On Fri, Jan 05, 2001 at 08:53:38AM -0500, Hal Haygood wrote:
> I'm working with a fairly large distribution system that was recently
> switched from rdist to rsync. I am having some incredibly frustrating
> problems with rsync dying.
>
> I have tried everything I can find any information about, but it continues to
> happen, sometimes in new and spectacular ways.
>
> The most common failure mode is an unexplained 'unexpected EOF in
> read_timeout'. That's with no special command-line options, just '-azvHl
> --timeout=900'. If I set the --rsh=/bin/rsh option (to avoid using the
> /bin/remsh symlink), I see the same error interspersed with timeouts.
>
> If I use --blocking-io to force the issue, I start getting all kinds of
> different 'unexpected tag' errors, with what appear to be random numbers
> between -127 and 127 as the unexpected tag.
Hmm, I thought that blocking-io was needed to *avoid* the unexpected tag
errors with rsh. By default blocking-io is turned on only if the --rsh
command is exactly equal to the default "rsh" so by your setting it to
/bin/rsh you turn blocking-io off. Hey, I see the man page does not
explain that behavior, that's an oversight. I'll make sure that gets
fixed.
> I've tried 2.2.1, 2.3.1, and 2.4.6; all exhibit similar behavior.
>
> Help! I'm at my wit's end dealing with this. What combination of rsync
> version and command-line options actually *works* on Solaris 2.5.1/2.6/2.8
> with rsh?
Yesterday I discovered a hanging on Solaris 2.8 when simply rcp'ing a
particular file. Do all of your timeouts involve a Solaris 2.8 system?
Does it always hang on the same file?
- Dave Dykstra
