Re: Disk mirroring - rsync - backup methodology
hi oki
yes..yes...
your bnackup mechanism and backup implementation will
protect you against only certain failures... a single backup
methodology will NOT protect you against various failure modes
and yes... if you create a bad file or corrupt a file...
and you use raid1 to mirror your data...
than you have corrupted your data disk AND the backup !!!
-
- i do NOT like raid1 for data backup for this reaason
- raid1 by itself should NEVER(??) be used for data backup
-
- you lose your data on both drives
-
backup methodology ... fun stuff....
have fun
alvin
http://www.linux-backup.net
-- how much fault tolerance do you wanna implement ???
- what do you wanna protect against???
- to protect against power cable problems, and ethernet cable
problems...
- keep the sticky fingers away from the cables
- to protect against fan failures... ( next most commmon failures )
- have more than one fan...
-
- dual power supply may or may not solve your problems
- how often has things failed on you...
and what was the reason for the failure...
-
- its very seldom the disks itslef
-
- change your partition on the disks...
- monitor/admin your server more often... ???
-
- to protect against any disk failures
( i always assume throw the diska away kind of failures )
- use raid mirroring or raid5
- to protect against boot failures
- make a boot floppy
- make a bootable cdrom
- make a raid1 or raid5 root/raid system
- to protect against network failures...
-
- have more than one gateway between the server and its backup
-
have a local disk for backup... although if the power supply
goes bonkers...you lose the data drive and its backup disks
but you do have time to copy off the data to another backup system
- most people's backup requirements are satisfied with the above
methodologies ...
- to protect against data being written and updated while your
backup is still updating/working and your disk dies...
-
- geez...that kind of (realtime) backup is expensive
-
make sure that the program writting data is writting
to multiple servers to guarantee zero data loss
even if a disk/server crashes in the middle of writing
now you start needing multiple raid systems to backup
other raid systems
now we are at...
-
- the 500Gb Raid5 in 1U ....
-
On Tue, 15 May 2001, Oki DZ wrote:
> Alvin Oga wrote:
> > each time youupdate lilo.conf or add new kernels and modules,
> > you'd have to remember to rerun lilo on the backup disk
> > ( assuming that is to be a hard disk bootable replacement
> > - order of magnitude easier ot do raid1 in this case
>
> I don't think that I'd copy the kernel and the bootloader config files.
> I mean, it is sufficient to copy them once and have the disk bootable,
> and do the rest (data, applications) using rsync.
>
> > - if you just want ot backup your main hd to a mirror/backup disks...
> > there is no reason to worry about boot issues as it will not be
> > able to boot ... ( /etc/fstab, /etc/lilo.conf, etc would be wrong,
>
> Well, I think it depends how bad the first disk failure would be. If it
> were a total failure, then the SCSI number (the /dev/sd<x>) would be
> changed; /etc/fstab, etc. (on the second disk) would be all right. But
> if it were a data failure, then the setup wouldn't work.
>
> I think it would be easy to make the second disk bootable, especially if
> it's done from a floppy. The difficult thing is how to predict what kind
> of failure you expect from the first disk; hardware failure?; corrupted
> filesystem? data loss? Each one of it needs a different setup in order
> to make the second disk bootable. My other concern is that, if I make
> the file copying done by cron, how would I know that the copy is not
> done while the first disk is about to fail? I'd end up with another
> non-working disk. It seems that copying it during the night is not a
> good idea. Am I right? Am I wrong?
>
> Oki
>
Reply to: