer a chroot'ed environment to which scp shall copy the backups. And the server can for example daily rotate that chroot.

The push can then push one thing: daily backups. It cannot log in. It cannot overwrite older backups.

Short of a serious SSH exploit where the ransomware could both re-configure the server to accept all ssh (and not just scp) and escape the chroot box, the ransomware is simply not destroying data from before the ransomware found its way on the system.

My backup procedure does that for the one backup server that I have on a dedicated server: a chroot'ed ssh server that only accepts scp and nothing else. It's of course just one part of the backup procedure, not the only thing I rely on for backups.

P.S: it's not incompatible with also using read-only media

I'm trialing a NAS with Immich, and then backing up the media and Immich DB dump daily to AWS S3 Deep Archive. It has Android and iOS apps, and enough of the feature set of Google Photos to keep me happy.

You can also store photos/scans on desktops in the same NAS and make sure Immich is picking them up (and then the backup script will catch them if they get imported to Immich). For an HN user it's pretty straight-forward to set up.

Is '25 years of photos' a North American measure of data I was previously unfamiliar with?

As bambax noted, you do in fact need a backup system -- you just don't realise that yet.

And you want a way of sharing data between devices. Without knowing what you've explored, and constraints imposed by your vendors of choice, it's hard to be prescriptive.

FWIW I use syncthing on gnu/linux, microsoft windows, android, in a mesh arrangement, for several collections of stuff, anchored back to two dedicated archive targets (small memory / large storage debian VMs) running at two different sites, and then perform regular snapshots on those using borgbackup. This gives me backups and archives. My RPO is 24h but could easily be reduced to whatever figure I want.

I believe this method won't work if Apple phones / tablets are involved, as you are not allowed to run background tasks (for syncthing) on your devices.

(I have ~500GB of photos, and several 10-200GB collections of docs and miscellaneous files, as unique repositories - none of these experience massive changes, it's mostly incremental differences, so it is pretty frugal with diff-based backup systems.)

Downloads and scans are generally trash unless deemed important.

For the phones and cameras, setup Nextcloud and have it automatically sync to your own home network. Then have a nightly backup to another disk with a health check after it finishes.

After that you can pick either a cloud host which your trust or get another drive of ours into someone else's server to have another locstion for your 2nd backup and you're golden.

I use syncthing... it's great for that purpose, Android is not officially supported but there is a fork, that works fine. Maybe you want to combine it with either ente.io or immich (also available for self-hosted) for photo backup.

I would also distinguish between documents (like PDF and TIFF) and photos - there is also paperless ngx.

PhotoPrism or Immich are solid self-hosted options that handle deduplication and provide good search/tagging for family photos. For cloud, Backblaze B2 + Cryptomator can give you encrypted storage at ~$1/TB/month with DIY scripts for uploads.

Others have already mentioned Syncthing[^1]. Here's what I'm doing on a budget since I don't have a homeserver/NAS or anything like that.

First you need to choose a central device where you're going to send all of the important stuff from other devices like smartphones, laptops, etc. Then you need to setup Syncthing, which works on linux, macos, windows and others. For android there's Syncthing-fork[^2] but for iOS idk.

Setup the folders you want to backup on each device, for android, the folders I recommend to backup are DCIM, documents, downloads. For the most part, everything you care about will be there. But I setup a few others like Android/media/WhatsApp/Media to save all photos shared on chats.

Then on this central device that's receiving everything from others, that's where you do the "real" backups. I my case, I'm doing backups to a external HDD, and also to a cloud provider with restic[^3].

I highly recommend restic, genuinely great software for backups. It is incremental (like BTRFS snapshots), has backends for a bunch of providers, including any S3 compatible storage and if combined with rclone, you have access to virtually any provider. It is encrypted, and because of how it was built, can you still search/navigate your remote snapshots without having to download the entire snapshot (borg[^4] also does this), the most important aspect of this is that you can restore individual folders/files. And this crucial because most providers for cloud storage will charge you more depending on how much bandwidth you have used. I have already needed to restore files and folders from my remote backups in multiple occasions and it works beautifully.

[^1]: https://github.com/syncthing/syncthing [^2]: https://github.com/Catfriend1/syncthing-android [^3]: https://github.com/restic/restic [^4]: https://github.com/borgbackup/borg

>I still haven't found anything good.

I have used pCloud for years with no issue.

Also external "slow" storage drives are fairly inexpensive now as a third backup if your whole life's images and important documents are at stake.

Always best to keep multiple copies of photos or documents that you care about in multiple places. Houses can flood or burn, computers and storage can fail. No need to be over-paranoid about it, but two copies of important things isn't asking too much of someone.

Struggling too.

For me one win/mac with backblaze. Dump everything to that machine. Second ext. Drive backup just in case.

Check out ente.io - it is really good!

I recently found that Nextcloud is good enough to "collect" the photos from my family onto my NAS. And my NAS makes encrypted backups to a cloud using restic.

A NAS running Immich, maybe?

