Re: Backup server disaster recovery
On Fri, Jul 21, 2006 at 06:16:38AM -0700, Joe Donner (sent by Nabble.com) wrote:
>
> Dear all,
>
> say your amanda backup server itself dies, and you need to
> reinstall/recreate it from scratch.
>
> You want the new backup server to have available the information needed to
> find and restore data from tapes, i.e. the info you get when running:
> amadmin config find server_name disk
> or like when you restore something from a specific date, when the server
> tells you which tapes you need to do the restore (while using amrecoger).
>
> What do you need to back up on the current backup server to enable you to
> get the new server to this state?
>
Top of head list
(some may overlap,
some may not be needed for basic recovery):
amanda_user home directory
amanda config directory
config files not in config directory
(ex. I breakout dumptypes, tapetypes, and others
into a common directory for all configs)
logfile directory
index directory
curinfo directory
gnutar-lists directory
/etc/amandates
/etc/dumpdates
.amandahosts
{x}inetd config
crontab entries
/???/bin/
/???/sbin/
/???/libexec/
/???/lib/
/???/lib/
?? system tape and changer config file ??
(ex. on Solaris I've modified st.conf and
sgen.conf, on Fedora I've modified stinit.def)
exclude or include list files
?? holding disk in case what you want is not flushed ??
You might bundle these with tar and transfer to a different
host for safe-keeping after each amdump. An alternative
advocated by at least one is to append the bundle to the
end of the tape just used by amdump. So it would be a
tape file after all the amanda tape files. Easy to locate,
advance to end and backup one file. This assumes sufficient
space on tape (testable).
--
Jon H. LaBadie [EMAIL PROTECTED]
JG Computing
4455 Province Line Road(609) 252-0159
Princeton, NJ 08540-4322 (609) 683-7220 (fax)
Re: Backup server disaster recovery
On Fri, July 21, 2006 9:16 am, Joe Donner (sent by Nabble.com) wrote: > > Dear all, > > say your amanda backup server itself dies, and you need to > reinstall/recreate it from scratch. > > You want the new backup server to have available the information needed to > find and restore data from tapes, i.e. the info you get when running: > amadmin config find server_name disk > or like when you restore something from a specific date, when the server > tells you which tapes you need to do the restore (while using amrecoger). > > What do you need to back up on the current backup server to enable you to > get the new server to this state? > > Will appreciate your insight. > > Joe > -- > View this message in context: > http://www.nabble.com/Backup-server-disaster-recovery-tf1980202.html#a5433481 > Sent from the Amanda - Users forum at Nabble.com. > Hello: Here is what we have done at our site. We re-mastered DSL (Damn Small Linux) with all the tools we need (and some we won't ever need). The bootable CD includes: ssh, netcat, etc., SCSI modules for our card, raid stuff, and most important, Amanda client. As a test, we trashed the drives on the backup server rendering it un-bootable and after the restoration a few minutes later the server was running again. What I did was, boot from the CD, fdisk the drives, create the file systems, enable mirroring, mount the partitions, and restore the last level 0 to the drive. I didn't have to mess with any incrementals because part of my test included a level 0 right before destroying the server. The only other thing I did was to copy /dev from the CD to the drive and mount /proc in order to install lilo on the hard drives. In short, if you could invest the time in creating a "Live CD" with amanda client now, it will pay good dividends later. You would need nothing more than the CD and a cup of coffee. Later, / Ronald Vincent Vazquez Senior Unix Systems Administrator Senior Network Manager Christ Tabernacle Church Ministries http://www.ctcministries.org (301) 540-9394 Home (240) 401-9192 Cell For web hosting solutions, please visit: http://www.spherenix.com/
Re: Backup server disaster recovery
On Fri, Jul 21, 2006 at 06:16:38AM -0700, Joe Donner (sent by Nabble.com) wrote: > > Dear all, > > say your amanda backup server itself dies, and you need to > reinstall/recreate it from scratch. http://www.charlescurley.com/Linux-Complete-Backup-and-Recovery-HOWTO.html This does not cover Amanda; see Jon's reply for that. -- Charles Curley /"\ASCII Ribbon Campaign Looking for fine software \ /Respect for open standards and/or writing? X No HTML/RTF in email http://www.charlescurley.com/ \No M$ Word docs in email Key fingerprint = CE5C 6645 A45A 64E4 94C0 809C FFF6 4C48 4ECD DFDB pgpP7txUFbjt4.pgp Description: PGP signature
