Does this sort of hang with 2.6.9 look familiar?

[Whit Blauvelt]

Hi,

I've had 2.6.9 running to sync a directory tree between two Ubuntu 8.04 LTS
systems for well over a year, every hour, without trouble until today.
Today, it runs through the whole tree, deletes, moves files over, then hangs
at the very end - sending the system load, which normally is minimal, up
over 10 for an extended time until killed.

This happens consistently. If I move just a smaller part of the tree, it
happens with one section, not with another. But when it happens it's
consistent. I'm suspecting it may be triggered by something in the
contents of the branch it hangs on, but there are too many files there, and
too many new today, to narrow it down to anything obvious.

Obviously this is an older rsync, and I should try the most recent version.
But it's the current rsync for Ubuntu 8.04 Long Term Service, so if this is
something known bad then I should file a bug with Canonical about it so they
can patch or upgrade appropriately.

I've researched enough to know there have been various rsync hangs dealt
with in the past. What I haven't found is what looks like a description of
the circumstances of this hang - one that occurs not in the middle of
transfer, but when all the transfers have been done. At the point of hanging
it seems all that's pending is the final message about the statistics of the
operation. But it drives the cpu "%wa" (in the "top" readout) up to around
99% at this point, and keeps it near there, running the CPU load up above 10
persistently, until killed. It can hang that way for at least an hour, on a
process where the checking, deleting and transferring to update the files
took less than a minute.

Thanks,
Whit