Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 1 Feb 2009 08:59:02 -0400 (AST)
From:      Andrew Hamilton-Wright <andrewhw@ieee.org>
To:        freebsd-questions@freebsd.org
Subject:   dump(8) using snapshot + "recommended" cache
Message-ID:  <alpine.BSF.2.00.0902010838030.1180@qemg.org>

next in thread | raw e-mail | index | archive | help

Hi All;

I regularly use dump(8) with snapshots to back up my server.

While this seems to have been working perfectly well so far,
upon (re)reading the man page for dump(8), I have noticed a
somewhat scary pair of lines in the paragraph describing
the option for -C <cachesize> (emphasis with stars mine):

 	[Use of this option] will greatly improve performance
 	at the cost of ***dump possibly not noticing changes in
 	the file system*** between passes.

 	***It is recommended that you always use this option when
 	dumping a snapshot.***

Does anyone know what, exactly, this means?

In particular, is the first statement actually trying to say:

 	Use of this option will greatly improve performance;
 	however it may be that changes made to the filesystem
 	made between _dump_ passes will be ignored.  ***The resulting
 	dumped filesystem image will be consistent and correct
 	based on a timestamp no later than that of the point
 	at which the dump was started***.

Is this a fair statement?  Is this guaranteed?  Or are we
trying to say that:

 	The resulting filesystem will contain images of individual
 	files based on a timestamp no later than that of the
 	point at which the dump was started, however any individual
 	files modified after the dump begins may be stored using
 	any of the version that appeared written to disk during
 	the period of the dump.



As far as the second line goes, I am not at all clear on what
this is trying to say.  Why is the cache recommended?  For
speed?  Stability?  Output correctness?  In particular, if a
snapshot dump is made without a cache option, is it potentially
corrupt?

In particular, if the second attempt above is more true than
the first, it seems to me that we should _not_ recommend the
use of a cache with snapshots, as it seems to erode the utility
of the snapshot itself.  It is for this reason that I am
suspecting that there is more here than meets the eye, which
is why I am keen to make sure that this is clear.


I am very happy to put in an update to the docs if we can make
sure that we know exactly what we are trying to say here.


Thanks,
Andrew.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.0902010838030.1180>