rsync stuck in infinite loop? (was How do I make rsync ignore unreadable files (damaged sectors)?)

[Carlos Carvalho]

Wayne Davison (wayned at samba.org) wrote on 22 December 2009 08:02:
 >On Tue, Dec 22, 2009 at 4:18 AM, Tomas Gustavsson <tomplast at gmail.com> wrote:
 >> Still, I do think that rsync should give up after a long time, but it doesn't.
 >
 >Yeah, if it gets a read error for a part of a file, it substitutes
 >zeros for the data and keeps trying to read the file.  That could make
 >it take an eternity for a large file that is entirely unreadable.  It
 >would also mark the file bad, and retry sending it, so expect double
 >the fun.

About a month ago something very strange happened here: rsync
apparently got stuck in an infinite loop, consuming a lot of cpu but
doing no I/O. Total cpu times ranged from about 40min to over 2h when
I killed them manually. It was on the destination machine, I don't
know if it was the receiver or generator. It happened a few times,
with both 3.1 and 3.0.6, and with 2 origin machines; one of them runs
protocol 29 and the other 30. It was a standard recursive/delete run,
on a repository with about 800,000 files. The problem disappeared
magically and never happened again.

Are there situations where rsync may get trapped in infinite loops?
Might it have any relation to the above issue?