Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 6 Dec 2002 12:05:47 -0500 (EST)
From:      Jerry McAllister <jerrymc@clunix.cl.msu.edu>
To:        lists-freebsd-questions@biaix.org (Joan Picanyol i Puig)
Cc:        freebsd-questions@FreeBSD.ORG
Subject:   Re: clarification of man dump
Message-ID:  <200212061705.gB6H5lP02054@clunix.cl.msu.edu>
In-Reply-To: <20021206154739.GB69019@grummit.biaix.org> from "Joan Picanyol i Puig" at Dec 06, 2002 04:47:39 PM

next in thread | previous in thread | raw e-mail | index | archive | help
> 
> Hi,
> 
> I have some trouble understanding the last part of 'man dump', from "An
> efficient method of...". My understanding is that the example means:
> 
> 1. Take a level 0 dump once a month
> 2. Take a level 1 dump once a week
> 3. Take a level 3 dump on mondays, a level 2 on tuesdays, a level 5 on
> wednesdays and so on until a level 9 on sundays.
> 
> Am I missing something? What's so "efficient" from doing it this way
> (why)?

It really depends on your needs.   

Basically, a leval 0 dumps everything.
A level '1' dumps everything that has changed since a level '0' dump
A level '2' dumps everything that has changed since a level '1' or '0' dump.
A level '3' dumps everything that has changed since a level '2' or lower
dump, etc.   This is modified by nodump flags and maybe some other things,
but that is the general pattern.

So, if you do a level '0' once per month then you start fresh, so to
speak, every month.  Then if you do a level '1' each week you can
get everything restored from just one monthly  and one weekly dump,
except anything that has changed since the last weekly dump, etc.

Then for the rest of the days you have to juggle the remain dump levels.
It used to be systems only recognized up to level '5' so that added to
the complexity.

The complicated pattern they give which will tend to reduce the total 
number of tapes you have to write for your dumps or read back in case 
of a restore.

That was all thought up when tapes didn't hold very much and neither
did disks.  First of all, people had to keep as little as possible online 
and so they were always moving things off and on disk from some other 
storage.  So, the online data pool changed a lot every day.  Because
tapes were small, dumps could take a lot of them.    Because a lot 
changed often - not only new data, but stuff that had been pushed off 
and on.

Not only full dumps, but change dumps (levels 1 thru whatever) would
take a lot of tapes.  It was not uncommon for a full dump to take
say 40 to 100 tapes and a change dump to take 20 or 60 tapes and that
would be on a fairly ordinary system.  Really big ones were awful.

When there was a failure one had to go back to the last full dump and
then through whatever series of change dumps were needed to get the
full latest stuff restored.    So, if a dump set repeatedly dumped
certain files (doing a level '0' monthly and then '1' daily for example)
you could end up using a lot more tapes than necessary.   But, if you
did a higher level each day, you quickly run out of levels and may
also have a lot of tapes with only a little bit on them.  So, they
tried to come up with a pattern of tape use that was most efficient.

But, now days, people have giant disks and leave most things online
all the time.  The vast majority of it doesn't change.   Plus tapes 
are relatively huge too, so a full dump may take as little as one tape 
on a smaller system and maybe only ten or twenty on a bigger system 
(ignoring the problem of those huge huge databases for now).  Change 
dumps (level '1' or higher) generally take only one or maybe two tapes
except for some either very big or very busy systems.   So, the needs 
have changed and what efficiency means has also changed.  

Our main systems do a full dump once per week and use just one single
higher level (eg level '1') each day in between, ignoring level '2', '3',
etc.   That means a lot of files that have changed only once get redumped 
each day, but the change dump all fits on one or two tapes, so who cares.   
It is less complicated and easier than managing all those interacting levels. 

Actually, for some of my machines that are just used lightly - not so much 
change going on, I just do a weekly level 0 dump and don't even bother
with the daily change dumps.   

By now you probably have all this figured out anyway, but that is
some general perspective.   Probably others know the arithmetic of
how they decided the pattern better than me, but,...

////jerry

> 
> tks
> -- 
> pica
> 

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200212061705.gB6H5lP02054>