On 2/5/19 5:24 PM, Debra S Baddorf wrote:
On Feb 5, 2019, at 4:14 PM, Chris Hoogendyk <[email protected]> wrote:
I'm running Amanda 3.4.5 on Ubuntu LTS Server 14.04 with an Overland NeoSeries
48 slot library with two LTO7 drives.
I have near 100TB of external drives in various mdadm raid configurations with
LVM managing the storage and about 110 DLEs in my disklist.
One of the groups using the storage on this server has about 11TB of data
representing more than 4 years of work by a group of people. It would be a real
disaster to lose it, and they want an extra set of tapes as an archive.
It seems pretty clear that amvault will do this. However, I'm trying to figure out the
best way to configure it and get it done. If I set up a storage and label some archive
types to belong to that, then I can get amvault to copy the latest fulls of a set of DLEs
that correspond to their work from the regular backups to the archive tapes. I don't see
any specific examples that show how to do that in detail (specifying particular DLEs);
but, I'm also concerned that it would tie up two tapes drives (from and to), require me
to load the appropriate tapes whenever it needed them, and probably take long enough that
it would get in the way of the regular backups. I'm thinking it is likely to work like
amrecover, where it tells me what tapes it needs and then I do `amtape -o
storage="research" daily slot 7` (after looking at tapelist and changer status
to figure out the slot).
Is there some other way of configuring or doing this that is more efficient? Or
that would only tie up one drive? I haven't done this yet. I'm just beginning
to dig into it. Amanda 3.4.5 has amvault, but is it missing anything that would
be gained by upgrading to, say, 3.5.1?
Note that I do have holding disk space, but it is not excessively large in
terms of what I am normally backing up or want to do for this archive job. It
is made up of two 4TB SSDs. I could perhaps add some more holding disk by
carving it out of LVM, but that would probably be slower with respect to the
speed of the LTO7s.
Is there any good documentation online beyond the man pages for the amvault and
storage features? With examples?
Chris Hoogendyk
For my own archival storage, I’ve found it easiest to create a second
configuration. No amvault.
Just a second cronjob that I run one weekend per month. It ONLY does level
0’s, and uses the same
DLE that my “daily” job uses. But a different amanda.conf so that I can
specify no-incremental, no
re using tapes. More detailed info, if this option appeals to you.
It takes longer than a daily job, for me, but uses the same tape drive.
(Daily jobs don’t get to run
on the weekend when this occurs).
But I think you’ve already considered this option, and didn’t go that route.
Deb Baddorf
Fermilab
I'll take that into consideration.
I was assuming the amvault would be more efficient and faster, since it doesn't have to go through
the tar pipe. Instead of tarring up a DLE, it would simply dd it from an existing LTO7. I'm not sure
if amanda would be smart enough to construct a single dd from one LTO7 to another LTO7, but it would
seem that that would be rather fast.
Some time ago, I quit using gzip and just let the LTO7 do its own compression. I lose some planning
and information that way (can't tell how well it actually compressed), but it seems to do well and
is substantially faster. My server doesn't get overloaded with gzip processes.
Also, I'm hoping I can configure an amvault archive, add tapes to it in the future, and specify
different DLEs that people want to archive, all without changing the configuration, just marking
tapes as don't re-use after I have an archive on them.
--
---------------
Chris Hoogendyk
-
O__ ---- Systems Administrator
c/ /'_ --- Biology & Geosciences Departments
(*) \(*) -- 315 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst
<[email protected]>
---------------
Erdös 4
