RE: [gentoo-user] Admin system documentation
I just can't get my head around why you would want or need to do a total rebuild. I myself can't imagine _wanting_ to do a rebuild, but needing to, yes. Flood, tornado, theft, etc . Users have work to do. Let's see, what's my SAMBA configuration? And those SQL databases looked like what? How quickly could you bring a system with many users back online _and_ functional? If you are in a position of being responsible for thoses entities, you had better have your ass covered. That's what I'm trying to do. If your only method of getting a production system back online is by doing a full rebuild, then you are royally screwed. Most linux systems (especially gentoo-based) are in a constant state of flux. Users get added and removed, emerge updates packages and config files, databases have rows added/removed, etc. Your primary disaster recovery tool is the system backup. That will be the only way to get your system back to the state it was in before the failure occurred. Planning to rebuild from scratch rather than restore from backup ensures that a) you're going to need a huge window of time to get the system back to a functional state, b) you're going to need another huge window of time to get all of your configuration back in line with the current setup, and c) your data from databases will be basically lost (recreating tables is just the start of trying to bring it back online). So forget the 'rebuild' idea, it won't work and will be prone to failure. Instead focus on the appropriate tool to build system backups. Test your backup procedures and restoration procedures earnestly to ensure that what you plan to do for disaster recovery really will work. If you can get to a static system state (i.e. you stop emerging packages/updates), you could get away with a full system backup performed once followed by incremental backups from /etc, /var, and /home. And don't forget to take copies of the backup'd media offsite to another location; that way if the building goes under (with your system in it) you'll be able to get another system online based off of the offsite media. Dave -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] Admin system documentation
John J. Foster wrote: Good afternoon, I had intended on starting my conversion from Suse 9.1 to Gentoo over the weekend, but the weather turned out to be way to nice to remain indoors. But in my planning stages I realized I have a bit of a longer learning curve than I initially anticipated. So, I'm going to remedy this by starting off with: Question #1 What do the professional (and amateur) admins among you consider to be essential system documentation in the event of a disaster. I am fairly well versed on the requirements of a M$ based system and network, but have only been dabbling in Linux for a couple years. Backups I know, but what is considered a fairly necessary _paper_ trail in the event that the unexpected happens, and a total rebuild is necessary in the shortest time possible. Much of it depends on the size of your network and what you're doing. I run mostly clusters of the same box. So I don't back up the OS because I can lose one without causing trouble and can easily add a new server in the case of a prolonged outage. I do tar.gz the data and relevant config files. You'll want to make a judgment call on how much infrastructure you think you need to dedicate to backups. However in the event that I need to add a new server I do have a few build docs and scripts. It's really the little stuff that is a pain in the ass. I keep track of it in a Wiki, which seems less annoying than they were a few years ago, but whatever works for you. Mine has stuff like this: New Server build /etc/resolv.conf Use the linked file. Never change the domain search order or old broken stuff that the wed devs idiots won't grep out will break. The rotate stuff distributes the lookups so we don't overload ns1. The timeouts cause lookups to timeout faster and move on the next server in case we lose a name server. /etc/rc.conf Change default editor to vim Name Servers only Add symlinks to deal with Gentoo bind/named nonsense ln -sf /etc/bind/named.conf /etc/named.conf Make the Redhat admins happy ln -sf /var/bind /var/named And then ask, why does Gentoo use named in /var/run/ instead bind? and so on. I would however seriously look at the Catalyst tool for building stage3 or what people are calling stage4 builds. I haven't gotten it working exactly right, but the idea would be to include most of the little nonsense in a stage3 and then use that as my base to generate stage4 build of particular kinds of server. So I'd be able to lay down a new name server, web server, mail server, db server etc in under an hour. Then import or rsync any data over and you're set. A few people I know are using Feather Linux to try to ghost partitions and then lay them down on new machines. Haven't heard how well that's working, but there was no reason why it would be an issue. The roundabout point on this is a Linux environment can get away from you if you let it. Use the same USE variables on everything and document the reason why you use those. And don't forget to update the docs and machines if that ever changes or you'll be fighting for three hours on a Sunday when something isn't quite kosher. kashani -- gentoo-user@gentoo.org mailing list
