Why not advertise ssh/scp save possibility with sshfs?

Messages about how to use partimage across a network (with the partimaged server, samba, nfs, ...)

Moderator: feffer

Post Reply
Posts: 1
Joined: Thu Sep 04, 2008 10:20 pm

Why not advertise ssh/scp save possibility with sshfs?

Post by LightKnight » Thu Sep 04, 2008 10:24 pm

I just booted on the Ubuntu live CD, installed partimage and sshfs, mounted a remote linux partition with:

$ sudo su -
# mkdir /tmp/server
# sshfs [email protected]:/home/username /tmp/server
# cd /tmp/server
# partimage

And then just saved files as if they were local. No need for complicated nfs setup or getting a client/server thing to work.

sshfs rocks. Also for partimage. Why not advertise it in the docs?

Posts: 126
Joined: Tue Mar 17, 2009 4:30 am

Post by feffer » Wed Apr 29, 2009 12:10 am

I scripted partimage to run in cron using the client-server model in the partimage docs. This works, but takes a long time on my rather slow NAS server. As LightKnight suggested, using ssh/scp to run partimage from a server works well. Here is my server-side script:

Code: Select all

# Script to image client machine partition locally, then mv to slug
# Makes image of Ubuntu laptop partition (/dev/sda5) from mint_lap or kubby_lap
# Saves locally, then moves it to slug server

# Added so cron will directly run script
set > /tmp/cron.environment

# The target partition must be unmounted or partimage will fail
# "laptop-bak" is set in /home/.ssh/config to authorize ssh with public/private keys
# There are many online tutorials that explain how to do this
ssh laptop-bak umount /dev/sda5 > /dev/null 2>&1;

Today="`date +%m-%d-%Y-%M`"

# This command saves the partition locally to the client machine
# It ends with "> /dev/null 2>&1" to prevent partimage progress feedback
# The "--fully-batch" flag needs the dummy follower "x=y" 
ssh laptop-bak partimage save --compress=1 --nodesc --finish=3 --fully-batch x=y /dev/sda5 /home/pep/data/tmp/ubun_images/ubun_lap_$Today.gz > /dev/null 2>&1 &&

# The following copies the finished image from the client machine to the NAS server
# When copied, it deletes the image from the client
# This sequence is faster than using partimaged on the server and saving the image there -- at least on my LAN -- ymmv
cd /home/bak/laptop/ubun_lap/images;
scp -q laptop-bak:/home/pep/data/tmp/ubun_images/*gz* . &&
ssh laptop-bak rm /home/pep/data/tmp/ubun_images/*gz*

# I want a maximum of 3 images of any given partition
# The following deletes the oldest if more than 3
while [ `find -iname '*.gz*' -not -path '*saved*' | wc -l` -ge 4 ];
        do ls -t -1 ./*.gz* | tail --lines 1 | xargs rm;

# Finally, the image is made executable -- ready to restore
chmod +x *`date +%m-%d-%Y`*gz* 
This script finished almost twice as fast as the partimaged one on my LAN. One advantage of the above script is that I can trigger it from cron on my client laptop and direct it to image an unbooted partition. Hope some find this useful.


Post Reply