Re: StorEdge Enterprise Backup vs Veritas NetBackup

From: Len Philpot (len_at_philpot.org)
Date: 07/14/03 

Date: Sun, 13 Jul 2003 23:56:21 -0000

In comp.unix.admin "Michael wrote:
> overhead the catalog will take. Another big problem is when backups
> "expire" out of the catalog, maintenance must be run regularly to
> reclaim the expired space. In a couple of the products I've used with

This used to be an issue with us and SBU, but it's not anymore. When the
index format changed between versions 5.x and 6, the indexes shrank
quite a bit. I used to run nsrck regularly to reclaim space, but now it
stays at 100% used (i.e., no gaps in the data, a good thing). Usage in
the index filesystem dropped by close to 50% with the change.

> very large catalogs (and yours will be right up there), this can take
> over 8 hours, during which backups cannot be run nor the process halted
> or interrupted. Corrupting this database is _very_ bad as it could not

You're right about the time - With ~90 clients, some with few and some
with many (xxx,xxx) files, it generally took us about 6 hours to check
them all. With the new format, it's much more efficient. In general,
we've been advised by Sun not to run index utilities as a maintenance
tool, but rather only as needed as a repair tool. These days, the
indexes pretty much take care of themselves. However, nsrck can be
interrupted without harm. In fact, until I upped the data size and file
handles limit with ulimit, it would regularly die about halfway through,
but without any damage.

> allow you to do restores. Be sure there are tools aplenty to repair it.
> Be sure you can run backups while the catalog is in maintenance mode or
> do maintenance on only part of the catalog. If it's multiple files
> rather than a single monolothic file, that's also better.

SBU (Legato) has two levels of indexing (catalogs): The media index
contain what savesets (typically filesystems) are on what media and the
file index contains what files are in a given saveset. When the Browse
Policy is passed, the file indexes expire and its' no longer possible to
browse for files during a restore. However, you can still recover entire
filesystems and (I think) even a given file if you know the file's
location. When the Retention Policy is passed, nothing changes except
the saveset it considered Recyclable. When all savesets on a given media
pass the Retention Policy, that media is recyclable and will be
automatically reused if physically available. However, the media index
entry for that volume (tape) and saveset remain intact until the media
is relabeled (erased). To get the file index back for an expired volume,
you have to scan it in, which is generally about a 3-4 hour process for
a DLT7000 volume.

For simplicity, we generally set both policies to the same value.

> Some of the products just "forget" the files on an expired tape and
> there's no way to recover them. I know you can get them off with Legato
> but I don't know about Netbackup. This could be a problem if you're
> doing archiving. I'd stay with ufsdump for that.

I understand your point, but see above re: media index and scanning. The
data will stay on the tape and even if the indexes are deleted, scanning
it in will rebuild all that. Still, having backups on more than one
media never hurts. In fact, before I upgrade Solaris and Veritas Volume
Manager next week on our production server, I'll run backups on both DLT
and 8mm Mammoth tape. 

-- 
-- Len Philpot                          ><>  --
-- len@philpot.org       http://philpot.org/ --
-- lphilpot01@yahoo.com  (alternate email)   --

Relevant Pages

  • Re: StorEdge Enterprise Backup vs Veritas NetBackup
    ... > overhead the catalog will take. ... Another big problem is when backups ... SBU has two levels of indexing: The media index ... file index contains what files are in a given saveset. ...
    (comp.unix.solaris)
  • Re: VMS Encrypt key handling
    ...  The media goes offsite for up to 8 weeks after ... Likely we'll be using a single 'key' for backups with AES encryption ... course making sure we can restore both). ...
    (comp.os.vms)
  • Re: LS120 Capacity - Is 122880 the correct capacity for Tar Backup ?
    ... >> drives had started to become affordable and offered better reliability ... >> and performance as well as dirt cheap media (if you stuck with CDR ... when I used 4GB DAT backups years ago to keep ... drive upgrades overwhelmed the capability of the tape system. ...
    (uk.comp.os.linux)
  • Re: VMS Encrypt key handling
    ...  The media goes offsite for up to 8 weeks after ... be used for daily and weekly backups. ... The first question to consider is "Do you really *need* to encrypt your ... Current tape storage consists of 60 archived DAT 12/24 tapes, ...
    (comp.os.vms)
  • Re: NTBackup Seems to be Missing Files
    ... all subsequent backups contain empty ... >> How can backups of empty directory structures take up 100GB? ... > See what happens when you catalog. ... >>> How can backups of empty directory structures take up 100GB? ...
    (microsoft.public.windowsxp.general)