* Adar Dembo ([EMAIL PROTECTED]) [021218 03:02]:
> I'm trying to create a backup script that will, when run, connect to 
> another computer, and rsync all of its partitions into the local computer.
> 
> In order to be able to rsync properly and copy all the files, the user 
> logging in must be root. However, this poses a problem, as 
> PermitRootLogin was "no" in sshd_config. So here is how I went about 
> trying to solve the problem.
> 
> First, some names. The source computer is called "dh3". The target 
> computer is "dh2".
> 
> On dh2, I ran:
> ssh-keygen -t rsa1
> ssh-keygen -t rsa
> ssh-keygen -t dsa

Just use one key, not 3.  3 keys is 3 times as many ways to get root
on that other machine; your backup script only needs one.  Let's just
use "-t rsa" to make an ssh protocol version 2 rsa key.

> Note the use of PermitRootLogin forced-commands-only which should allow 
> me to ssh in as root, using my keys, as long as I run a command afterwards.
> 
> The actual command being run on dh2 (as root) is something to the effect of:
> 
> ssh dh3.doggus.com rsync .....

No, that's not quite how it works.  forced-commands-only will allow
login with that key and will automatically execute the command listed in
the authorized_keys file, ignoring anything specified on the ssh command
line[*].

Your authorized_keys should look something like this:

from="other.host",command="rsync --server 
something",no-port-forwarding,no-X11-forwarding,no-agent-forwarding ssh-rsa 
1231241234123(key data) backup-key(comment)

Just use one key, not 3, and set up some options to make it more secure,
as in the above example.  Invoke rsync from the client as rsync, which
will try to use rsh (which is really ssh on your system, right? Or give
rsync -e /usr/bin/ssh).  That is to say you shouldn't be invoking "ssh
other host rsync something"

To figure out the rsync command to place in your authorized_keys file,
use ssh -v to see the command rsync tries to run on the server side
(it'll  be rsync --server something) and stick that in the command=""
option in your authorized_keys.

[*] The original command line is placed in SSH_ORIGINAL_COMMAND in the
environment of the executed command.  This can be used e.g. in a wrapper
script that needs variable arguments: it can parse $SSH_ORIGINAL_COMMAND
and use the args specified on the ssh command line.  For a backup
script, you shouldn't need variable arguments at all: it'll be run the
same all the time (assuming you're always backing up the same
directories).

good times,
Vineet
-- 
http://www.doorstop.net/
-- 
One nation, indivisible, with equality, liberty, and justice for all.

Attachment: msg19916/pgp00000.pgp
Description: PGP signature

Reply via email to