On 2011-03-09 07.48, Brett Lymn wrote:
> On Wed, Mar 09, 2011 at 01:49:04AM +0100, Benny Lofgren wrote:
>> I don't know about the rest of you guys, but to me that sounds exactly
>> like a bug... especially since nothing is mentioned of such behaviour in
>> the dump man page.
> Well, you would have to totally redo the guts of dump if you want to
> "fix" this well known behaviour.  Dump scans and writes the meta-data
> first and then writes the data for the files.  If you have ever used
> restore you would note that it builds the directory structure first and
> then puts the data back.  Dump comes unstuck when the file system meta
> data changes during the backup - when files are added or deleted.  The
> safest way to do a dump is when the machine is in single user mode for
> this very reason but many people play fast and loose because they cannot
> wear the outage for a backup, in that case you must quiesce the file
> systems as best you can.  Really, this is well known unix sys admin
> procedure.

Oh, I'm well aware of how dump and restore works. I only objected to the
behaviour the OP described not being labeled as a bug. (Ok, in fact
there is a mention in man restore, where it states in the bugs section
that "restore can get confused when doing incremental restores from
dumps that were made on active file systems". The definition of
"confused" however, is... well, confusing. :-) )

I may have misread the OP (english not being my first language, after
all), but my interpretation was that the question of whether there is a
bug was concerning the endless loop due to not recognizing "none" rather
than restore:s confusion as to how many dump volumes there were in the
dump set.

> Have a go at fixing it, by all means, but note that people _like_ the
> interactive restore mode where you can select the files to restore by
> browsing and would find it unacceptable to wait for a full tape scan
> before they can perform this task.  It will be interesting to see how
> you go about handling files appearing and disappearing during the
> backup.

Not having looked at the source I'm probably putting my foot in my mouth
here, but it wouldn't seem that impossible to fix restore so that it
recognizes the "none" input properly in the OP:s described context
(especially since it recommends it itself in the prompt!).


Regards,

/Benny

-- 
internetlabbet.se     / work:   +46 8 551 124 80      / "Words must
Benny LC6fgren        /  mobile: +46 70 718 11 90     /   be weighed,
                    /   fax:    +46 8 551 124 89    /    not counted."
                   /    email:  benny -at- internetlabbet.se

Reply via email to