On Sat, Mar 09, 2002 at 09:29:30AM -0500, Jenn Sturm wrote:
[...]
> My initial thought was to have several different configurations, and
> do a full with each configuration once every two weeks, then an
> incremental or two with each configuration every few nights. It seems
> to me, though, that the way Amanda is best set up is to have one
> giant configuration with every single machine and disk in there, and
> let it do a full once per dumpcycle, then handle the incrementals of
> all machines on a nightly basis.
more or less.
> Is my understanding of this correct? And if so, why is it preferrable
> to have one giant config with every machine in it instead of a number
> of configurations broken down by logical machine groupings?
at my site we have 100+ hosts and 500+ disk partitions with total disk
data amounting to 500+ Gb.
at one time we tried syncronizing our "archival" (level-0 on everything)
runs with the regular amanda run which does it's own balancing of
full dumps and incremental dumps. we did full dumps on weekends in
*separate configurations*, and then used amadmin Daily force to tell
amanda to force all the disks which had been full-dumped on the weekend
archival run with the Daily config set to "skip full" backups...
sigh. there ain't not graceful way. stuff like this is just too error-
prone and admin-intensive. you really need to spend alot of time making
sure things go smoothly, like do not force your daily to skip disks
which failed the archival.
what we do now is to simply run the Daily config without trying to sync
it with the archivals. again the Archival must have it's own separate
configuration. then for the archival backups, use the "no-record"
option for all the disktypes. this supresses the archivals being
recorded in /etc/dumpdates, which would otherwise confuse the Daily
configuration's normal operation.
make sure you do not run your Daily configuration at the same time as
your Archival! our archival run takes 30 to 40 hours to run. also, you
might not have room in your tape changer for both the archival tapes and
daily tapes.
> Thanks in advance,
>
> Jenn
cheers,
chris
--
Christopher Linn, <[EMAIL PROTECTED]> | By no means shall either the CEC
Staff System Administrator | or MTU be held in any way liable
Center for Experimental Computation | for any opinions or conjecture I
Michigan Technological University | hold to or imply to hold herein.
