$Header: /CVSROOT/scylla-charybdis/README.backup,v 1.6 2003/11/02 22:35:39 tino Exp $
THIS IS MORE A CONCEPTUAL PAPER. BUT READ ON! AS THERE ARE NO MORE FUTURE RELEASES OF S&C THIS IS ALL YOU HAVE.
I don't need a restore capability, as I only back up files for web servers or dumps of databases, where the raw data is all you need. And there will be a successor to S&C, however this will take quite a time. For more about why S&C, read the file RATIONALE.
- Assumptions
dial-up.provider.at.eu is your location in Europe. this.computer.is.ko.us is the computer located in America. scylla-charybdis.ok.au is the backup server located in Australia. You don't want to fly to America nor Australia to get the backup done. All you have are some "remote hands" at both locations.
- Prerequisites
-
- You have an internet connection 1a) All computers have an Internet connection. 1b) this.computer.is.ko.us can reach scylla-charybdis.ok.au via Internet. 1c) There is no firewall blocking the port where scylla listens on.
- All computers work properly 2a) this.computer.is.ko.us is running and must be backed up. 2b) scylla-charybdis.ok.au has scylla installed as an inetd service. 2c) Scylla listens on the port 999 (change this accordingly). 2d) Only one instance of the backup script runs in parallel.
- You have all tools in place 3a) this.computer.is.ko.us has charybdis installed in /usr/local/bin/ 3b) The S&C distribution with all files lies at /home/src/sc/ 3c) scylla-charybdis.ok.au has scylla installed in /usr/local/sbin/ 3d) You have a scylla.pem to make SSL work 3e) You know how to use your brain
Process description for the lazy admin:
On scylla-charybdis.ok.au install scylla as an inetd service (as root):
/home/src/sc/example/make-ssl-certificate
ln -s /some/big/filesystem /scylla
chown nobody /scylla/.
chmod 1777 /scylla/.
echo "999 stream tcp nowait nobody /usr/sbin/tcpd /usr/local/sbin/scylla /scylla password /etc/scylla.pem" >> /etc/inetd.conf
killall -1 inetd
From dial-up.provider.at.eu open a shell to this.computer.is.ko.us
Run screen and within screen do a su to root.
Create a fresh directory not under /tmp and change into it. Run:
echo password |
/home/src/sc/example/backuploop.sh scylla-charybdis.ok.au:999 /
Be sure
- to replace password above by a real password.
- to use an absalute path (starting with /) at the second argument.
- that /bin/bash exists to run the script.
- to echo the correct Password. The script only waits 1 second to read the input from STDIN and then displays the Usage.
- netcat is installed. If you don't want netcat, UNCOMMENT the return 0 within the script.
- there is enough free room in /tmp .
- not to run two instances of backuploop.sh in parallel.
Note that backuploop allows you to give some parameters to it:
- You can give more than one path argument to the script
- You can give throttle information parameter in the form host:port@throttle:blocksize Throtteling means, roughly transfer only throttle KB/s with packets of a given blocksize (:blocksize can be left away)
- If you interrupt the script, you can (nearly) restart it where it
broke by specifying the working-timestamp. This looks like:
/tmp/scylla.backup/YYYYMMDDhhmmss `hostname -f`/tmp/scylla.backup/YYYYMMDDhhmmss YYYYMMDDhhmmss BE SURE ALL IS SYNCED ON scylla-charybdis.ok.au!
- If you want to speed things up (weird) by running several charybdis in parallel, try following helper (this is thought for LANs only): Be sure you have compiled /home/src/sc/example/read-nonblocking.c like this: cd /home/src/sc/example && make read-nonblocking echo password | /home/src/sc/example/backuploophelper.sh scylla-charybdis.ok.au:999 /tmp/scylla.backup/YYYYMMDDhhmmss.tmp X where YYYYMMDDhhmmss.tmp is the (existing) timestamp working directory and X is a number from 1 to "number of lines of your terminal less 1". THIS IS MESSY, I KNOW. But it works and does not break anything.
If you are happy how this works, you can call the script from cron as well.
Note that backuploop.sh is a script which was programmed such, that it even backs up a server in case the connection is extremely instable. For example scylla-charybdis.ok.au. can be on a dialup line reachable over dyndns and is disconnected and changes it's IP in more or less regular intervalls (like once a day). I even do it somewhat more insane in opening an SSH port tunnel from my backup server at home to the machine to back up.
That's it.
-Tino
$Log: README.backup,v $
Revision 1.6 2003/11/02 22:35:39 tino
CVS keywords added (I always forget this, sigh)