[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Restoring large archive with network failures

Hi Colin & Graham,

I restored a large archive using the --resume-extract option and it worked beautifully. I ended up writing a throwaway node script for comparing the written files with the metadata in the list and it all looked good. Thank you for the great work!

вс, 16 сент. 2018 г. в 21:34, Fede Pereiro <fpereiro@gmail.com>:
Hi Colin & Percival!

I think I never subscribed to the list -- I nonchalantly assumed that no subscription was necessary : ). Just subscribed.

Thank you for the response and the PR. I think I will use that code to recompile tarsnap and give it a try. I will also write a _javascript_ (nodejs) script to compare the list output from tarsnap against the file sizes & dates, to detect any discrepancies - hopefully this will catch any bugs from the new code. When I am done with the script I will post it in github and send you the link.


сб, 15 сент. 2018 г. в 15:16, Graham Percival <gperciva@tarsnap.com>:
Hi Fede,

I've just uploaded a PR that adds a --resume-extract option for precisely this

If you're comfortable recompiling Tarsnap, by all means give it a try!  There's
always the warning that this is relatively untested code right now.

- Graham

On Sat, Sep 15, 2018 at 11:37:41AM -0700, Colin Percival wrote:
> Hi Fede,
> I just found this in my spam filter -- it looks like your mailing list post
> may have been eaten (are you subscribed to the list with this email address?)
> as well.
> Feel free to re-send this to the list, but I think the answer to your question
> is (a) yes you're right that using --keep-newer-files for this has a problem
> relating to files which are partially downloaded when an extract fails, and
> (b) there's a work in progress for adding a --resume-extract option which will
> catch cases like this... I think Graham (CCed) may have a patch you can try,
> if you're comfortable recompiling Tarsnap.
> Colin Percival
> On 9/10/18 8:28 AM, Fede Pereiro wrote:
> > Hi everyone!
> >
> > I'm in the process of restoring (-x) a large archive to my local disk. During
> > download, I have experienced network errors that have interrupted the download.
> >
> > If I initialize the process again, using the *--keep-newer-files *to only
> > download the missing files, the restore will eventually be complete. My main
> > concern, however, is that a network error can leave a file partly downloaded.
> > In this case, when I re-run the command, said file won't be re-downloaded and
> > I will be left with an incomplete file.
> >
> > To avoid this, I would have to manually delete the last file downloaded before
> > the network failure and re-enter the command. Another option would be to
> > concoct a script that compares the tarsnap -t output with that of a recursive
> > ls and spot files with different sizes (then I could manually delete them and
> > run the command again).
> >
> > Before I do either of the above, I ask: is there a more reliable and efficient
> > (both in personal time and in bandwidth cost) to restore large backups when
> > using a connection that experiences network failures?
> >
> > I take this opportunity to thank cperciva and the community for creating and
> > maintaining Tarsnap.
> >
> > Thanks!
> --
> Colin Percival
> Security Officer Emeritus, FreeBSD | The power to serve
> Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid