Re: Amanda 2.6.0 spanning
2008-06-23 12:13:01
Sweet. That's a nice setup, Paul.
I originally pictured having a separate archive configuration; but,
since I have a tape library and run tapes every night, I ended up
deciding that it was easier to do it with my regular daily
configuration. I only do archives periodically (like end of semester or
before some significant event like a server change). When I want to do
that, I just use `amadmin daily force ...` for all the servers.
Typically, I'll do that at the end of the day on Friday. When Amanda
runs that night, it does fulls of everything. Then Monday morning I
check the report and do an `amadmin daily no-reuse daily-072` with the
appropriate tape label and take that tape out of the cycle.
This means my regular daily incrementals depend upon the archive full
until Amanda gets around to scheduling additional fulls. But, my
"off-site storage" is just in another building segment (with an actual
physical firewall between segments). So, if I had to retrieve the
archive tape, it would just take 5 minutes.
I only have one drive in my library; but, with significant holding disk
space, a drive outage of a day or two is no big deal. Amanda will
automatically drop back and do incrementals to the holding disk
depending on the "reserve" specification. That gets flushed out to tape
automatically when the drive is back in place.
---------------
Chris Hoogendyk
-
O__ ---- Systems Administrator
c/ /'_ --- Biology & Geology Departments
(*) \(*) -- 140 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst
<hoogendyk AT bio.umass DOT edu>
---------------
Erdös 4
Paul Bijnens wrote:
On 2008-06-23 15:49, Johan Booysen wrote:
That's very interesting. Not sure I completely understand your setup -
one "normal" config from Monday to Thursday, and another config doing a
level 0 only on Fridays?
Yes, I have two configs for amanda.
On Tuesday to Saterday at 0:04 early morning, I run the "daily" config.
That config uses vtapes on disk as backup using the chg-disk changer.
This is a classical run having fulls and incrementals mixed each day.
It includes all the disks except some large volumes (that contain
very static files only).
Using vtapes means there is no hassle to juggle tapes each day, and
restores are easy and fast as well.
On sunday morning, I run the "archive" config. This config does only
full dumps using LTO2 tapes. Those tapes are stored offsite.
I have 2 LTO2 tape drives, joined together with chg-multi to emulate
a changer with 2 tapes.
The config has more to dump than those 2 tapes, so I instruct amanda
to collect the rest to holdingdisk using the parameters "maxdumpsize"
and "reserve".
Here are specialised settings in "archive/amanda.conf":
dumpcycle 0 # results in mostly full dumps, if enough capacity
# to store them on tape and holdingdisk
runspersycle 1
tapecycle 5 tapes # I have many more, but allow overwriting fast
runtapes 2
tpchanger "chg-multi"
changerfile "chg-multi.conf"
tapetype LTO2
amrecover_do_fsf on
amrecover_check_label on
amrecover_changer "changer"
taperalgo "largestfit" # make better use of the tape capacity
reserve 0 # Important; default= 100% reserved for incrementals
maxdumpsize 800g # Important; default= runtapes times tapecapacity
define dumptype global { # included by every dumptype
program "GNUTAR"
index yes
record no # Important; do not confuse the "daily" config!
compress client fast
skip-incr yes # never do incrementals, even when out of space
strategy noinc # avoid calculating estimates for incrementals
}
For some slower DLE's, I even have "estimate calcize" and even
"estimate server" to avoid wasting time doing the estimates.
The file "chg-multi.conf", in the same directory as amanda.conf contains:
multieject 0
firstslot 1
lastslot 2
gravity
needeject 0
ejectdelay 0
statefile /var/amanda/archive/changer-status
slot 1 /dev/nst0
slot 2 /dev/nst1
Anyway, I happen to have 2 identical tape drives, and the idea of using
both to "emulate a changer with 2 tapes" (like you do) sounds like a
good plan.
How did you set that up in amanda.conf (I'd imagine that I would have
/dev/nst0 and /dev/nst1 when both are installed)?
E.g.
runtapes (??)
tpchanger "chg-multi"
tapedev "tape:/dev/nst0" (Specify 2 tape drives?)
changerfile "/etc/amanda/daily/changer.conf" (Need any parameters set in
this file?)
changerdev "/dev/null" (??)
See above (chg-multi ignores the changerdev parameter - you may omit it)
How do you handle restores? If one tape drive dies, and you have to
restore but Amanda thinks there should be 2 slots, how would you handle
that?
As you noticed I used the parameter "amrecover_changer" to give my
changer a name. Then I can do:
amrecover -d changer
(Actually, I made that the default builtin device)
And then amanda knows how to handle the two slots in my changer
all by itself.
When one tape drives dies, you can do two things.
Explicitly mention the device you want to use for recovery:
amrecover -d /dev/nst1
or (temporarily) change the settings in chg-multi-conf
to:
...
firstslot 1
lastslot 1 # until the drive comes back from repair
...
# slot 1 /dev/nst0
# slot 2 /dev/nst1
slot 1 /dev/nst1 # drive nst0 is broken
Thank you.
Johan
-----Original Message-----
From: Paul Bijnens [mailto:Paul.Bijnens AT xplanation DOT com] Sent: 23 June
2008 13:21
To: Johan Booysen
Subject: Re: Amanda 2.6.0 spanning
On 2008-06-23 13:43, Johan Booysen wrote:
Is there anyone who is actually using chg-manual for their Amanda
setup,
who would be willing to give me an example of their config files, such
as amanda.conf and changer.conf (changerfile parameter)?
Can't help with chg-manual, but...
I avoid the use of chg-manual by having a large holdingdisk (4 disks
500GB each in raid5).
I set "maxdumpsize" to a value large enough to dump what is needed
in the run (800 Gbyte currently).
For my archive run (full dumps only on weekends), I have two LTO2
tapedrives (using chg-multi to emulate a changer with 2 tapes), but
the total backup volume now spans 5 tapes.
The first two tapes are written during the normal dump during
the weekend.
On monday morning I insert the next two tapes, and run amflush.
Somewhere early afternoon, I get a mail from amanda saying still
some files are left on holdingdisk, so I insert the next tape
and run amflush once again for last bits.
Works good enough for me.
And does not need manual intervention DURING a backup, like chg-manual
does.
I prefer having 2 tapedrives, instead of one (usually expensive)
changer. That way i'm not completely blocked when a tapedrive needs
repair.
|
<Prev in Thread] |
Current Thread |
[Next in Thread>
|
- Re: IO Errors backing up to new LTO3, (continued)
- Re: IO Errors backing up to new LTO3, Michael Loftis
- Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Ian Turner
- RE: Amanda 2.6.0 spanning, Johan Booysen
- RE: Amanda 2.6.0 spanning, Johan Booysen
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Paul Bijnens
- Message not available
- Re: Amanda 2.6.0 spanning, Paul Bijnens
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Paul Bijnens
- Re: Amanda 2.6.0 spanning,
Chris Hoogendyk <=
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Paul Bijnens
- RE: Amanda 2.6.0 spanning, Johan Booysen
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Dustin J. Mitchell
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Dustin J. Mitchell
- RE: Amanda 2.6.0 spanning, Johan Booysen
- Re: Amanda 2.6.0 spanning, Paul Bijnens
- krb5 auth problem, Chad Kotil
|
|
|