Backup strategy for home user (involving a FreeNAS)

[Jonas Camillus Jeppesen]

Hi all

My basic question is the next paragraph, read past that if you want my thoughts, wishes and a bit of discussion.

What is a good backup strategy for a 100-500 GB of important data (think photos) stored on a FreeNAS system? *

* No need to discuss off-site vs. on site backup. I know the difference and I think most solution being usable on-site could be implemented off-site instead, if that is wished for.

RAID is not a backup, neither are ZFS snapshots. Admin error might delete all data, or 4 disks might die simultaneously in a raidz3 setup. To discuss a backup strategy let me talk about my data and how I store them now:

1. Unimportant data like TV-shows and movies are stored on varies local clients (my laptop, my desktop, media center, etc.) and it is straight forward to migrate that to the more centralized FreeNAS system. I guess these TB of unimportant data are exempt from the backup strategy.
2. Important data like family photos are stored on a multitude of external hard drives laying in drawers and on local live systems (my laptop, my desktop etc.). It would be nice to store all these in one safe place. Hopefully that safe place could be FreeNAS + a backup strategy.
3. Important documents, research data, code, etc. are mostly stored on local live systems (laptops, desktops etc.) and they are needed locally while on the move (I need my work related documents while on the train without an internet connection). So the FreeNAS system should probably serve as backup location for those, not a centralized location for daily use.

After writing this down I realize that only 2. (important data being stored, not backed up, on the FreeNAS system) needs a backup strategy. 1. is not worth backing up and 3. use the FreeNAS system as its backup.

So what is a good backup strategy for those few hundred gigabytes of important data (again, think photos) stored on the FreeNAS? Here is what I've come up with

• A second FreeNAS system. It seems like a good, albeit expensive, solution.
• Blu-Ray. Dual layer discs hold 50GB - not very attractive to store sets of 10 of these.
• External hard disks. Offline disks seems like a very unreliable media (compared to an active ZFS file system).

Thank you for your input!

/ Jonas

 

[depasseg]

The simplest backup would be using ZFS replication. Whether that is to a dataset (Drive or drives) in the same machine, another local FreeNAS server, or an off-site FreeNAS server.

Other options are a backup software package like crashplan, or a cloud service like S3cmd.

 

[Jonas Camillus Jeppesen]

The simplest backup would be using ZFS replication. Whether that is to a dataset (Drive or drives) in the same machine, another local FreeNAS server, or an off-site FreeNAS server.

Other options are a backup software package like crashplan, or a cloud service like S3cmd.

Thank you for your comments.

So you are recommending another live system (e.g. a second FreeNAS system) as backup? Should this second system be even more reliable than the first one for this to make sense? I.e. should it use smaller disks with better URE rate, or does a duplicate of the first system make sense?

Should I understand that there is no good media for backup which can rest on shelves without being powered (with 500GB-1.0TB capacity)? So backup is almost always done to powered/live file systems?

 

[depasseg]

I haven't seen anyone using Tape or optical media to do backups.

There are a couple threads kicking around where users have used ZFS replication to a locally attached datastore (like a hotplug drive in your primary FreeNAS) and then detached it to put on a shelf. As long as you are good about connecting, attaching, running replication, detaching, putting away, this should be good enough.

If you aren't very schedule oriented, then a second FreeNAS box which is configured to do this automatically would be beneficial. It wouldn't need to be any more reliable than the first. And if you didn't think you were going to change a lot of data, you could just turn it off after a replication task. If your data changes frequently and you have many snapshots, then the power-down idea might not be useful, because once you power it on, it will sync all the snapshots in the dataset.

The way I think of it is: Take snapshots in case I do something stupid. Use replication in case the FreeNas does something stupid (like crashes).

 

[Jonas Camillus Jeppesen]

So assuming that failure of the two NAS systems are independent of one another then the probability of total failure (data loss) is P_total = P_1 * P_2. Putting management (my admin skills), environment and FreeNAS bugs asside this would be the same as mirroring a raidz2 vdev, like this http://i.imgur.com/W0x0sPT.png , right?

But management, environment and FreeNAS bugs are probably worth making independent by putting half of the drives in a second FreeNAS system, compared to mirroring a raidz2 vdev.

 

[depasseg]

The end effect (total data loss) is the same, but in my opinion, the total failure "P_total" would not be due to equipment failure of both servers. It would be due to theft or fire. Hence the need for offsite. Seriously, look into Amazon S3. It's cheap and would handle your use case of only 100-500GB of very important data.

If that isn't feasible, and you have the ability (money and space), then 2 servers would be my second choice.

Last choice would be a hotplug drive in the primary server.

