On Fri, 2009-04-10 at 15:26 -0400, Gene Heskett wrote:
> On Friday 10 April 2009, Dustin J. Mitchell wrote:
> >On Fri, Apr 10, 2009 at 1:54 PM, Gene Heskett <gene.heskett AT verizon DOT 
> >net> 
> wrote:
> >> Dustin;  It is looking like its catching up, after several runs with
> >> intervening versions, it has caught up with whatever change made that made
> >> it think everything was new.  I don't believe it was device major/minors
> >> that did it, since when the first time I ran the 0331 version it was a
> >> total failure, not 'to big' failures like I have had with 0321 & newer
> >> till its caught up again.  Reverting to 0319 from 0331 made it run normal.
> >>  But every newer version has acted like the device mapper had moved all
> >> the disks around, and the option --no-device-check was being ignored.  So
> >> I have NDI what the real problem is/was.  I am continuing to build & test,
> >> 0327 is building now.
> 
> You recall that when I went from 0319 to 0331, it wasn't this, it was a 
> total, 
> all results missing failure, and reverting to 0319 fixed that, so I believe I 
> am looking at 2 distinct problems.
> 
> >As in the past, this problem really exists between tar and the kernel,
> >so Amanda versions aren't relevant.  The one thing that you can check
> >for us (as opposed to for the kernel or tar developers) is that Amanda
> >is properly sending the --no-device-check to tar.  You should see this
> >either in your sendbackup debug logs or your runtar debug logs.
> >
> >Dustin
> 
> I can see it in the htop report while its actively running, but haven't 
> looked 
> at the runtar debug logs yet, hang on.  Mummmmmmm, no runtar logs. I am 
> running amgtar.
> 
> From the htop screen: No, I can't show it, amgtar has finished all DLE's for 
> this run, and the printer just spit out a normal report.
> 
> I just grepped all the amgtar.*.debug files, and it is not noted in any of 
> them.  This begins to quack like amgtar is ignoring that option?  Please say 
> it isn't so...  If so, it seems like it should log that the option wasn't 
> done, somehow.

If I'm following at all, for amgtar, you should be setting this as a
property field in your amgtar application section (in my case "yes")

        property "CHECK-DEVICE" "yes"

and it should be showing up in the amgtar.*.debug log as

1239319047.814503: amgtar: CHECK-DEVICE yes

Paul

> 
> It any event, I haven't hit the total failure of 0331 yet, so I will proceed 
> with the next one, 0330, as 0327 seemed to run normally.  Stay tuned...
> 
-- 
Design Engineer
Zmanda, Inc.
http://www.zmanda.com/