Jon LaBadie wrote:
> On Thu, Jun 27, 2002 at 09:30:17AM +0100, Martin Hepworth wrote:
> 
>>Jon
>>
>>Well it's not happy - amanda is constantly trying to do a level 0. So 
>>it's seeing this as a fail.
>>
> 
> 
> 
> I find it is generally easier to follow a thread
> if new comments are added at the bottom.
> 
> 
>>--
>>Martin
>>
>>Jon LaBadie wrote:
>>
>>>On Wed, Jun 26, 2002 at 03:47:41PM +0100, Martin Hepworth wrote:
>>>
>>>
>>>>Hi all
>>>>
>>>>well I've been fiddle with my new shiney iMac G4 trying to get amanda to 
>>>>back the thing up. Currently I'm  getting this from the backup report..
>>>>
>>>>FAILURE AND STRANGE DUMP SUMMARY:
>>>>stuartdmg4 / lev 0 FAILED [/usr/bin/gnutar returned 2]
>>>>
>>>
>>>  [ snipped part and rearranged next line ]
>>>
>>>
>>>>sendbackup: argument list: gtar --create --file - --directory / 
>>>>--one-file-system
>>>>  --listed-incremental /var/amanda/gnutar-lists/stuartdmg4__0.new 
>>>>  --sparse
>>>>  --ignore-failed-read --totals .
>>>
>>>
>>>  [ snipped part ]
>>>
>>>
>>>
>>>>sendbackup: index created successfully
>>>>error [/usr/bin/gnutar returned 2]
>>>>sendbackup: pid 588 finish time Wed Jun 26 01:11:36 2002
>>>
>>>
>>>Note: amanda uses the "--ignore-failed-read" option to tar.  To not use 
>>>this
>>>would cause tar to abort if it ever failed to read a file successfully.
>>>Even something as benign as file removed before tar got to it.  But
>>>generally permissions problems.
>>>
>>>When using this flag, and encountering one or more failed reads,
>>>tar exits with a status of 2 rather than 0.  This is what amanda is
>>>telling you.  The interpretation is that tar had to skip one or more
>>>files.  I don't know of any way to determine which files.  The rest
>>>of the dump should be fine.
>>>
>>
>>
>>
>>
>>>>End of included message <<<
>>>
> 
> Well, I hate to suggest this, but ...
> 
> On my unix system I use gnutar 1.13.25.
> I use the same gnutar version on an amanda client I run on a Win2K box under CYGWIN.
> 
> The unix amanda is version 2.4.2, on the w2k box it is 2.4.3b2.
> The unix box never exhibits the problem you are seeing, the w2k box always "did".
> 
> I don't know if my unix box, version 2.4.2, never sees any failed reads, or whether
> there is code in amanda 2.4.2 that accepts the exit status of 2 without considering
> it a failure.  But the 2.4.3b2 under w2k/cygwin is sensitive to an exit status of 2.
> 
> So I had to remove the sensitivity or change gnutar so it did not exit 2.
> I chose to modify gnutar.  I always have amanda use its own copy of tar
> so if I want to customize it, or shell wrapper it, I can without affecting
> the system version of tar.
> 
> If you are compiling or can compile your own gnutar, it is a simple change.
> 
> At the end of main(), at the very end of the file tar.c is an if statement.
> 
> Nominally it says "if the exit status is 2 (aka TAREXIT_FAILURE) print an
> error message".  I simply added an exit(0) in the if statement.  Actually,
> the curley braces, a comment, and the exit(0);.
> 
> Here is the resulting code with the changes (at the end of tar.c).
> 
>         if (exit_status == TAREXIT_FAILURE)
>         {
>             error (0, 0, _("Error exit delayed from previous errors"));
>                 /* added for amanda */
>             exit(0);
>         }
>         exit (exit_status);
>     }
> 
> 
> I scanned the gnutar code and as far as I could see, the only place the
> "exit_status" variable is set to TAREXIT_FAILURE (defined as 2) is when a
> failed read occured and was ignored.  So I don't think any other failure mode
> would be affected.  But I've been wrong before :(
> 

Jon

re reply areas: yeah I know - lack of caffene :-)

Re the other stuff.

The client/server are both using amanda 2.4.2.p2

Maybe I;m being totally stupid but I couldn't find anything above 1.13 
(ie I couldn't find the beta's anywhere, just mentions of it.). I'll 
have another dig tomorrow (knocking off time now). However if you have a 
URL to hand :-)

I'll have another dig tomorrow.

--
Martin


Reply via email to