Re: Backup server disaster recovery
On Fri, July 21, 2006 3:57 pm, Jon LaBadie wrote: > On Fri, Jul 21, 2006 at 02:44:30PM -0400, Ronald Vincent Vazquez wrote: >> >> Hello: >> >> Here is what we have done at our site. We re-mastered DSL (Damn Small >> Linux) with all the tools we need (and some we won't ever need). The >> bootable CD includes: ssh, netcat, etc., SCSI modules for our card, raid >> stuff, and most important, Amanda client. >> >> As a test, we trashed the drives on the backup server rendering it >> un-bootable and after the restoration a few minutes later the server was >> running again. What I did was, boot from the CD, fdisk the drives, >> create >> the file systems, enable mirroring, mount the partitions, and restore >> the >> last level 0 to the drive. I didn't have to mess with any incrementals >> because part of my test included a level 0 right before destroying the >> server. The only other thing I did was to copy /dev from the CD to the >> drive and mount /proc in order to install lilo on the hard drives. >> >> In short, if you could invest the time in creating a "Live CD" with >> amanda >> client now, it will pay good dividends later. You would need nothing >> more >> than the CD and a cup of coffee. >> > > Great sounding system. Couple of questions if I might: > > 1. The resulting live CD, is it limited to the restoration of a single >host, presumably the amanda server. And within that limit, if it >exists, is the live CD in anyway tied to specific amanda configs? > > 2. Over the years/months, what sorts of things would necessitate a >rebuild of the CD? Change of OS release? Update of any specific >packages on the host? hardware changes? amanda config changes? >amanda updates? > > 3. How scriptable might creating such a CD be? I'm thinking of two >scenarios. An "automated" build would be useful if regular remakes >of the CD were needed (question 2 above). And it would be a nice >distributable script. Give instructions for how to get DSL and >how to put together any of the extra pieces (netcat, etc.) into a >form/location the script can use. Then setup a config file that >says where the pieces are and anyother params needed. > >After they've done those steps, even neophytes of building a live CD, >might be able to run the script and create a .iso to burn as their >rescue/recovery disk. > > Thanks, > jon > -- > Jon H. LaBadie [EMAIL PROTECTED] > JG Computing > 4455 Province Line Road(609) 252-0159 > Princeton, NJ 08540-4322 (609) 683-7220 (fax) > Hello Jon: 1. Actually, the resulting CD is universal, we could restore any of our x86 machines here. Our CD displays a prompt where you can choose form a series of options: 1. Restore a cluster node, 2. Restore the backup server itself (will drive the changer), 3. Restore any of our servers, 4. Obtain a "bash" shell. All you do to restore any of the other servers is configure your ethernet interface and ssh to the backup server or use netcat. 2. The only reason I see which will force us to re-master the CD soon will be to upgrade amanda itself from vers 2.4.X to 2.5.X. Other than that all I can think is that you may want your "universal" CD to perform a new task. In that case, another 20 cents and 5 minutes of your time... ;-) 3. The process of creating the disk can be turned into a script very easily. For the software we add to the CD, we are compiling it on another machine and then moving the binaries and libaries to the chrooted environment manually. Let me see if I have enough time this weekend to "cook" a document in order to share our efforts. RV / Ronald Vincent Vazquez Senior Unix Systems Administrator Senior Network Manager Christ Tabernacle Church Ministries http://www.ctcministries.org (301) 540-9394 Home (240) 401-9192 Cell For web hosting solutions, please visit: http://www.spherenix.com/
Re: Backup server disaster recovery
On Fri, Jul 21, 2006 at 04:32:04PM -0400, Ronald Vincent Vazquez wrote: > > On Fri, July 21, 2006 3:57 pm, Jon LaBadie wrote: > > On Fri, Jul 21, 2006 at 02:44:30PM -0400, Ronald Vincent Vazquez wrote: > >> > >> Hello: > >> > >> Here is what we have done at our site. We re-mastered DSL (Damn Small > >> Linux) with all the tools we need (and some we won't ever need). The > >> bootable CD includes: ssh, netcat, etc., SCSI modules for our card, raid > >> stuff, and most important, Amanda client. > >> > >> As a test, we trashed the drives on the backup server rendering it > >> un-bootable and after the restoration a few minutes later the server was > >> running again. What I did was, boot from the CD, fdisk the drives, > >> create > >> the file systems, enable mirroring, mount the partitions, and restore > >> the > >> last level 0 to the drive. I didn't have to mess with any incrementals > >> because part of my test included a level 0 right before destroying the > >> server. The only other thing I did was to copy /dev from the CD to the > >> drive and mount /proc in order to install lilo on the hard drives. > >> > >> In short, if you could invest the time in creating a "Live CD" with > >> amanda > >> client now, it will pay good dividends later. You would need nothing > >> more > >> than the CD and a cup of coffee. > >> > > > > Great sounding system. Couple of questions if I might: > > > > 1. The resulting live CD, is it limited to the restoration of a single > >host, presumably the amanda server. And within that limit, if it > >exists, is the live CD in anyway tied to specific amanda configs? > > > > 2. Over the years/months, what sorts of things would necessitate a > >rebuild of the CD? Change of OS release? Update of any specific > >packages on the host? hardware changes? amanda config changes? > >amanda updates? > > > > 3. How scriptable might creating such a CD be? I'm thinking of two > >scenarios. An "automated" build would be useful if regular remakes > >of the CD were needed (question 2 above). And it would be a nice > >distributable script. Give instructions for how to get DSL and > >how to put together any of the extra pieces (netcat, etc.) into a > >form/location the script can use. Then setup a config file that > >says where the pieces are and anyother params needed. > > > >After they've done those steps, even neophytes of building a live CD, > >might be able to run the script and create a .iso to burn as their > >rescue/recovery disk. > > > > Thanks, > > jon > > -- > > Jon H. LaBadie [EMAIL PROTECTED] > > JG Computing > > 4455 Province Line Road(609) 252-0159 > > Princeton, NJ 08540-4322 (609) 683-7220 (fax) > > > Hello Jon: > > 1. Actually, the resulting CD is universal, we could restore any of our > x86 machines here. Our CD displays a prompt where you can choose form a > series of options: 1. Restore a cluster node, 2. Restore the backup server > itself (will drive the changer), 3. Restore any of our servers, 4. Obtain > a "bash" shell. All you do to restore any of the other servers is > configure your ethernet interface and ssh to the backup server or use > netcat. > > 2. The only reason I see which will force us to re-master the CD soon will > be to upgrade amanda itself from vers 2.4.X to 2.5.X. Other than that all > I can think is that you may want your "universal" CD to perform a new > task. In that case, another 20 cents and 5 minutes of your time... ;-) > > 3. The process of creating the disk can be turned into a script very > easily. For the software we add to the CD, we are compiling it on another > machine and then moving the binaries and libaries to the chrooted > environment manually. > > Let me see if I have enough time this weekend to "cook" a document in > order to share our efforts. > > That would be great. I have a spare box totally unused with a small disk. I'll try to use it to trash and recreate from a live CD based on your docs. -- Jon H. LaBadie [EMAIL PROTECTED] JG Computing 4455 Province Line Road(609) 252-0159 Princeton, NJ 08540-4322 (609) 683-7220 (fax)
Re: Backup server disaster recovery
> Dear all, > > say your amanda backup server itself dies, and you need to > reinstall/recreate it from scratch. > > You want the new backup server to have available the information needed to > find and restore data from tapes, i.e. the info you get when running: > amadmin config find server_name disk > or like when you restore something from a specific date, when the server > tells you which tapes you need to do the restore (while using amrecoger). > > What do you need to back up on the current backup server to enable you to > get the new server to this state? I understand that you are interested in the possibility to reuse the back-up on tape, not by restoring the amanda server. I assume that you'd know how to rebuild an Amanda server from scratch, but then you'd need all the previous tape to be usable (that's how I set-up the strategy we are currently using). What we keep up is: - Amanda index database (from amanda user home directory/Daily-Setxxx) - Amanda config files (from usuall /usr/local/etc/amanda?) I do that by: - rsync'ing the above mentionned directories to another machine - emailing to Amanda operators the result of amadmin export See: https://wwws.cs.ait.ac.th/amanda/operator.shtml#database The command amdatabase is a home coocked script. Best regards, Olivier
