A timely discovery

I have been interested in automated backups of computer data since the mid 90’s, when I had a very well-timed hard disk failure. By pure chance, I had been working on a script that would copy my “important files” from my hard disk to a 100MB “Zip Drive”. I finished my script after testing it several times, and then I went to bed. The next morning, I woke up to find that my hard disk had crashed. Fortunately, I had a very recent backup!

I have often marveled at how easy it would be to lose invaluable files in a single mishap… countless memories, photos, financial records and project work.  Backups are important.

BackupPC

When I worked at “the oven place” (TMIO), I was tasked with evaluating backup schemes for their factory and office PC’s. So I looked at several open source packages, with emphasis on being server-centric and automatic. That is, the backup server would decide when to make the backups, and the employees would never have to remember to do anything special. Any process that relies on a human to remember to kick it off is destined to be run once-a-year.

We ended up choosing “BackupPC“, which runs on a modest server with a large storage disk. It would wake up every so often and run through its list of clients and pick one to back up.

For several years, I ran BackupPC at home, too. At first, I ran it on a discarded PC. But later, I migrated to low-power fanless embedded boards.

Flashback

In 2013, I decided that BackupPC was taking too long to make backups. I would bring my laptop home from work and turn it on, and BackupPC would notice it and start backing it up. But the backups were taking so long that they would still be running when I was ready to leave for work the next morning!  I ran a few tests with rsync to see if the problem was with BackupPC or the file compression or their crazy idea of how “incremental backups” should work.  So I wrote what started out to be a speed test, and then a wrapper around “rsback”, and finally a very minimal python script that I named “Flashback“.  “Flash” because it’s fast.  My laptop backup, which was taking all night using BackupPC, usually completes in a half hour or less.

You can find Flashback on github.

The Pogo Plug v4

pogoplugv4This week, I tried out a new hardware gadget called a Pogo Plug.  It is a very close cousin to the SheevaPlug, an embedded Linux board which I had been BackupPC and Flashback on.  What caught my attention about the Pogo Plug v4 was:

  • It has two USB3 ports.
  • It has gigabit ethernet.
  • It was on sale for just $20.

The only bad part is that it only has 128MB of RAM… that’s only a quarter of what the SheevaPlug has.  But I am not really using the memory for anything.  I am just running rsync.

I did not spend any time using the stock firmware.  Instead, I immediate enabled SSH and then followed these instructions for installing Arch Linux on a USB stick, which I plugged into the top plug (the bootable USB2 one).  I plugged the 1-terabyte USB2 hard disk into the back of the Pogo Plug.

Then I installed Flashback and I modified the monitor script to take advantage of the three-color LED on the front (green for sleeping, yellow for backing up, red for error).

It’s been running for about a week now, and I think it has proven itself worthy.

I’d like to try it with a USB3 hard disk, and see if it’s any faster.