Open Source Backups Ignoring Open Standards
We have a lot of servers. Not a lot as might be defined by a former employer of mine, but certainly in terms of data, there's a lot. We back up a couple terabytes of data every week.
We've cobbled together this backup system over the years (and by "we" I mean the college and my predecessors). It works great, but it's definitely sub-optimal, requiring loving care and attention, which is something you really don't want from your backup system. You want your backup system to just work (and of course, you schedule random tests of your backup system to assure that this is the case).
But we started to need some features we couldn't cobble together any more, like NDMP support, and so we started looking for alternative solutions. Things more sophisticated than a collection of Perl and Bash scripts.
So I went to yesterday afternoon's "Open Source Backups" session with a lot of hope. I hoped I'd find something that would meet our organization's needs. They're not that complicated -- back up the various servers to our TonsO'Disk® servers. Archive that disk storage to tape for offsite vaulting. Then back up our (separate) NetApps to tape using NDMP, and offsite those tapes. That's not that hard, I wouldn't think. Day-to-day restores come from the local disk backups, and the tapes are strictly for DR purposes.
Except, of course, that there's no open source product that does these things. Bacula doesn't support backing up to disk(?!?). (Updated: OK, it does support backing up to disk, it's just not really in an intuitive place in the manual so I didn't see it.) Amanda does, but it does so in such a kludgy fashion that I wouldn't even consider it. Oh, and Amanda likes to think it knows better than I do when it should be doing full backups instead of incrementals, etc. Because why should I know better than it does when the system is under lower demand, or when the disk isn't being hammered by other tasks, I'm only the SysAdmin after all.
But worst of all these crimes against humanity is that neither of them supports NDMP. There's an open standard out there -- and one that's been around for quite some time -- and neither of the open source products supports it. If I want support for the open standard, I have to used closed-source software. Where's the logic of that?
Sure, the standard mantra of the Open Source community is "patches welcome", but what that really means is "unless you've got programmers on staff who can make this work for you, you're going to send money off to (Veritas,IBM,CA)[rand] and pay them to do it for you, and get something that's going to make management more warm and fuzzy than the Open Source solution was going to, anyway."
Because, sadly, that's what I have to do when I go back to work next week - find which closed-source company I plan to give money to.
Comments on this weblog
1 to 2 of 2
1 to 2 of 2
Return to weblogs.oreilly.com.