Comment by prof-dr-ir

16 hours ago

I am very interested in moving my photos and data to a self-hosted solution but am a little anxious about backups.

Do you simply trust hetzner to not lose the data on your 1TB storage box?

(I am aware that I am currently trusting google and dropbox to do just that.)

Set up your Hetzner boxes in a European location so that they are in the same network zone. Activate automatic snapshots and Hetzner does 7 snapshots (a full image of your box) a day. The snapshot is never saved at the same location as the server running your box, but at one of the other locations in the same network zone.

It is still viable to self-host everything from photos to mail yourself and sync to cloud/storage services as disaster recovery. It helps if you have an infrastructure background but anyone can set this up. Never trust just one service; no company is too big to fail and durability is always best effort, even if that effort is very good. Mail is the most annoying service to self-host, not because it's technically difficult but because deliverability is a long-term reputation function that easily deteriorates from misconfiguration or neglect. Nevertheless I've been my own MX and storage provider since the early '90s and it's too late to change my ways now, you just have to keep up with the gold standard as it varies.

The biggest hazard, especially if the whole family uses your stuff, is key-person risk, since infrastructure requires maintenance. The second biggest is being out of your depth in securing it.

My only regret in all my years of self-hosting was that time I returned a portable /24 to APNIC. Still stings even if it was the right thing to do, civically speaking.

I retain gmail & hotmail accounts for deliverability checks and as signup swamps.

To be fair if both google and dropbox can't take care of 1TB of data, who can?

My solution against photo anxiety is to actually look at them and decide to physically print the best ones every year. More likely to be used as gifts or just fun to look through them in a photo album, nobody is going to sit next to you on a phone or computer but bring out an old photo album and everyone is on it.

  • I do professional wedding photography as a side business.

    Yes, please print your photos! I love it when my clients print their photos, and I print my favorites as well. There's still something magical about a real, physical photo vs. digital.

    I have vast archives of digital photos and you know what? I barely look at them, but I have prints up all over my walls, in my wallet, etc and I enjoy them all the time.

Back it up to S3 glacier, or to Backblaze. The cost of it is pretty low, much lower than a VPS / bare metal box + 1 TB cost for the photo app hosting.

  • Technically I have no big doubts about S3 Glacier.

    But what happens if you don't use that stuff for a long time. You are in hospital when the bill needs to get paid. Your credit card gets stolen and the number needs to changed. Whatever personal crisis that you are not able to take care of life as usual for some weeks. They will just delete your data before you are back in business.

    Does anyone know how long it takes, how many warning mails will come? I have very little data in AWS, but I more or less constantly feeling it might happen to me. Maybe not because of such big crisis, but just the simple fact that my bank will reject the automatic payment requiring a PSD2 second factor and I miss the email...

    • It takes a couple of months for an unpaid AWS account to get it suspended. Then you have 30 days to reactivate it. Then you have 90 days before the data are actually wiped from the Glacier. You have half a year, or maybe more, to get your backup data.

      The price of Glacier Deep Archive is roughly $1/mo per terabyte. (I struggled to produce 500 GB of photos in 15 years.) Set up a dedicated AWS account, put $50 on it, set up a yearly auto-payment of $10, and you're likely safe for several years of nonpayment.

      Retrieval is not free though, something like $20-40 for retrieval from tape, and about $90 for a terabyte of egress traffic. Okay for the rare occasion of a full restore.

      Backblaze B2 is $6/mo per terabyte, and they only give you 44 days of grace period before deletion for nonpayment. But the traffic is free either way, up to 3x the amount stored per month. They are good for frequent full backups, and for doing full restores periodically.