 

[JoeVulture]

I use rsync to backup my data to external HDDs via an eSATA dock. Every Sunday, I put in a new drive for a full backup, and then Monday through Saturday, incremental backups are performed. I keep four weeks of drives.

 

[Jonas Camillus Jeppesen]

I use rsync to backup my data to external HDDs via an eSATA dock. Every Sunday, I put in a new drive for a full backup, and then Monday through Saturday, incremental backups are performed. I keep four weeks of drives.

Which file system do you use on your external hard drives, ZFS so you can do snapshots and replication? Do you have any redundancy on your backup drives, e.g. do you have the external hard drives mirrored?

What do you incremental backups consist of throughout the week? Do you leave the Sunday-hard drive in its dock for a whole week end perform snapshots?

 

[Jonas Camillus Jeppesen]

depasseg: From what you say I gather that most people would use live file systems for backup (FreeNAS, 2nd FreeNAS, Amazon Web Services, etc.), and that few would consider using optical disks or tape for backup.

So I will hopefully stop thinking about rotating BD-RE disks, and start thinking about a second live file system to backup to. Right now rotating external hard disk drives, which can easily be moved off-site, seems appealing, Afterall my internet connection is only 100 Mbit/s ~ 12 MB/s, and best case is ~12 hours to transfer 500GB. So manually moving drives once a month might be a better option.

 

[depasseg]

I haven't seen any posts asking about backing up to Tape or optical media. Of course, I've only been here a little while and I haven't searched the forums.

How often does your 500GB change? And how much changes? Usually, the initial backup takes the longest, and then the replication will send only whatever has changed. Way back when I had a couple TB's and it took around a month over my connection, then the delta's were much quicker.


Look into Crashplan and giving a USB drive to a family or friend. You can do the initial backup locally, take the drive to someone else's computer, install crashplan, and then continue backups to their machine.

 

[JoeVulture]

Which file system do you use on your external hard drives, ZFS so you can do snapshots and replication? Do you have any redundancy on your backup drives, e.g. do you have the external hard drives mirrored?

What do you incremental backups consist of throughout the week? Do you leave the Sunday-hard drive in its dock for a whole week end perform snapshots?

Disclaimer: I'm a home user - I don't know if I would consider this acceptable for a small business, let alone an enterprise. That said, my scheme works fine for me, YMMV.

I have snapshots setup on my RAID-Z2 every 15 minutes with a two-week retention; they might make it to the backup drive, I haven't verified that (it's not too important). Rsync creates hard links so I can go to any given day directory (Sun-Sat) and get the full snapshot for that day.

Each drive is formatted as UFS (no mirroring) with the same label so that they can easily be mounted. The drive is unmounted automatically at the of the rsync, although it is still in the eSATA dock for the week, powered on. I'm not using built-in rsync tasks, it's my own custom rsync script that I developed years ago when I used Linux for my NAS.

When the week is over (Sunday morning), I remove the drive, place it in a storage case and use it again four weeks later. Every now and then, I do a diff on the external drives and the RAID-Z2 array for consistency.

 

[Scharbag]

Interesting discussion. Made me look at my approach to data availability and, for a home user, I think it is a good approach.

Here is what I do:

• Single FreeNAS server with:
  • RaidZ2 production pool that stores all types of data
    • HTPC files
    • Data
    • Music
  • RaidZ1 backup pool
    • all data is backed up once a day

My strategy involves using a variety of periodic snapshots (retained for at least a week) on the production pool so I can restore previous version if I need to. To backup the files, I use Rsync each day to archive all files to the backup pool. Prior to the Rsync tasks, I take a snapshot of the backup pool that is retained for a year. This way, if an issue with the production array causes data to be corrupted, I can restore a snapshot of the backup system to recover the files.

I use Rsync instead of ZFS replication as it allows me a little more flexibility in scheduling via cron. I am sure that replication could offer the same functionality if I knew more about it but I have used uni-directional Rsync for years (before ZFS was born) without any issues. Old dog, new tricks and all that.

The risk of total data loss is fairly low in this setup and would occur if:

• server theft
• at least 5 concurrent HDD failures in (3 in production pool, 2 in backup pool)
• flood, fire etc. (I have bigger fish to fry if that happens)
• malicious access (physical or remote)

To take this system to the next level, I would need a second server in a second physical location and all that. Not something I need for home use IMHO but that is how it is done where I work.

I think this strategy could be used by the original poster. I personally choose to have enough backup storage for all of my files but, for 500GB of data, you could make a small mirrored pool of spinning or SSD disks to suit the backup needs. Furthermore, given the limited size of the backup needed, you could make a second FreeNAS server at very little cost using older equipment and a couple small drives.

Finally, IMHO, any backup strategy that requires human interaction at regular intervals is bound to fail because humans are unreliable :) I say this from both personal and professional experience...

Cheers,