On Fri, 2005-04-01 at 13:09 -0500, Gene Heskett wrote: > On Friday 01 April 2005 03:19, Erik P. Olsen wrote: > [...] > >Hm. I would like to challenge this statement. It may be true if you > > have a large network of systems to back up but if you only have one > > system I doubt if it's true. Before my switch to Linux I backed up > > my system (OS/2) on a weekly schedule with one full back-up and > > four incremental back-ups to one tape only. > > If useing tapes, and they are big enough, set the holding disk > reserved value to some low percentage like 20%, and only put in a > tape once a week, with the autoflush option set in your amanda.conf. > That will put the whole weeks worth of backups on one tape. It does > have the disadvantage of leaving that weeks stuff subject to a disk > failure though. But thats something I've not had in about 2 years, > no failures out of about 7 drives here when they are all spinning. > But I'll lose one yet today just because I mentioned it, Murphy is > listening. :(
Last year I had head crash with my two 60 GB disks when they where still under warranty. I had them replaced but couldn't wait the 3 weeks it took, so I had to buy new ones. Last time it happened (there were 6 months in between) the tape streamer also died and took the tape with it. I'm sure Murphy has a law for that. As a consequence I could not restore the system even with a new and well functioning tape station in place. Statistically this problem should never occur, but it did. > > > It worked extremely > > well, if I crashed my system - which I did very often - it took me > > about half an hour to recover either using a stand-alone recover > > program or my maintenance OS/2 if it was alive and I kept an > > archive of up to 8 weeks of back-ups. > > I was always told that OS2 was stable. And I stay quite bleeding edge > in terms of the kernel I run on this FC2 system, which is also > backing up my RH7.3 firewall box. Currently running 2.6.12-rc1, the > smoothest running, snappiest kernel yet in the 2.6 series. I can't > recall the last time I actually crashed a running system. Several > months ago in any event. OS/2 is stable though not as stable as FC3. The frequent crashes were mainly because I did a lot of testing with new system software. > > > Now with Linux and Amanda I > > use 9 tapes mainly because Amanda won't add today's back-up to > > yesterday's tape. > > Thats a security risk amanda won't take. When amanda is done, and has > released the drive, there is nothing to prevent someone from removing > the tape, and either reinserting it, in which case the tape is > rewound and will be totally overwritten, or even the wrong tape might > be reloaded. Either way, amanda has no ironclad assurance that the > tape will be sitting in the same position it was left in, ready to > append new files to it. Yes, most drives today can do an 'mt > -d/dev/nst0 seof' and hit within a quarter of an inch of it. But > some drives cannot, and that locks amanda out of useing that feature > for all users. At some point, the last legacy drive that cannot do > that might die, but we have no idea when that might be... I suppose the back-up software placed a sort of end-of-tape mark after each back-up and just searched for that tapemark when a new back-up was about to be run. I've used it for 5 years and it never failed. It would even ask for another tape to continue the back-up if the first tape ran full. > > Maybe that should be the subject of a questionaire at some point? > > > I could probably do with less tapes but I feel > > more confident with a large tape pool. > > [...] > > I answered this in a previous post. > -- Regards, Erik P. Olsen