Rsync forwarding using Balance/Jumpgate hanging on large directories

[Bachmeier, Marc]

I've got a puzzling issue, thought I'd ask to see if anyone else has had
a similar problem.

We have two machines that are synchronizing directories one way - the
data files are being pushed from the source to the destination. On the
destination system, I have multiple (6) modules set up to receive data
based on type. The source system is running rsync 2.6.2, the destination
is running 2.6.3rc1.

Between the two hosts is a system running jumpgate in order to forward
the rsync TCP packets from the source to the destination:

Source - rsync TVP port 873 -> relay (jumpgate) -> TCP 873 ->
destination

The relay is in order to comply with internal security requirements.

The data that is being sent is performance data from network nodes,
which is collected and stored on the source. This results in lots (some
directories have 800K+ files in them) of small files - don't ask why,
it's vendor software that mandates the directory structure.....

All the rsync pushes work perfectly going through the jumpgate relay
except one - the one that is failing will start the rsync process, pass
the filelists back and forth, do the checksum matches, and then start to
transfer files. After in transfers anywhere from 2-50 files, the rsync
hangs with no error message. The receiving end still has a sub-process
froked from the master daemon, but no data is being passed.

I've gone through all the logs, debugging, -vvvv and I'm stumped as to
why one works and another doesn't. Thought I'd see if another set of
brains would help.

Command used to send data from source server is:

usr/local/bin/rsync -vptr /lab/data/vendor/
rsync://destination.host.attws.com/module1

Any suggestions appreciated

Marc

Marc Bachmeier
Systems Engineer
NSS Platform Engineering
d) 425.702.3600
c) 206.618.3602
mailto: marc.bachmeier AT attws.com