How I backup my servers
blog.sapico.me
blog.sapico.me
So backup server has rights only to read your prod server and prod server should not have write rights on backup server or at least “create only” if other option is too much hassle.
The feature you linked will only work with smart backends (e.g. VM running borg), so if I'm using S3 etc to store my backups I can't use that. That's a big caveat given the pricing difference between object storage (e.g. S3) and block storage (where I can run `borg serve --append-only`).
edit: Actually it seems borg doesn't support object storage at all.
Borg was just one example. And you can still run it effectively append-only by mounting the s3 endpoint locally and setting the right IAM permissions on the bucket.
What is needed (of the top of my head):
- runs on a nas, so it can coordinate the backups of the different machines
- nice browser based gui, with adequate security.
- can accomodate storage extensions for when space runs low
- backs up arbitrarily large files.
- uses VSS on Windows for snapshot consistency and to back up (permanently) open files
- uses global deduplication to save up significant space
- can do disk image as well as arbitrary file/folder selection backups
- offers advanced retention policies with good out of the box defaults
- allows both server and user side throtteling and pauses
- offers bandwith limiter
- offers archiving
- auto client agent updates
- bootable usb smart restore to changed hardware
- no 'call us for getting a price quotation'
- no complicated extras license needs for PC/Server/VM/DB/... , just a simple one site family non commercial use price.
Synology active backup for business, which I use now, goes there about 70% of the way. Any suggestions for alternatives?
Do we really assume that the "typical home/family" is using a NAS ? I think it far more common that they use "the cloud" like OneDrive, Google Drive, Apple iCloud, etc.
That is not a bad thing (privacy issues aside) as data resilience is much much better in the cloud, than almost any NAS you can setup at home. A paid OneDrive subscription gives you dual datacenter redundance (erasure coding) with unlimited versions for 30 days of all files, copy on write, and much more.
And regarding privacy, that is also not an issue that the "typical" user is concerned about. All the big vendors pretty much swear to never look at your data _unless you share it_, and for the "nerds", things like Cryptomator exists.
In my book, and one i use myself, a modern setup will have the cloud as primary storage, and "all you need" is a local box that can synchronize cloud data locally with versions, or make a versioned backup locally (the sync is useless if upstream gets corrupted). In theory a "real time" sync of data and scheduled BTRFS/ZFS/APFS snapshots every N minutes would be a versioned backup.
On top of that you then add a cloud backup of the synchronized data to a different cloud provider, and you have 3-2-1 backup in place.
This will, for most users, be cheaper than buying and running a NAS, as the average family probably has less than 2TB data, and comes with less risk of data loss due to hardware failure. The biggest threat to your data in the cloud is loss of access, which you can mitigate with the local mirror and the remote backup.
You can get Microsoft Family 365 for around €65/year (Microsoft HUP program and other deals), which includes 6 x 1TB OneDrive storage. Add to that a Backblaze Backup subscription at €65/yeaer and you have both primary cloud storage and cloud backup at €130/year. Where i live, just powering a 4 bay NAS for a year costs around €200.
> Synology active backup for business, which I use now, goes there about 70% of the way. Any suggestions for alternatives?
I personally use Arq for my client backups, but it does not support global deduplication, only per marchine.
For global deduplication, you might want to look at Restic (open source, stable), Duplicacy (commercial license) or Kopia (open source, beta state)
"the average family probably has less than 2TB data"
Backups here are 23TB, reduced to 7.5 after dedup. Maybe we're a bit more into tech than the average family, but 2TB does not sound reasonably if you have some retention.
We do have the O365 family sub. The Onedrive portion is used sparingly, and I would never count on it for backups.
I've always had something NAS like. Had Microsoft's WHS for a while on a spare PC until that was discontinued, now on my second Synology. It's not perfect but saved the day several times. Putting in the usb after a disk replacement and restoring to yesterday's point in time quickly is vital as disk crahes tend to happen at the least convenient moment.
Depends how you use it I guess. Mine works as a backup target for Arq backup, and holds encrypted, versioned client backups.
> Backups here are 23TB, reduced to 7.5 after dedup.
How much of that data is stuff you can easily download again ?
While we’re nowhere near professional photographs, we do take around 10k photos every year, and our family photo library stretches back almost two decades, and yet our family photo backup is “only” around 3TB, and 1.3TB compressed and deduplicated (quite a few RAW images there). Yearly growth is below 100GB, I know because I make yearly M-disc archives of the previous years photos, and they fit on a 100GB BD-XL.
Besides the photo backup, our documents, school work, music collection (ripped from CDs), purchased ebooks and more is less than 500GB (compressed, deduplicated)
I could probably save a couple hundred GB by leaving out music and ebooks, as the music is still on the CDs, and ebooks can (probably) be downloaded again, or they’re outdated enough that it doesn’t matter.
The absolute minimum would be our photos, which naturally are irreplaceable. Even documents will be outdated rather fast, and are mostly only kept for nostalgia.
Have you tested your ability to recover these backups at all levels? e.g. OS, DBMS and data.
If so, it would be worth including that in your article. Appreciating that not everyone has the same use case or the same risk appetite but it is important.
> By default, I already use the backup option on my VPS that is configurable per Cloud provider ( in my case, my Windows Servers run on Vultr and Linux servers on Hetzner). So I won’t go into detail on this.
Which basically handles backup and restore from systems within the cloud provider I'm using.
My data and db's are stored 'externally'. And I restore them regularely (locally) to test things ( eg. before I deploy to prod and/or repro a bug). So I test my backups ~every 2 weeks since I've added them in my dev workflow.
> Being able to reproduce a production issue in dev under 2 minutes from any device that I own, has been a time-saver
An extra precaution would be to simple sync the BackBlaze bucket to my local NAS, but that doesn't seem to be required atm.
Tldr: yes, I test my backups since I use them in my dev workflow.
Obligatory Tao of Backup: http://taobackup.com/history.html
Also 20% of your server's total cost seems pricey.
[1]: https://docs.hetzner.com/cloud/servers/getting-started/enabl...
As far as I'm aware, 20% seems to be the default ( reference: digital ocean).
It's very rare that I would need a server backup older than 7 days in practice. Because I backup my data a month. It could be better if it remained 1 backup per month though ( eg. In Sunday)
7 days is also within the period of time people can be on vacation. It's way too short.
Take a backup, migrate, test and restore, if required.
For the rare cases that I needed it, it's been worth it.
Imo handling day to day docs/data/mutables is harder. Hence im happy to use simple DVFS repo manager. No need to do backups of those :)
Stuff is stored usualy on desktop, laptop and server.
I'll check next week and reply here.