Wayne Davison wrote:
On Thu, Sep 08, 2005 at 01:39:39PM +1000, Mark Inkley wrote:
You will note, that I am using 2.6.6, previously I used 2.6.2 which
had the same issues.
Is the version on each machine 2.6.6?
Yes, I am using the -rsync-path switch to target the 2.6.6 binary. Our
binaries are on AFS so this is pretty easy to do.
If the remote version is older
than 2.6.4 then rsync cannot use its keep-alive protocol to try to keep
the connection from timing out during lulls in the sending of data (such
as when one side is busy deleting files). (Also, 2.6.4 only has partial
keep-alive support.)
If both machines are running 2.6.6, it is still possible for the session
to timeout if the keep-alive logic doesn't get called often enough for
the speed of your system and the largeness of the files involved (e.g. a
local copy of a large file that can occur when backing up deleted files
can slow down the checking). In such situations, a workaround is to
slightly increase the timeout value to avoid the accidental triggering.
I must admit, playing with the timeouts did not produce the desired
results. I have found a vast improvement by using the -delete-during
option, and now we have a much higher sucess rate of jobs completing
(100% so far) and it seems to take less time to run. I would have
liked to use the delete-after switch only because in some cases this
reduces process time on the source and makes for a shorter work-window
on the source. However the timeout problem still exists.
Yes, the keep-alive logic in rsync could be improved to help it avoid
such extended lull sections, so if someone wants to look into this,
feel free.
..wayne..
|
--
To unsubscribe or change options: https://lists.samba.org/mailman/listinfo/rsync
Before posting, read: http://www.catb.org/~esr/faqs/smart-questions.html
