1

For archiving and backup, I created the following strategy:

  • Server 1 (Local Network in Office) running 24/7, Linux Ubuntu NAS on DIY Odroid XU4 with Cloudshell 2 and Raid-1 (2x 8TB)
    • this is my primary backup and archive, important directories on my computers/smartphone etc. get directly synced once they're connected to the local network or if I select files manually for archive
  • because this is not a true backup (imagine my house burns, server gets stolen), I have another server (2), which is physically (far) separated from Server 1
    • this server is online only once a week, when it rsyncs all data from Server 1 or when I need to perform other more intensive tasks
    • it has 32 GB ECC RAM

I know this setup should be the other way around, but the XU4 is pretty silent and I don't want that monster server running 24/7 in my house. I was wondering how bad this setup is in terms of data corruption:
- I assume server 2 will sync files, no matter if they are corrupted or not from server 1 (e.g. due to non-ECC RAM in Server 1)
- is there a feasible approach to tell Server 1: only sync files that changed on purpose, do not sync files corrupted by RAM errors
- maybe Server 1 could even check with server 1, e.g.: "I do not have any live connections currently, File xx still changed - lets check with Server 2 if this was due to a RAM error"

Any help appreciated!

Alex
  • 171
  • 8
  • Thanks Sven for pointing this out and apologies for my unclear description. This question IS business related - Server 1 is running in my Office (as Freelancer, this Office is located at Home) – Alex May 27 '18 at 10:17
  • I suggest having one computer at home with a good sized disk, good incremental backup software, and an external hard drive that's stored offsite unless you're running backups. Add to that nightly / constant cloud backups for anything critical that's not huge and you're almost done. Read about the 3-2-1 backup rule - from memory it's 3 copies, 2 media, at least one offsite location. – Tim Jun 11 '18 at 09:19

1 Answers1

1

ECC RAM protects you from some memory errors. The files you're transferring are on disk. They will only travel through the RAM momentarily, if at all, when you are synchronising them to your other computer. It doesn't have an API where you can "query" for in-memory errors for a specific file.

Also, this kind of corruption just generally doesn't happen. What you should be concerned about is file system corruption, not memory corruption. And honestly there's not much you can do about that. There's no way a backup software can know if a file is corrupted, because it just copies files byte for byte.

As long as you test your restores you are way over thinking all of this. Copy your files to your remote server, test restores occasionally, and sleep easy.

As an aside, you may want to rethink copying the to a server you own. The price of running that server, maintenance, patching, internet links that can sustain the bandwidth in both directions, the inconvenience of having a piece of hardware in your home. A cloud backup service such as backblaze can do this for $50/year. You can put terrabytes of data into Amazon Glacier for pennies a day if you have backup software that supports this.

I used to be like you - wanted super control over all aspects of my backups etc. Then I saw the light a few years ago and I have one on-site backup (usually a USB hard disk that can be quickly transferred to a new machine in the event of a total hardware failure), and a cloud backup (my personal stuff is in Backblaze, non-personal stuff in AWS S3 and AWS Glacier).

BaronSamedi1958
  • 12,510
  • 1
  • 20
  • 46
Mark Henderson
  • 68,316
  • 31
  • 175
  • 255
  • Awesome, thanks for the explanation. This makes sense and find that I am probably good with what I have. I understand that, for backup only, Internet servers are much better, but I'm running many jobs on these 2 Servers and, for me, it is cheaper doing it all on my own instead of paying a monthly fee. In addition, it is difficult finding a cheap cloud backup in Europe, and backing up to US based servers will almost get you in jail here with the new GDPR (well, not exactly, but its crazy at the moment and my setup saves me a lot of headache). – Alex May 29 '18 at 06:09