Post: rsync howto rsync network backup tips
April 29th, 2005 by jamFri April 29, 2005
Post: rsync howto rsync network backup tips
Finally got back to this. Firstly, I have to say I still think tapes have a place in backups, but more as the weekly/monthly archive than the primary method of backing stuff up.
Naturally, I can’t find a good howto that’s specific to the kind of rsync backups I use (server, no encryption, no ssh)
But this one sets up a server very simply:
http://transamrit.net/docs/rsync/
Additionally:
* Yes, the rsync server wont function without the correct permissions on the secrets file
* To backup TO an rsync server, obviously you change read only = no
* To synchronise a volume to a server I use rsync -av –delete /path/ |username@server::modulename(usually the name of the server you’re backing up)
* To synchronise from a server (servers set up on machines to be backed up, this helps to centrally manage a few of these):
  rsync -av –delete username@server::modulename /destination_path/
* To script this activity, the following options come in handy:
–password-file=file (with password to use in it, must have similar brutal permissions if I recall)
–bwlimit=X (kilobytes per second) handy if you’re rynchronising a couple of gig over an xDSL link during the day and don’t have QOS.
–stats produces some useful output for your logs
–exclude-from=FILE can be handy if you’re backing up / and don’t want to include the CEO’s mp3 collection or the backups folder
–max-delete=NUM the one time this would have come in handy, I really wished I’d used it. Basically if you have an insane problem (disk not mounted), this helps prevent rsync helpfully synchronise a blank volume to your backup machine
Performance issues:
* Using rsync, roughly 500GB of user files (400ish users) seems to move about 5GB of data a day. That’s about 7 hours over a 2Meg xDSL link. (slightly less as rsync also compresses data in transit). These are from memory, I don’t post from work so don’t have the exact data, or the pretty graph I made.
* Our mail server backups (Groupwise) are also a couple of gig per day.
* Doesn’t work well on MS-SQL backup files, sends entire file every time (I think MS do something really creative with the backup file that makes it impossible to track what’s changed
* Works very well on VMware virtual disk files (which must be either unmounted or a snapshot), including virtual disks that CONTAIN live MS-SQL databases
* 400GB of files seems to consume about 1Gig of RAM while rsyncing (again, from memory, YMMV). Going heavily into swap is unlikely to effect performance if you’re doing it over a slow network link anyway.
Storage requirements:
* The last 1TB array I built using commodity hardware cost about $3k. (and took me about eight hours of fighting with RAID controller drivers, but that is another story
* I have a quote for a Promise 15drive iSCSI unit with 3.2TB of SATA drives (WD Raid Edition, 1M hours MTBF, 7200rpm SATA) for about $12k.
Cool stuff:
* Any linux machine whose data has been synchronised to another linux server can have its *functionality* brought up on the backup server, in the event of a disaster. I have previously managed to merge two file servers and a web server using this method. As such, your local backup box can become a handy Business Continuity server. Any Virtual machine… even easier. If you’re doing the same thing to an offsite machine, well, there’s your DR done.
* It’s really quite handy to have Yesterday’s Data on disk all the time. Cuts down the number of restores we have to do.
Regards,
James

Posted in Nerdy Shit, Work | No Comments »