rsync - Dec 2002 - rsync 2.5.5, "unexpected tag failures", Solaris 2.6 vs. 2.8, "--b locking-io" workaround

Hi,

I have been having a problem that seems related to previously reported
problems
but persists.

I am syncing from a 2.6 Solaris box to a 2.8 solaris box.  All are running
rsync 2.5.5.

When I sync (commands simplified for example, always using rsh):

2.8box% rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/

i.e., the 2.6 box is the server, I get spurious, persistent but inconsistent
"unexpected tag <some number>" failures.  This is,
unfortunately, on the 2.8
box, the way I have to do it.

When I sync:

2.6box% rsync -e rsh /path/path/stuff/ 2.8box:/path/path/newstuff/

(the 2.8box is the server), it works.  But the 2.6 box isn't where the
script runs.

The first command ALSO works if I use the "--blocking-io" option, 

2.8box%rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/
--blocking-io

but I'm not sure what the possible side-effects of that might be.  (I picked
up from the CVS archive that there was some kind of relationship between a
"socketpair" Solaris bug and the "--blocking-io" option, so
tried it, but
that seemed many versions ago).

I am not doing the builds myself - sysadmins are doing it.  Is there
anything special they
need to do in the 2.6 build to avoid my having to use the
"--blocking-io"
flag?

Thanks,

Dave Madole

All,

I am using rsync to sync multiple files in the background of some other
tasks from server to clients. It is launched dynamically from any of the
client web cgi-s and in order for apache to continue and not delay user
pages I am running in in a forked chiled with all IN/OUT/ERR closed on
it. 

Now after Dave's suggestion I am testing "--ignore-existing" knob,
but
unfortunately rsync just logs a core. How can I troubleshoot such a case
? If I leave OUT/ERR open it works OK. (of course -v option is not
present). 

Any other tip in reading rsync core files would be welcome as well !

Thx,
R.

The problem is in your use of "-e rsh"; --blocking-io is assumed if
the -e value is equal to the RSYNC_RSH define which is usually "remsh"
on solaris but maybe it isn't on solaris 2.8.  The --blocking-io option
is required for most versions of rsh, but it's sometimes difficult for
rsync to know when rsh is being used.  It's very confusing, I know.

- Dave

On Tue, Dec 03, 2002 at 10:42:19AM -0800, Madole, Dave BGI SF
wrote:
> 
> Hi,
> 
> I have been having a problem that seems related to previously reported
> problems
> but persists.
> 
> I am syncing from a 2.6 Solaris box to a 2.8 solaris box.  All are running
> rsync 2.5.5.
> 
> When I sync (commands simplified for example, always using rsh):
> 
> 2.8box% rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/
> 
> i.e., the 2.6 box is the server, I get spurious, persistent but
inconsistent
> "unexpected tag <some number>" failures.  This is,
unfortunately, on the 2.8
> box, the way I have to do it.
> 
> When I sync:
> 
> 2.6box% rsync -e rsh /path/path/stuff/ 2.8box:/path/path/newstuff/
> 
> (the 2.8box is the server), it works.  But the 2.6 box isn't where the
> script runs.
> 
> The first command ALSO works if I use the "--blocking-io" option,
> 
> 2.8box%rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/
> --blocking-io
> 
> but I'm not sure what the possible side-effects of that might be.  (I
picked
> up from the CVS archive that there was some kind of relationship between a
> "socketpair" Solaris bug and the "--blocking-io"
option, so tried it, but
> that seemed many versions ago).
> 
> I am not doing the builds myself - sysadmins are doing it.  Is there
> anything special they
> need to do in the 2.6 build to avoid my having to use the
"--blocking-io"
> flag?
> 
> Thanks,
> 
> Dave Madole
> -- 
> To unsubscribe or change options:
http://lists.samba.org/mailman/listinfo/rsync
> Before posting, read: http://www.tuxedo.org/~esr/faqs/smart-questions.html

Thanks!

In other words, it would be using --blocking-io anyway, and I want it too 
use --blocking-io, so the worst the --blocking-io option would be is
redundant if the *SERVER* box was compiled with RSYNC_RSH=rsh (which, in my
case, it wasn't).  The --blocking-io option can't hurt and would be the
default if the builds were consistent and should have no side-effects.

This all seems silly, given that on all the boxes the "remsh"s are
just
links to the "rsh"s, but that's Solaris' silliness, not
rsync's.

Thanks again,

Dave
> -----Original Message-----
> From: Dave Dykstra [mailto:dwd@drdykstra.us]
> Sent: Thursday, December 05, 2002 12:28 PM
> To: Madole, Dave BGI SF
> Cc: 'rsync@lists.samba.org'
> Subject: Re: rsync 2.5.5, "unexpected tag failures", Solaris 2.6
vs.
> 2.8, "--b locking-io" workaround
> 
> 
> The problem is in your use of "-e rsh"; --blocking-io is assumed
if
> the -e value is equal to the RSYNC_RSH define which is usually
"remsh"
> on solaris but maybe it isn't on solaris 2.8.  The 
> --blocking-io option
> is required for most versions of rsh, but it's sometimes difficult for
> rsync to know when rsh is being used.  It's very confusing, I know.
> 
> - Dave
> 
> On Tue, Dec 03, 2002 at 10:42:19AM -0800, Madole, Dave BGI SF wrote:
> > 
> > Hi,
> > 
> > I have been having a problem that seems related to 
> previously reported
> > problems
> > but persists.
> > 
> > I am syncing from a 2.6 Solaris box to a 2.8 solaris box.  
> All are running
> > rsync 2.5.5.
> > 
> > When I sync (commands simplified for example, always using rsh):
> > 
> > 2.8box% rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/
> > 
> > i.e., the 2.6 box is the server, I get spurious, persistent 
> but inconsistent
> > "unexpected tag <some number>" failures.  This is, 
> unfortunately, on the 2.8
> > box, the way I have to do it.
> > 
> > When I sync:
> > 
> > 2.6box% rsync -e rsh /path/path/stuff/ 2.8box:/path/path/newstuff/
> > 
> > (the 2.8box is the server), it works.  But the 2.6 box 
> isn't where the
> > script runs.
> > 
> > The first command ALSO works if I use the "--blocking-io"
option,
> > 
> > 2.8box%rsync -e rsh 2.6box:/path/path/stuff/ /path/path/newstuff/
> > --blocking-io
> > 
> > but I'm not sure what the possible side-effects of that 
> might be.  (I picked
> > up from the CVS archive that there was some kind of 
> relationship between a
> > "socketpair" Solaris bug and the "--blocking-io"
option, so
> tried it, but
> > that seemed many versions ago).
> > 
> > I am not doing the builds myself - sysadmins are doing it.  Is there
> > anything special they
> > need to do in the 2.6 build to avoid my having to use the 
> "--blocking-io"
> > flag?
> > 
> > Thanks,
> > 
> > Dave Madole
> > -- 
> > To unsubscribe or change options: 
> http://lists.samba.org/mailman/listinfo/rsync
> > Before posting, read: 
> http://www.tuxedo.org/~esr/faqs/smart-questions.html
>