ADSM-L

Re: Archive and symbolic links

1998-12-10 19:51:03
Subject: Re: Archive and symbolic links
From: Jason Blyth <Jason.Blyth AT ASPECT.COM DOT AU>
Date: Fri, 11 Dec 1998 11:51:03 +1100
I agree with your recommendation regarding the need for an option to allow
archives to treat symbolic links as links and not to follow into them.

I also have clients who require this feature.

The only workaround is to register the same machine as a different node and
do a full incremental backup with serialization mode of absolute. However
this requires additional licenses.

Would someone please reply with details on how to put forward this
suggestion for formal consideration/review.

> -----Original Message-----
> From: Bruce Elrick [SMTP:belrick AT HOME DOT COM]
> Sent: Wednesday, December 09, 1998 1:47 PM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Archive and symbolic links
>
> Well...
>
> PTF6 is out, the symlink.doc file is gone from index.storsys, and the
> README does not mention anything new about symbolic links and archives.
> I seem to recall that symlinks and archives were going to be revisited
> by development; is this correct and does anyone recall when?
>
> An option for 'dsmc archive' like '-symlinkaslink' would be really
> handy.  AIX 4.3 html-based documentation has a looped symlink along the
> lines of:
> /usr/share/html/some/path/with/a/lang/en_US/foo ->
> /usr/share/html/some/path
> which throws an archive of /usr deep into a neverending loop.
>
> Also, a client of mine has a setup where a number of links cross
> filesystems; this causes a archive of one filesystem to cross into other
> filesystems, which seems contrary to the spirit of ADSM archives _not_
> crossing filesystem boundaries.
>
> Also, I've found the process of scheduled archives difficult at best
> because with archives, the filespec is a filespec and not a
> filesystemspec (as it is for incremental backup) so it is almost
> impossible to have a generic schedule with action=archive, since the
> objects depend on the client system.  You pretty well have to write a
> script that generates the filespecs and then schedule the script to
> run.  If the archive takes a long time, it can then push a subsequent
> scheduled incremental back past its window so the latter is missed.
>
> I've found that I have to write a script with artificial intelligence in
> it to generate the filespec and wrap it in a wrapper script that runs my
> script like a daemon, in the background with a nohup, so that the
> scheduler comes back right away and is ready for the later incremental
> backup.  If I don't wrap it, then the scheduler starts a session at the
> appointed time, starts the script which itself runs one or more sessions
> serially, while the scheduler waits for the script to finish.   While
> the scheduler is waiting its session with the server is idle and gets
> punted, so that in the event log the scheduled script shows up as a
> failure.
>
> With my wrapper, on the other hand, the scheduled archive script
> (wrapper) always shows up as successful since the success of the
> underlying script is decoupled from the success of the wrapper script.
> Either always a failure or always successful, regardless of any true
> measure of success.
>
>
> Sigh...
>
> What do you do when faced with this...?
>
> Cheers...
> Bruce
> --
> Bruce Elrick, Ph.D.
> ADSM/AIX/SP Certified
<Prev in Thread] Current Thread [Next in Thread>