Apple devices and a family iCloud storage plan.

You do need a backup. But before that, you need a family NAS. There are plenty of options. (But a NAS is not a backup.)

What does dirvish do better or simpler than rsync?

I built a disaster recovery system using python and borg. It snapshots 51 block devices on a SAN and then uses borg to backup 71 file systems from these snapshots. The entire data set is then synced to S3. And yes, I've tested the result in a offsite: recovering files systems to entirely different block storage and booting VMs, so I'm confident that it would work if necessary, although not terribly quickly, because the recovery automation is complex and incomplete.

I can't share it. But if you contemplate such a thing, it is possible, and the result is extremely low cost. Borg is pretty awesome.

If you die suddenly tomorrow, what would you want your family to recover? What would you want your grandchildren to have access to in a few decades? That's your valuable data. They may not need or want to inherit from hundreds of thousands of files. Chances are that a few key photos, videos, and text would be enough.

This comment made me rethink what I have that is actually valuable data. My photos alone even if culled down to just my favorites would probably be at least a few gigs. Contacts from my phone would be small. Other than that I guess I wouldn't be devastated if I lost anything else. Probably should put my recovery keys somewhere safer but honestly the accounts most important to me don't have recovery keys.

Curious what you consider valuable data?

Edit: I should say for pictues I have around 2Tb right now (downside of being a hobby photographer)

> All external disks are Samsung T7 Shield

And make sure to use at least a different batch, or better a different model. Same batch or same model tend to fail at the same time (usually if you need to restore data and the disk is under heavy load).

I think generally speaking, databases are resilient to this so taking a snapshot of the disk at any point is sufficient as a backup. The only danger is if you're using some sort of on-controller disk cache with no battery backup, then basically you're lying to the database about what has flushed and there can be inconsistencies on "power failure" (i.e. live snapshot).

But for the most part as especially in the cloud, this shouldn't be an issue.

It's not clear if there are other places that application state is being stored, outside your database, that you need to capture. Do you mean things like caches? (I'd hope not.)

pg_dump / mysqldump both solve the problem of snapshotting your live database safely, but can introduce some bloat / overhead you may have to deal with somehow. All pretty well documented and understood though.

For larger postgresql databases I've sometimes adopted the other common pattern of a read-only replica dedicated for backups: you pause replication, run the dump against that backup instance (where you're less concerned about how long that takes, and what cruft it leaves behind that'll need subsequent vacuuming) and then bring replication back.

[dead]

Other things that matter (some overlap):

* Is the file userland-compressed, filesystem-or-device-compressed, or uncompressed?

* What are you going to do about secret keys?

* Is the file immutable, replace-only (most files), append-only (not limited to logs; beware the need to defrag these), or fully mutable (rare - mostly databases or dangerous archive software)?

* Can you rely on page size for (some) chunking, or do you need to rely entirely on content-based chunking?

* How exactly are you going to garbage-collect the data from no-longer-active backups?

* Does your filesystem expose an accurate "this file changed" signal, or better an actual hash? Does it support chunk sharing? Do you know how those APIs work?

* Are you crossing a kernel version that is one-way incompatible?

* Do you have control of the raw filesystem at the other side? (e.g. the most efficient backup for btrfs is only possible with this)

It works in theory, but usability is almost non-existent with this approach unless someone creates an app that interacts with this database and provides file system-like access to users. Any normal human would be better off with Dropbox or Google Drive.

Almost like GMail Drive back in the day but worse.

https://en.m.wikipedia.org/wiki/GMail_Drive

artifactory is great, and i do something very similar

Occasionally you need to try restoring from your backups.

Obviously a full restore gives you full confidence, and it goes down from there.

Ideally try to restore about 10% of your content every month,but really it depends on how high stakes your backups are.

Backblaze is great, but restores can be a bit time consuming, even on a fast FTTP connection.

I do have BackBlaze on my desktop, but I also have UrBackup running on all the computers in the house which backs up to a RaidZ2 array, and then a daily offsite backup of the "current" backup (which is just the files stored in a directory in UrBackup) via restic and rclone to JottaCloud.

VM's and containers backup to Proxmox Backup Server and the main datastore of that is also shipped offsite every day, as well as a second Proxmox Backup Server locally (but separate from the rack).

I test restores monthly and so far so good.

Do you mean $99?

  > without mentioning RPO, RTO, or even RCO

  > Those terms are handy for anyone not familiar with the space to go do some further googling.

  RPO: Recovery Point Objective 
  RTO: Recovery Time Objective
  RCo: Recovery Consistency 

FWIW, I didn't know these terms until you mentioned them, so I'm not an expert. Please correct me if I'm misunderstanding or being foolishly naive (very likely considering the previous statement). But as I'm only in charge of personal backups, should I really care about this stuff? My priorities are that I have backups and that I can restore. A long running rsync is really not a big issue. At least not for me.

https://francois-encrenaz.net/what-is-cloud-backup-rto-rpo-r...