Re: [Duplicity-talk] Out of space error while restoring a file

[edgar . soldin]

On 26.09.2012 11:42, Laurynas Biveinis wrote:
[SNIP]
>>>
>>>>> So, "twice the size of the restored file + one volume?" But originally
>>>>> it failed with 80GB free for a 30GB file, so it's even more than that.
>>>>
>>>> right, we would have to disect filesystem usage during restoring to find 
>>>> out what is going on here. actually the v9 dump tells you when temp files 
>>>> are created and deleted (though not their size) so you can work through 
>>>> that if you want.
>>>>
>>>> alternatively you have to observe what happens in the tmp folder during 
>>>> restore ;)
>>>
>>> Here's the open file snapshot around volume 704 (out of ~750):
>>> http://pastebin.com/pGFH6Wju
>>>
>>> Note that "Sandėlys" has been replaced with "Elements", that's another
>>> volume (so that locally-downloaded backup sets would fit)
>>>
>>> Thanks again,
>>>
>>
>> ok, just wanted to check the command line output promised above, just to see 
>> there was nothing attached ;)
> 
> Oops. Fixed.
> 
>> wrt to the lsof output. it shows
>> duplicity 8946 laurynas    7u   REG   8,33 25211305984     8622 
>> /media/Elements/Laurynas/tmp/duplicity-DL3ypc-tempdir/tmpgxn0pY (deleted)
>> and
>> duplicity 8946 laurynas   12u   REG   8,33 34465812480     8614 
>> /media/Elements/Laurynas/tmp/duplicity-DL3ypc-tempdir/tmpYZgzig (deleted)
>>
>> and some others, all of them have the tag (deleted), so i assume the space 
>> was already freed again?
> 
> Nope, it's taken space.
> 
>> even if not this is what we assumed: 2 times the 30GB file plus some.
>>
>> what would be interesting is a "multiplexed" output of 'duplicity -v9' _and_ 
>> and a periodic 'ls -la' of TMP . so one could see what files duplicity 
>> creates/deletes and how this reflects in the TMP folder. i believe this 
>> would be possible with bash magic.
> 
> I'l see about producing this. But, ls -la will not show those big
> deleted files. I will try to include lsof as well, perhaps filtered to
> remove open net sockets and the like.
> 

first off, the v9 output does not show any tmp files named like tmpYZgzig but 
only mktemp-MRoS1d-741 ones. so i assume the default python tempfile does not 
inform us 

or

you are not using the latest patch?
http://bazaar.launchpad.net/~duplicity-team/duplicity/0.6-series/revision/860

regarding the used space. did you monitor via du/ls/df the duplicity tmp folder 
and can confirm that it uses up much more than the 2xbiggest file plus some? 
how much?

..ede/duply.net

..ede