[OpenAFS] backup strategy

Russell Button russell@telmate.com
Mon, 10 Nov 2014 11:53:33 -0800


--089e01538c487d27cb0507868423
Content-Type: text/plain; charset=UTF-8

I'm the new guy on the sysadmin team at the company.  Of course I get
annointed to be King of Backups.  Nobody else wanted to do it and I'm a
sucker for always being willing to do the dirty and un-fun work.   I've
never worked with AFS before, let alone configured a backup strategy for
it.  So this is a case of the blind leading the blind.  I just got in a
copy of Richard Campbell's "Managing AFS" book because it had a chapter on
backups.  The backup tool I've been reading about reminds me a lot of
bacula.   But bacula, like most UNIX/Linux backup tools keeps a database of
what gets backed up by hostname/file system/date stamp and such.  Which is
fine for backing up traditional machines, but not necessarily sufficient
for our situation.

Like I said, I'm a total NOOB with regards to AFS.   From what I
understand, we have something a bit over 200TB spread out over a number of
different physical machines, spread out over 3 geographically distant data
centers.   Furthermore, this is a volume which will continue to grow
indefinitely.  Chances are we'll have to keep on-line at least 3 to 5 years
of data, and unlike e-mails at the White House,  everything will have to
permanently archived and retrievable.

I get the impression that AFS is this amorphous cloud of data storage.
 So when you backup stuff, it's not as if it's organized by machine and
file system.  With this much data, spread out over 3 geographically distant
data centers, it's not as if you can do a full dump on the 1st of the month
and then do daily incrementals for the month, and then start over again
next month.  And that's where the backup paradigm sort of breaks down for
me.  Effectively you have to keep everything forever, but you really only
want to backup a given file once.

Does anyone Out There have a similar problem, and if so, what strategy did
you use?

Thanks!

Russ

--089e01538c487d27cb0507868423
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">I&#39;m the new guy on the sysadmin team at the company.=
=C2=A0 Of course I get annointed to be King of Backups.=C2=A0 Nobody else w=
anted to do it and I&#39;m a sucker for always being willing to do the dirt=
y and un-fun work. =C2=A0 I&#39;ve never worked with AFS before, let alone =
configured a backup strategy for it.=C2=A0 So this is a case of the blind l=
eading the blind.=C2=A0 I just got in a copy of Richard Campbell&#39;s &quo=
t;Managing AFS&quot; book because it had a chapter on backups.=C2=A0 The ba=
ckup tool I&#39;ve been reading about reminds me a lot of bacula. =C2=A0 Bu=
t bacula, like most UNIX/Linux backup tools keeps a database of what gets b=
acked up by hostname/file system/date stamp and such.=C2=A0 Which is fine f=
or backing up traditional machines, but not necessarily sufficient for our =
situation.<div><br></div><div>Like I said, I&#39;m a total NOOB with regard=
s to AFS. =C2=A0 From what I understand, we have something a bit over 200TB=
 spread out over a number of different physical machines, spread out over 3=
 geographically distant data centers. =C2=A0 Furthermore, this is a volume =
which will continue to grow indefinitely.=C2=A0 Chances are we&#39;ll have =
to keep on-line at least 3 to 5 years of data, and unlike e-mails at the Wh=
ite House, =C2=A0everything will have to permanently archived and retrievab=
le.</div><div><br></div><div>I get the impression that AFS is this amorphou=
s cloud of data storage. =C2=A0 =C2=A0So when you backup stuff, it&#39;s no=
t as if it&#39;s organized by machine and file system.=C2=A0 With this much=
 data, spread out over 3 geographically distant data centers, it&#39;s not =
as if you can do a full dump on the 1st of the month and then do daily incr=
ementals for the month, and then start over again next month.=C2=A0 And tha=
t&#39;s where the backup paradigm sort of breaks down for me.=C2=A0 Effecti=
vely you have to keep everything forever, but you really only want to backu=
p a given file once.</div><div><br></div><div>Does anyone Out There have a =
similar problem, and if so, what strategy did you use?</div><div><br></div>=
<div>Thanks!</div><div><br></div><div>Russ</div></div>

--089e01538c487d27cb0507868423--