Uh, I don't know exactly which partition is it. Strangely, I could not
reproduce this with a mock-up load (find -exec cat {} \; >/dev/null),
but odds are that it usually happens while reading a half-filled 500GB
partition.
Anyhow, I can put it on my public-facing NAS and give you either a
download link (my uplink is 50 Mbps wide) or ssh access, if you would
be kind enough to have a look at it :)
BTW, if I enable checkpoints in borg (so it saves internal state
periodically) and restart the backup from a checkpoint after the crash,
it always finishes successfully. Does that confirm or refute your
hypothesis?
Thanks,
--
Ivan Shapovalov / intelfx /
On 2017-06-02 at 23:37 +0300, Edward Shishkin wrote:
> Alas, can not reproduce it.
> I suspect garbage on the leaf level. For some reasons
> fsck can not detect it (on semantic pass).
> How big is your partition?
> Could you dd it to a file and let me download that file?
>
> Thanks,
> Edward.
>
> On 06/02/2017 01:59 AM, Ivan Shapovalov wrote:
> > Looking forward to your results.
> >
> > And by "looking at the tree" you mean the debugfs output, right?
> >
> > Thanks,
>
> Attachment:
signature.asc
Description: This is a digitally signed message part
