Difference between revisions of "FreeBSD Network Backup Thing from Scrap"

From Run Your Own
Jump to: navigation, search
(Backup configuration)
(Borg Usage)
Line 100: Line 100:
 
'''Note:''' In theory it means that the RPi1 could be used as one big central repository for different machines to maximise the deduplicatings mechanism. However in practice, each machine needs to maintain a cache of the repos index, so every time a machine creates a new archive, the other machine will need to update their local copy, which will add some time to do to the backup process. Most importantly, because the borg process locks the repos when being used by a machine, it means that no concurrent backup can be made on the same repos. For these reasons each machine were given their own directory, for their own repos, in the client setup above.
 
'''Note:''' In theory it means that the RPi1 could be used as one big central repository for different machines to maximise the deduplicatings mechanism. However in practice, each machine needs to maintain a cache of the repos index, so every time a machine creates a new archive, the other machine will need to update their local copy, which will add some time to do to the backup process. Most importantly, because the borg process locks the repos when being used by a machine, it means that no concurrent backup can be made on the same repos. For these reasons each machine were given their own directory, for their own repos, in the client setup above.
  
 +
=== Examples ===
 +
==== Repository ===
 
* Initialize backup repository (this needs to be done '''only once''')
 
* Initialize backup repository (this needs to be done '''only once''')
 
  borg init --encryption=repokey-blake2 freebsd@192.168.1.XXX:/sendhelp/${HOST}/
 
  borg init --encryption=repokey-blake2 freebsd@192.168.1.XXX:/sendhelp/${HOST}/
  
 +
==== Archives ====
 
Examples of archive creation:
 
Examples of archive creation:
  

Revision as of 22:24, 12 July 2018

This document shows how to setup a dedicated backup machine for LAN use, made from bit and pieces most people don't care about these days: a Raspberry Pi 1 (it seemed such a great idea at the time... ahem...), a 32MB SD Card, and a 2GB USB key. For the software, we will use FreeBSD and BorgBackup.

Prepare the OS

Why waste a 8GB or 16GB fancy SD Card when you probably have some old SD and USB sticks lying around? Also reading performances on the RPi1 is much better on USB than with the crappy SD reader. Finally, FreeBSD on the RPi1 works great, and very light on resources which is a plus given the limitation of the RPi CPU and RAM.

What needs to be done in a nutshell is to have the mandatory MS-DOS /boot partition on the 32MB SD card and have the rest of the filesystem on the USB stick.

  • Get the FreeBSD RPi1 image (update to latest RPI-B image)
wget https://download.freebsd.org/ftp/releases/arm/armv6/ISO-IMAGES/11.2/FreeBSD-11.2-RELEASE-arm-armv6-RPI-B.img.xz
  • Plug USB stick and copy the FreeBSD image to it (adjust /dev/sdX)
xzcat FreeBSD-11.2-RELEASE-arm-armv6-RPI-B.img.xz | sudo dd bs=1M of=/dev/sdX status=progress
  • Plug the 32MB sdcard and create a W95 FAT32 partition with fdisk or something (Hex code b) then format it:
sudo mkfs.vfat /dev/sdY1
  • Mount both the USB key's first partition and the SD card partition and copy the boot content from the USB key to the SD card:
sudo mount /dev/sdX1 /media/usb-key
sudo mount /dev/sdY1 /media/sd-card
sudo cp /media/usb-key/* /media/sd-card/
umount /dev/sdX1
umount /dev/sdY1

Boot FreeBSD and basic system configuration

Booting and ssh'ing

  • Insert the SD card and USB key on the RPi1
  • Optionally: connect the composite out to something for easier troubleshooting
  • Power the RPi1, the booting process will automagically find the USB key and the FreeBSD system partition (actually this is a little known feature and it's super nice).
  • Optionally: if you can't figure out the IP of the RPi1, you can sweep your LAN with:
nmap -sP 192.168.1.0/24
  • ssh in the RPi1 (default user is freebsd:freebsd):
ssh freebsd@192.168.1.XXX

Basic system config

Note: The first time FreeBSD starts, it will expand the filesystem to the full available space of the USB key.

  • Change default password for freebsd user:
passwd
  • Become root with su (password is root)
  • Change default password for root user:
passwd
  • update the FreeBSD repos catalogue (will take some time to process):
pkg update
  • Install $YOUR_FAV_TEXT_EDITOR
pkg install vim-tiny
echo "set nocompatible" > /root/.vimrc
  • Enable NTP for setting up time at every startup (the RPi1 has no RTC to keep actual time), edit /etc/rc.conf and add the following:
ntpd_enable="YES"
ntpd_sync_on_start="YES"
  • While you're here, feel free to rename the RPi:
hostname="thevault"
  • Configure your time zone:
tzsetup
  • reboot to celebrate your escape from Groundhog Day
reboot

Prepare the USB hard-drive for backup

Note: I'm using a 4TB USB disk here, do whatever you want.

  • ssh back in the RPi1, and become root
ssh freebsd@192.168.1.XXX
su
  • Plug the USB drive
  • If all goes well, it will show up as da1 (da0 is the USB key)
dmesg | grep da1
  • Create a partition
gpart create -s GPT da1
gpart add -t freebsd-ufs -a 1M da1
gpart show da1
  • Format the new partition
 newfs -U /dev/da1p1
  • Edit /etc/fstab and add an entry for the disk's mount point:
/dev/da1p1      /sendhelp       ufs     rw      2       2
  • Create and mount the mount point
mkdir /sendhelp
mount /sendhelp
chown freebsd:freebsd /sendhelp

BorgBackup configuration

On the server side

  • ssh in the RPi1, and become root
ssh freebsd@192.168.1.XXX
su
  • Install BorgBackup
pkg install py36-borgbackup

On each client

ssh-copy-id freebsd@192.168.1.XXX
  • Make a directory for your machine's backup on the RPi1
ssh freebsd@192.168.1.99 mkdir /sendhelp/${HOST}

Borg Usage

Introduction

Borg is deduplicating backup program, which means that each file is split into chunks and only chunks that have never been seen before are added to the repository. Borg uses repositories to manage such information, and inside a repository several archives can co-exist, all sharing the same deduplicating database. The nice thing about borg is that it does not assume a particular workflow, it's really up to the user to make archives that are following an incremental model, or snapshots of specific disks, one time folder copies, and of course mix and match of all these.

Note: In theory it means that the RPi1 could be used as one big central repository for different machines to maximise the deduplicatings mechanism. However in practice, each machine needs to maintain a cache of the repos index, so every time a machine creates a new archive, the other machine will need to update their local copy, which will add some time to do to the backup process. Most importantly, because the borg process locks the repos when being used by a machine, it means that no concurrent backup can be made on the same repos. For these reasons each machine were given their own directory, for their own repos, in the client setup above.

Examples

= Repository

  • Initialize backup repository (this needs to be done only once)
borg init --encryption=repokey-blake2 freebsd@192.168.1.XXX:/sendhelp/${HOST}/

Archives

Examples of archive creation:

  • Create an archive of /something/to/backup with a timestamp as automatic name
borg create --compression none freebsd@192.168.1.XXX:/sendhelp/${HOST}::{now} /something/to/backup
  • Create a new archive of the same /something/to/backup with default compression
borg create freebsd@192.168.1.XXX:/sendhelp/${HOST}::{now} /something/to/backup
  • Create a new archive of the same /something/to/backup and /something/else with progress
borg create --progress --compression none freebsd@192.168.1.XXX:/sendhelp/${HOST}::{now} /something/to/backup /something/else

Note: we use --compression none to