New thread, trying to summarize what I try to achieve here.
amanda-3.5.1 on a gentoo linux server
1x LTO6 tape drive, 8 slot tape library
around 12 TB data to dump
More or less enough holding disk to keep a full lev0 of all the DLEs.
-
The normal schedule (let's call it amanda config "daily") runs MO-TU and
is run via a specific option:
I skip some DLEs on weekdays as they are huge and take too long to dump
... they are only dumped on FR and saturday (then the dump sometimes
runs for >20 hrs)
-
aside from that I need to "generate" a set of tapes with lev0 of all
DLEs on it that can be archived (and that should be done soon ... in
january or so)
So I have to make sure I collect lev0 dumps on holdingdisk and then run
one big "archive flush".
Several approaches so far and I get lost already:
* separate config "archive", shared disklist, tuned dumptypes
* separate storage definition "archive" inside config "daily" with
specific tape-labels defined
* amvault (with or without vtapes as intermediate step)
* RAIT (don't like that ...)
A requirement has been removed, that makes it easier: I don't have to
use WORM tapes anymore ... this gives me more attempts to get the set
written correctly.
-
Right now the 2nd approach looks best to me.
See my current storage definitions:
define storage daily {
tapepool "daily"
tapetype "LTO6"
tpchanger "robot"
#labelstr "^CMR[0-9][0-9]*$" # label constraint regex: all
tapes must match
autolabel "CMR%b" # label constraint regex: all tapes must
match
#labelstr "^ARC[0-9][0-9]*$" # label constraint regex: all
tapes must match
flush-threshold-dumped 500 # (or more)
flush-threshold-scheduled 500 # (or more)
#taperflush 500
taperflush 0
autoflush no # keep FULL DLEs for archive tapes
runtapes 4 # number of tapes to be used in
a single run of amdump
}
define storage archive {
tapepool "archive"
tapetype "LTO6"
tpchanger "robot"
labelstr "^ARC[0-9][0-9]*$" # label constraint regex: all
tapes must match
#autolabel "ARC%b" # label constraint regex: all tapes must
match
runtapes 4 # number of tapes to be used in
a single run of amdump
dump-selection ALL FULL
}
storage "daily"
----
The idea is to let the normal runs write to storage "daily" and
additionally keep the dumped files in the holdingdisk.
After having all DLEs there I want to do something like:
amflush daily -o storage=archive
and get all FULL backups flushed to that specific set of tapes.
What do you think?
Right now my values for the flush-thresholds etc are way off ... I got
lost somehow and could need some checking eyes here.
Thanks in advance, regards, Stefan
ps: I already managed to do that one or 2 years ago ;-)
